diff --git a/tools/explorer/test/models/llama_attention_no_rot_emb_ttir.mlir b/tools/explorer/test/models/llama_attention_no_rot_emb_ttir.mlir new file mode 100644 index 000000000..0c1cea1db --- /dev/null +++ b/tools/explorer/test/models/llama_attention_no_rot_emb_ttir.mlir @@ -0,0 +1,120 @@ +#any_device = #tt.operand_constraint +#loc = loc("SelfAttention":0:0) +module @SelfAttention attributes {} { + func.func @forward(%arg0: tensor<1x12x3200xf32> {ttir.name = "hidden_states_1"} loc("SelfAttention":0:0), %arg1: tensor<1x1x12x12xf32> {ttir.name = "attention_mask"} loc("SelfAttention":0:0), %arg2: tensor<1xf32> {ttir.name = "input_1_multiply_20"} loc("SelfAttention":0:0), %arg3: tensor<3200x3200xf32> {ttir.name = "model.q_proj.weight"} loc("SelfAttention":0:0), %arg4: tensor<3200x3200xf32> {ttir.name = "model.k_proj.weight"} loc("SelfAttention":0:0), %arg5: tensor<3200x3200xf32> {ttir.name = "model.v_proj.weight"} loc("SelfAttention":0:0), %arg6: tensor<3200x3200xf32> {ttir.name = "model.o_proj.weight"} loc("SelfAttention":0:0)) -> (tensor<1x12x3200xf32> {ttir.name = "SelfAttention.output_reshape_38"}) { + %0 = tensor.empty() : tensor<12x3200xf32> loc(#loc30) + %1 = "ttir.squeeze"(%arg0, %0) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc30) + %2 = tensor.empty() : tensor<12x3200xf32> loc(#loc31) + %3 = "ttir.matmul"(%1, %arg3, %2) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc31) + %4 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc32) + %5 = "ttir.reshape"(%3, %4) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc32) + %6 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc33) + %7 = "ttir.transpose"(%5, %6) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc33) + %8 = tensor.empty() : tensor<32x12x100xf32> loc(#loc34) + %9 = "ttir.squeeze"(%7, %8) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc34) + %10 = tensor.empty() : tensor<12x3200xf32> loc(#loc35) + %11 = "ttir.matmul"(%1, %arg4, %10) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc35) + %12 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc36) + %13 = "ttir.reshape"(%11, %12) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc36) + %14 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc37) + %15 = "ttir.transpose"(%13, %14) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc37) + %16 = tensor.empty() : tensor<32x12x100xf32> loc(#loc38) + %17 = "ttir.squeeze"(%15, %16) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc38) + %18 = tensor.empty() : tensor<32x100x12xf32> loc(#loc39) + %19 = "ttir.transpose"(%17, %18) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc39) + %20 = tensor.empty() : tensor<32x12x12xf32> loc(#loc40) + %21 = "ttir.matmul"(%9, %19, %20) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc40) + %22 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc41) + %23 = "ttir.unsqueeze"(%21, %22) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc41) + %24 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc42) + %25 = "ttir.multiply"(%23, %arg2, %24) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc42) + %26 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc43) + %27 = "ttir.add"(%25, %arg1, %26) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc43) + %28 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc44) + %29 = "ttir.softmax"(%27, %28) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc44) + %30 = tensor.empty() : tensor<32x12x12xf32> loc(#loc45) + %31 = "ttir.squeeze"(%29, %30) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc45) + %32 = tensor.empty() : tensor<12x3200xf32> loc(#loc46) + %33 = "ttir.matmul"(%1, %arg5, %32) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc46) + %34 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc47) + %35 = "ttir.reshape"(%33, %34) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc47) + %36 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc48) + %37 = "ttir.transpose"(%35, %36) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc48) + %38 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc49) + %39 = "ttir.transpose"(%37, %38) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc49) + %40 = tensor.empty() : tensor<32x100x12xf32> loc(#loc50) + %41 = "ttir.squeeze"(%39, %40) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc50) + %42 = tensor.empty() : tensor<32x12x100xf32> loc(#loc51) + %43 = "ttir.transpose"(%41, %42) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc51) + %44 = tensor.empty() : tensor<32x12x100xf32> loc(#loc52) + %45 = "ttir.matmul"(%31, %43, %44) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc52) + %46 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc53) + %47 = "ttir.unsqueeze"(%45, %46) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc53) + %48 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc54) + %49 = "ttir.transpose"(%47, %48) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc54) + %50 = tensor.empty() : tensor<12x3200xf32> loc(#loc55) + %51 = "ttir.reshape"(%49, %50) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc55) + %52 = tensor.empty() : tensor<12x3200xf32> loc(#loc56) + %53 = "ttir.matmul"(%51, %arg6, %52) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc56) + %54 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc57) + %55 = "ttir.unsqueeze"(%53, %54) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc57) + return %55 : tensor<1x12x3200xf32> loc(#loc29) + } loc(#loc) +} loc(#loc) +#loc1 = loc("forward":4294967295:63) +#loc2 = loc("forward":4294967295:65) +#loc3 = loc("forward":4294967295:66) +#loc4 = loc("forward":4294967295:67) +#loc5 = loc("forward":4294967295:68) +#loc6 = loc("forward":4294967295:70) +#loc7 = loc("forward":4294967295:71) +#loc8 = loc("forward":4294967295:72) +#loc9 = loc("forward":4294967295:73) +#loc10 = loc("forward":4294967295:74) +#loc11 = loc("forward":4294967295:75) +#loc12 = loc("forward":4294967295:76) +#loc13 = loc("forward":4294967295:78) +#loc14 = loc("forward":4294967295:80) +#loc15 = loc("forward":4294967295:81) +#loc16 = loc("forward":4294967295:82) +#loc17 = loc("forward":4294967295:84) +#loc18 = loc("forward":4294967295:85) +#loc19 = loc("forward":4294967295:86) +#loc20 = loc("forward":4294967295:87) +#loc21 = loc("forward":4294967295:88) +#loc22 = loc("forward":4294967295:89) +#loc23 = loc("forward":4294967295:90) +#loc24 = loc("forward":4294967295:91) +#loc25 = loc("forward":4294967295:92) +#loc26 = loc("forward":4294967295:93) +#loc27 = loc("forward":4294967295:95) +#loc28 = loc("forward":4294967295:96) +#loc29 = loc(unknown) +#loc30 = loc("reshape_6.dc.squeeze.0"(#loc1)) +#loc31 = loc("matmul_8"(#loc2)) +#loc32 = loc("reshape_9"(#loc3)) +#loc33 = loc("transpose_10"(#loc4)) +#loc34 = loc("reshape_11.dc.squeeze.0"(#loc5)) +#loc35 = loc("matmul_13"(#loc6)) +#loc36 = loc("reshape_14"(#loc7)) +#loc37 = loc("transpose_15"(#loc8)) +#loc38 = loc("reshape_16.dc.squeeze.0"(#loc9)) +#loc39 = loc("transpose_17"(#loc10)) +#loc40 = loc("matmul_18"(#loc11)) +#loc41 = loc("reshape_19.dc.unsqueeze.0"(#loc12)) +#loc42 = loc("multiply_20"(#loc13)) +#loc43 = loc("add_21"(#loc14)) +#loc44 = loc("softmax_22"(#loc15)) +#loc45 = loc("reshape_24.dc.squeeze.0"(#loc16)) +#loc46 = loc("matmul_26"(#loc17)) +#loc47 = loc("reshape_27"(#loc18)) +#loc48 = loc("transpose_28"(#loc19)) +#loc49 = loc("transpose_29"(#loc20)) +#loc50 = loc("reshape_30.dc.squeeze.0"(#loc21)) +#loc51 = loc("transpose_31"(#loc22)) +#loc52 = loc("matmul_32"(#loc23)) +#loc53 = loc("reshape_33.dc.unsqueeze.0"(#loc24)) +#loc54 = loc("transpose_34"(#loc25)) +#loc55 = loc("reshape_35"(#loc26)) +#loc56 = loc("matmul_37"(#loc27)) +#loc57 = loc("reshape_38.dc.unsqueeze.0"(#loc28)) diff --git a/tools/explorer/test/models/open_llama_3b_single_layer.mlir b/tools/explorer/test/models/open_llama_3b_single_layer.mlir index 5e17dc39e..677aeb3c7 100644 --- a/tools/explorer/test/models/open_llama_3b_single_layer.mlir +++ b/tools/explorer/test/models/open_llama_3b_single_layer.mlir @@ -1,7 +1,6 @@ #any_device = #tt.operand_constraint #loc = loc("LlamaForCausalLM":0:0) -#system_desc = #tt.system_desc<[{role = host, target_triple = "x86_64-pc-linux-gnu"}], [{arch = , grid = 8x8, l1_size = 1499136, num_dram_channels = 12, dram_channel_size = 1073741824, noc_l1_address_align_bytes = 16, pcie_address_align_bytes = 32, noc_dram_address_align_bytes = 32, l1_unreserved_base = 1024, erisc_l1_unreserved_base = 1024, dram_unreserved_base = 1024, dram_unreserved_end = 1073741824, physical_cores = {worker = [ 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7, 1x0, 1x1, 1x2, 1x3, 1x4, 1x5, 1x6, 1x7, 2x0, 2x1, 2x2, 2x3, 2x4, 2x5, 2x6, 2x7, 3x0, 3x1, 3x2, 3x3, 3x4, 3x5, 3x6, 3x7, 4x0, 4x1, 4x2, 4x3, 4x4, 4x5, 4x6, 4x7, 5x0, 5x1, 5x2, 5x3, 5x4, 5x5, 5x6, 5x7, 6x0, 6x1, 6x2, 6x3, 6x4, 6x5, 6x6, 6x7, 7x0, 7x1, 7x2, 7x3, 7x4, 7x5, 7x6, 7x7] dram = [ 8x0, 9x0, 10x0, 8x1, 9x1, 10x1, 8x2, 9x2, 10x2, 8x3, 9x3, 10x3]}, supported_data_types = [, , , , , , , , , , , ], supported_tile_sizes = [ 4x16, 16x16, 32x16, 4x32, 16x32, 32x32], num_cbs = 32}], [0], [3 : i32], [ 0x0x0x0]> -module @LlamaForCausalLM attributes {tt.system_desc = #system_desc} { +module @LlamaForCausalLM attributes {} { func.func @forward(%arg0: tensor<1x12xi32> {ttir.name = "input_1"} loc("LlamaForCausalLM":0:0), %arg1: tensor<1xf32> {ttir.name = "input_1_add_4"} loc("LlamaForCausalLM":0:0), %arg2: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_14"} loc("LlamaForCausalLM":0:0), %arg3: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_24.1"} loc("LlamaForCausalLM":0:0), %arg4: tensor<1xf32> {ttir.name = "input_1_multiply_25"} loc("LlamaForCausalLM":0:0), %arg5: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_26.1"} loc("LlamaForCausalLM":0:0), %arg6: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_38.1"} loc("LlamaForCausalLM":0:0), %arg7: tensor<1xf32> {ttir.name = "input_1_multiply_39"} loc("LlamaForCausalLM":0:0), %arg8: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_40.1"} loc("LlamaForCausalLM":0:0), %arg9: tensor<1xf32> {ttir.name = "input_1_multiply_48"} loc("LlamaForCausalLM":0:0), %arg10: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_49"} loc("LlamaForCausalLM":0:0), %arg11: tensor<1xf32> {ttir.name = "input_1_add_70"} loc("LlamaForCausalLM":0:0), %arg12: tensor<1xf32> {ttir.name = "input_1_add_90"} loc("LlamaForCausalLM":0:0), %arg13: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_100"} loc("LlamaForCausalLM":0:0), %arg14: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_110.1"} loc("LlamaForCausalLM":0:0), %arg15: tensor<1xf32> {ttir.name = "input_1_multiply_111"} loc("LlamaForCausalLM":0:0), %arg16: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_112.1"} loc("LlamaForCausalLM":0:0), %arg17: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_124.1"} loc("LlamaForCausalLM":0:0), %arg18: tensor<1xf32> {ttir.name = "input_1_multiply_125"} loc("LlamaForCausalLM":0:0), %arg19: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_126.1"} loc("LlamaForCausalLM":0:0), %arg20: tensor<1xf32> {ttir.name = "input_1_multiply_134"} loc("LlamaForCausalLM":0:0), %arg21: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_135"} loc("LlamaForCausalLM":0:0), %arg22: tensor<1xf32> {ttir.name = "input_1_add_156"} loc("LlamaForCausalLM":0:0), %arg23: tensor<1xf32> {ttir.name = "input_1_add_176"} loc("LlamaForCausalLM":0:0), %arg24: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_186"} loc("LlamaForCausalLM":0:0), %arg25: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_196.1"} loc("LlamaForCausalLM":0:0), %arg26: tensor<1xf32> {ttir.name = "input_1_multiply_197"} loc("LlamaForCausalLM":0:0), %arg27: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_198.1"} loc("LlamaForCausalLM":0:0), %arg28: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_210.1"} loc("LlamaForCausalLM":0:0), %arg29: tensor<1xf32> {ttir.name = "input_1_multiply_211"} loc("LlamaForCausalLM":0:0), %arg30: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_212.1"} loc("LlamaForCausalLM":0:0), %arg31: tensor<1xf32> {ttir.name = "input_1_multiply_220"} loc("LlamaForCausalLM":0:0), %arg32: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_221"} loc("LlamaForCausalLM":0:0), %arg33: tensor<1xf32> {ttir.name = "input_1_add_242"} loc("LlamaForCausalLM":0:0), %arg34: tensor<1xf32> {ttir.name = "input_1_add_262"} loc("LlamaForCausalLM":0:0), %arg35: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_272"} loc("LlamaForCausalLM":0:0), %arg36: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_282.1"} loc("LlamaForCausalLM":0:0), %arg37: tensor<1xf32> {ttir.name = "input_1_multiply_283"} loc("LlamaForCausalLM":0:0), %arg38: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_284.1"} loc("LlamaForCausalLM":0:0), %arg39: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_296.1"} loc("LlamaForCausalLM":0:0), %arg40: tensor<1xf32> {ttir.name = "input_1_multiply_297"} loc("LlamaForCausalLM":0:0), %arg41: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_298.1"} loc("LlamaForCausalLM":0:0), %arg42: tensor<1xf32> {ttir.name = "input_1_multiply_306"} loc("LlamaForCausalLM":0:0), %arg43: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_307"} loc("LlamaForCausalLM":0:0), %arg44: tensor<1xf32> {ttir.name = "input_1_add_328"} loc("LlamaForCausalLM":0:0), %arg45: tensor<1xf32> {ttir.name = "input_1_add_348"} loc("LlamaForCausalLM":0:0), %arg46: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_358"} loc("LlamaForCausalLM":0:0), %arg47: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_368.1"} loc("LlamaForCausalLM":0:0), %arg48: tensor<1xf32> {ttir.name = "input_1_multiply_369"} loc("LlamaForCausalLM":0:0), %arg49: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_370.1"} loc("LlamaForCausalLM":0:0), %arg50: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_382.1"} loc("LlamaForCausalLM":0:0), %arg51: tensor<1xf32> {ttir.name = "input_1_multiply_383"} loc("LlamaForCausalLM":0:0), %arg52: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_384.1"} loc("LlamaForCausalLM":0:0), %arg53: tensor<1xf32> {ttir.name = "input_1_multiply_392"} loc("LlamaForCausalLM":0:0), %arg54: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_393"} loc("LlamaForCausalLM":0:0), %arg55: tensor<1xf32> {ttir.name = "input_1_add_414"} loc("LlamaForCausalLM":0:0), %arg56: tensor<1xf32> {ttir.name = "input_1_add_434"} loc("LlamaForCausalLM":0:0), %arg57: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_444"} loc("LlamaForCausalLM":0:0), %arg58: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_454.1"} loc("LlamaForCausalLM":0:0), %arg59: tensor<1xf32> {ttir.name = "input_1_multiply_455"} loc("LlamaForCausalLM":0:0), %arg60: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_456.1"} loc("LlamaForCausalLM":0:0), %arg61: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_468.1"} loc("LlamaForCausalLM":0:0), %arg62: tensor<1xf32> {ttir.name = "input_1_multiply_469"} loc("LlamaForCausalLM":0:0), %arg63: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_470.1"} loc("LlamaForCausalLM":0:0), %arg64: tensor<1xf32> {ttir.name = "input_1_multiply_478"} loc("LlamaForCausalLM":0:0), %arg65: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_479"} loc("LlamaForCausalLM":0:0), %arg66: tensor<1xf32> {ttir.name = "input_1_add_500"} loc("LlamaForCausalLM":0:0), %arg67: tensor<1xf32> {ttir.name = "input_1_add_520"} loc("LlamaForCausalLM":0:0), %arg68: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_530"} loc("LlamaForCausalLM":0:0), %arg69: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_540.1"} loc("LlamaForCausalLM":0:0), %arg70: tensor<1xf32> {ttir.name = "input_1_multiply_541"} loc("LlamaForCausalLM":0:0), %arg71: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_542.1"} loc("LlamaForCausalLM":0:0), %arg72: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_554.1"} loc("LlamaForCausalLM":0:0), %arg73: tensor<1xf32> {ttir.name = "input_1_multiply_555"} loc("LlamaForCausalLM":0:0), %arg74: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_556.1"} loc("LlamaForCausalLM":0:0), %arg75: tensor<1xf32> {ttir.name = "input_1_multiply_564"} loc("LlamaForCausalLM":0:0), %arg76: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_565"} loc("LlamaForCausalLM":0:0), %arg77: tensor<1xf32> {ttir.name = "input_1_add_586"} loc("LlamaForCausalLM":0:0), %arg78: tensor<1xf32> {ttir.name = "input_1_add_606"} loc("LlamaForCausalLM":0:0), %arg79: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_616"} loc("LlamaForCausalLM":0:0), %arg80: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_626.1"} loc("LlamaForCausalLM":0:0), %arg81: tensor<1xf32> {ttir.name = "input_1_multiply_627"} loc("LlamaForCausalLM":0:0), %arg82: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_628.1"} loc("LlamaForCausalLM":0:0), %arg83: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_640.1"} loc("LlamaForCausalLM":0:0), %arg84: tensor<1xf32> {ttir.name = "input_1_multiply_641"} loc("LlamaForCausalLM":0:0), %arg85: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_642.1"} loc("LlamaForCausalLM":0:0), %arg86: tensor<1xf32> {ttir.name = "input_1_multiply_650"} loc("LlamaForCausalLM":0:0), %arg87: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_651"} loc("LlamaForCausalLM":0:0), %arg88: tensor<1xf32> {ttir.name = "input_1_add_672"} loc("LlamaForCausalLM":0:0), %arg89: tensor<1xf32> {ttir.name = "input_1_add_692"} loc("LlamaForCausalLM":0:0), %arg90: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_702"} loc("LlamaForCausalLM":0:0), %arg91: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_712.1"} loc("LlamaForCausalLM":0:0), %arg92: tensor<1xf32> {ttir.name = "input_1_multiply_713"} loc("LlamaForCausalLM":0:0), %arg93: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_714.1"} loc("LlamaForCausalLM":0:0), %arg94: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_726.1"} loc("LlamaForCausalLM":0:0), %arg95: tensor<1xf32> {ttir.name = "input_1_multiply_727"} loc("LlamaForCausalLM":0:0), %arg96: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_728.1"} loc("LlamaForCausalLM":0:0), %arg97: tensor<1xf32> {ttir.name = "input_1_multiply_736"} loc("LlamaForCausalLM":0:0), %arg98: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_737"} loc("LlamaForCausalLM":0:0), %arg99: tensor<1xf32> {ttir.name = "input_1_add_758"} loc("LlamaForCausalLM":0:0), %arg100: tensor<1xf32> {ttir.name = "input_1_add_778"} loc("LlamaForCausalLM":0:0), %arg101: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_788"} loc("LlamaForCausalLM":0:0), %arg102: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_798.1"} loc("LlamaForCausalLM":0:0), %arg103: tensor<1xf32> {ttir.name = "input_1_multiply_799"} loc("LlamaForCausalLM":0:0), %arg104: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_800.1"} loc("LlamaForCausalLM":0:0), %arg105: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_812.1"} loc("LlamaForCausalLM":0:0), %arg106: tensor<1xf32> {ttir.name = "input_1_multiply_813"} loc("LlamaForCausalLM":0:0), %arg107: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_814.1"} loc("LlamaForCausalLM":0:0), %arg108: tensor<1xf32> {ttir.name = "input_1_multiply_822"} loc("LlamaForCausalLM":0:0), %arg109: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_823"} loc("LlamaForCausalLM":0:0), %arg110: tensor<1xf32> {ttir.name = "input_1_add_844"} loc("LlamaForCausalLM":0:0), %arg111: tensor<1xf32> {ttir.name = "input_1_add_864"} loc("LlamaForCausalLM":0:0), %arg112: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_874"} loc("LlamaForCausalLM":0:0), %arg113: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_884.1"} loc("LlamaForCausalLM":0:0), %arg114: tensor<1xf32> {ttir.name = "input_1_multiply_885"} loc("LlamaForCausalLM":0:0), %arg115: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_886.1"} loc("LlamaForCausalLM":0:0), %arg116: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_898.1"} loc("LlamaForCausalLM":0:0), %arg117: tensor<1xf32> {ttir.name = "input_1_multiply_899"} loc("LlamaForCausalLM":0:0), %arg118: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_900.1"} loc("LlamaForCausalLM":0:0), %arg119: tensor<1xf32> {ttir.name = "input_1_multiply_908"} loc("LlamaForCausalLM":0:0), %arg120: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_909"} loc("LlamaForCausalLM":0:0), %arg121: tensor<1xf32> {ttir.name = "input_1_add_930"} loc("LlamaForCausalLM":0:0), %arg122: tensor<1xf32> {ttir.name = "input_1_add_950"} loc("LlamaForCausalLM":0:0), %arg123: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_960"} loc("LlamaForCausalLM":0:0), %arg124: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_970.1"} loc("LlamaForCausalLM":0:0), %arg125: tensor<1xf32> {ttir.name = "input_1_multiply_971"} loc("LlamaForCausalLM":0:0), %arg126: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_972.1"} loc("LlamaForCausalLM":0:0), %arg127: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_984.1"} loc("LlamaForCausalLM":0:0), %arg128: tensor<1xf32> {ttir.name = "input_1_multiply_985"} loc("LlamaForCausalLM":0:0), %arg129: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_986.1"} loc("LlamaForCausalLM":0:0), %arg130: tensor<1xf32> {ttir.name = "input_1_multiply_994"} loc("LlamaForCausalLM":0:0), %arg131: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_995"} loc("LlamaForCausalLM":0:0), %arg132: tensor<1xf32> {ttir.name = "input_1_add_1016"} loc("LlamaForCausalLM":0:0), %arg133: tensor<1xf32> {ttir.name = "input_1_add_1036"} loc("LlamaForCausalLM":0:0), %arg134: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_1046"} loc("LlamaForCausalLM":0:0), %arg135: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1056.1"} loc("LlamaForCausalLM":0:0), %arg136: tensor<1xf32> {ttir.name = "input_1_multiply_1057"} loc("LlamaForCausalLM":0:0), %arg137: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1058.1"} loc("LlamaForCausalLM":0:0), %arg138: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1070.1"} loc("LlamaForCausalLM":0:0), %arg139: tensor<1xf32> {ttir.name = "input_1_multiply_1071"} loc("LlamaForCausalLM":0:0), %arg140: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1072.1"} loc("LlamaForCausalLM":0:0), %arg141: tensor<1xf32> {ttir.name = "input_1_multiply_1080"} loc("LlamaForCausalLM":0:0), %arg142: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_1081"} loc("LlamaForCausalLM":0:0), %arg143: tensor<1xf32> {ttir.name = "input_1_add_1102"} loc("LlamaForCausalLM":0:0), %arg144: tensor<1xf32> {ttir.name = "input_1_add_1122"} loc("LlamaForCausalLM":0:0), %arg145: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_1132"} loc("LlamaForCausalLM":0:0), %arg146: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1142.1"} loc("LlamaForCausalLM":0:0), %arg147: tensor<1xf32> {ttir.name = "input_1_multiply_1143"} loc("LlamaForCausalLM":0:0), %arg148: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1144.1"} loc("LlamaForCausalLM":0:0), %arg149: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1156.1"} loc("LlamaForCausalLM":0:0), %arg150: tensor<1xf32> {ttir.name = "input_1_multiply_1157"} loc("LlamaForCausalLM":0:0), %arg151: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1158.1"} loc("LlamaForCausalLM":0:0), %arg152: tensor<1xf32> {ttir.name = "input_1_multiply_1166"} loc("LlamaForCausalLM":0:0), %arg153: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_1167"} loc("LlamaForCausalLM":0:0), %arg154: tensor<1xf32> {ttir.name = "input_1_add_1188"} loc("LlamaForCausalLM":0:0), %arg155: tensor<1xf32> {ttir.name = "input_1_add_1208"} loc("LlamaForCausalLM":0:0), %arg156: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_1218"} loc("LlamaForCausalLM":0:0), %arg157: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1228.1"} loc("LlamaForCausalLM":0:0), %arg158: tensor<1xf32> {ttir.name = "input_1_multiply_1229"} loc("LlamaForCausalLM":0:0), %arg159: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1230.1"} loc("LlamaForCausalLM":0:0), %arg160: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1242.1"} loc("LlamaForCausalLM":0:0), %arg161: tensor<1xf32> {ttir.name = "input_1_multiply_1243"} loc("LlamaForCausalLM":0:0), %arg162: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1244.1"} loc("LlamaForCausalLM":0:0), %arg163: tensor<1xf32> {ttir.name = "input_1_multiply_1252"} loc("LlamaForCausalLM":0:0), %arg164: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_1253"} loc("LlamaForCausalLM":0:0), %arg165: tensor<1xf32> {ttir.name = "input_1_add_1274"} loc("LlamaForCausalLM":0:0), %arg166: tensor<1xf32> {ttir.name = "input_1_add_1294"} loc("LlamaForCausalLM":0:0), %arg167: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_1304"} loc("LlamaForCausalLM":0:0), %arg168: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1314.1"} loc("LlamaForCausalLM":0:0), %arg169: tensor<1xf32> {ttir.name = "input_1_multiply_1315"} loc("LlamaForCausalLM":0:0), %arg170: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1316.1"} loc("LlamaForCausalLM":0:0), %arg171: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1328.1"} loc("LlamaForCausalLM":0:0), %arg172: tensor<1xf32> {ttir.name = "input_1_multiply_1329"} loc("LlamaForCausalLM":0:0), %arg173: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1330.1"} loc("LlamaForCausalLM":0:0), %arg174: tensor<1xf32> {ttir.name = "input_1_multiply_1338"} loc("LlamaForCausalLM":0:0), %arg175: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_1339"} loc("LlamaForCausalLM":0:0), %arg176: tensor<1xf32> {ttir.name = "input_1_add_1360"} loc("LlamaForCausalLM":0:0), %arg177: tensor<1xf32> {ttir.name = "input_1_add_1380"} loc("LlamaForCausalLM":0:0), %arg178: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_1390"} loc("LlamaForCausalLM":0:0), %arg179: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1400.1"} loc("LlamaForCausalLM":0:0), %arg180: tensor<1xf32> {ttir.name = "input_1_multiply_1401"} loc("LlamaForCausalLM":0:0), %arg181: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1402.1"} loc("LlamaForCausalLM":0:0), %arg182: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1414.1"} loc("LlamaForCausalLM":0:0), %arg183: tensor<1xf32> {ttir.name = "input_1_multiply_1415"} loc("LlamaForCausalLM":0:0), %arg184: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1416.1"} loc("LlamaForCausalLM":0:0), %arg185: tensor<1xf32> {ttir.name = "input_1_multiply_1424"} loc("LlamaForCausalLM":0:0), %arg186: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_1425"} loc("LlamaForCausalLM":0:0), %arg187: tensor<1xf32> {ttir.name = "input_1_add_1446"} loc("LlamaForCausalLM":0:0), %arg188: tensor<1xf32> {ttir.name = "input_1_add_1466"} loc("LlamaForCausalLM":0:0), %arg189: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_1476"} loc("LlamaForCausalLM":0:0), %arg190: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1486.1"} loc("LlamaForCausalLM":0:0), %arg191: tensor<1xf32> {ttir.name = "input_1_multiply_1487"} loc("LlamaForCausalLM":0:0), %arg192: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1488.1"} loc("LlamaForCausalLM":0:0), %arg193: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1500.1"} loc("LlamaForCausalLM":0:0), %arg194: tensor<1xf32> {ttir.name = "input_1_multiply_1501"} loc("LlamaForCausalLM":0:0), %arg195: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1502.1"} loc("LlamaForCausalLM":0:0), %arg196: tensor<1xf32> {ttir.name = "input_1_multiply_1510"} loc("LlamaForCausalLM":0:0), %arg197: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_1511"} loc("LlamaForCausalLM":0:0), %arg198: tensor<1xf32> {ttir.name = "input_1_add_1532"} loc("LlamaForCausalLM":0:0), %arg199: tensor<1xf32> {ttir.name = "input_1_add_1552"} loc("LlamaForCausalLM":0:0), %arg200: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_1562"} loc("LlamaForCausalLM":0:0), %arg201: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1572.1"} loc("LlamaForCausalLM":0:0), %arg202: tensor<1xf32> {ttir.name = "input_1_multiply_1573"} loc("LlamaForCausalLM":0:0), %arg203: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1574.1"} loc("LlamaForCausalLM":0:0), %arg204: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1586.1"} loc("LlamaForCausalLM":0:0), %arg205: tensor<1xf32> {ttir.name = "input_1_multiply_1587"} loc("LlamaForCausalLM":0:0), %arg206: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1588.1"} loc("LlamaForCausalLM":0:0), %arg207: tensor<1xf32> {ttir.name = "input_1_multiply_1596"} loc("LlamaForCausalLM":0:0), %arg208: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_1597"} loc("LlamaForCausalLM":0:0), %arg209: tensor<1xf32> {ttir.name = "input_1_add_1618"} loc("LlamaForCausalLM":0:0), %arg210: tensor<1xf32> {ttir.name = "input_1_add_1638"} loc("LlamaForCausalLM":0:0), %arg211: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_1648"} loc("LlamaForCausalLM":0:0), %arg212: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1658.1"} loc("LlamaForCausalLM":0:0), %arg213: tensor<1xf32> {ttir.name = "input_1_multiply_1659"} loc("LlamaForCausalLM":0:0), %arg214: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1660.1"} loc("LlamaForCausalLM":0:0), %arg215: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1672.1"} loc("LlamaForCausalLM":0:0), %arg216: tensor<1xf32> {ttir.name = "input_1_multiply_1673"} loc("LlamaForCausalLM":0:0), %arg217: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1674.1"} loc("LlamaForCausalLM":0:0), %arg218: tensor<1xf32> {ttir.name = "input_1_multiply_1682"} loc("LlamaForCausalLM":0:0), %arg219: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_1683"} loc("LlamaForCausalLM":0:0), %arg220: tensor<1xf32> {ttir.name = "input_1_add_1704"} loc("LlamaForCausalLM":0:0), %arg221: tensor<1xf32> {ttir.name = "input_1_add_1724"} loc("LlamaForCausalLM":0:0), %arg222: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_1734"} loc("LlamaForCausalLM":0:0), %arg223: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1744.1"} loc("LlamaForCausalLM":0:0), %arg224: tensor<1xf32> {ttir.name = "input_1_multiply_1745"} loc("LlamaForCausalLM":0:0), %arg225: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1746.1"} loc("LlamaForCausalLM":0:0), %arg226: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1758.1"} loc("LlamaForCausalLM":0:0), %arg227: tensor<1xf32> {ttir.name = "input_1_multiply_1759"} loc("LlamaForCausalLM":0:0), %arg228: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1760.1"} loc("LlamaForCausalLM":0:0), %arg229: tensor<1xf32> {ttir.name = "input_1_multiply_1768"} loc("LlamaForCausalLM":0:0), %arg230: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_1769"} loc("LlamaForCausalLM":0:0), %arg231: tensor<1xf32> {ttir.name = "input_1_add_1790"} loc("LlamaForCausalLM":0:0), %arg232: tensor<1xf32> {ttir.name = "input_1_add_1810"} loc("LlamaForCausalLM":0:0), %arg233: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_1820"} loc("LlamaForCausalLM":0:0), %arg234: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1830.1"} loc("LlamaForCausalLM":0:0), %arg235: tensor<1xf32> {ttir.name = "input_1_multiply_1831"} loc("LlamaForCausalLM":0:0), %arg236: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1832.1"} loc("LlamaForCausalLM":0:0), %arg237: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1844.1"} loc("LlamaForCausalLM":0:0), %arg238: tensor<1xf32> {ttir.name = "input_1_multiply_1845"} loc("LlamaForCausalLM":0:0), %arg239: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1846.1"} loc("LlamaForCausalLM":0:0), %arg240: tensor<1xf32> {ttir.name = "input_1_multiply_1854"} loc("LlamaForCausalLM":0:0), %arg241: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_1855"} loc("LlamaForCausalLM":0:0), %arg242: tensor<1xf32> {ttir.name = "input_1_add_1876"} loc("LlamaForCausalLM":0:0), %arg243: tensor<1xf32> {ttir.name = "input_1_add_1896"} loc("LlamaForCausalLM":0:0), %arg244: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_1906"} loc("LlamaForCausalLM":0:0), %arg245: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1916.1"} loc("LlamaForCausalLM":0:0), %arg246: tensor<1xf32> {ttir.name = "input_1_multiply_1917"} loc("LlamaForCausalLM":0:0), %arg247: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1918.1"} loc("LlamaForCausalLM":0:0), %arg248: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1930.1"} loc("LlamaForCausalLM":0:0), %arg249: tensor<1xf32> {ttir.name = "input_1_multiply_1931"} loc("LlamaForCausalLM":0:0), %arg250: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1932.1"} loc("LlamaForCausalLM":0:0), %arg251: tensor<1xf32> {ttir.name = "input_1_multiply_1940"} loc("LlamaForCausalLM":0:0), %arg252: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_1941"} loc("LlamaForCausalLM":0:0), %arg253: tensor<1xf32> {ttir.name = "input_1_add_1962"} loc("LlamaForCausalLM":0:0), %arg254: tensor<1xf32> {ttir.name = "input_1_add_1982"} loc("LlamaForCausalLM":0:0), %arg255: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_1992"} loc("LlamaForCausalLM":0:0), %arg256: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_2002.1"} loc("LlamaForCausalLM":0:0), %arg257: tensor<1xf32> {ttir.name = "input_1_multiply_2003"} loc("LlamaForCausalLM":0:0), %arg258: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_2004.1"} loc("LlamaForCausalLM":0:0), %arg259: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_2016.1"} loc("LlamaForCausalLM":0:0), %arg260: tensor<1xf32> {ttir.name = "input_1_multiply_2017"} loc("LlamaForCausalLM":0:0), %arg261: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_2018.1"} loc("LlamaForCausalLM":0:0), %arg262: tensor<1xf32> {ttir.name = "input_1_multiply_2026"} loc("LlamaForCausalLM":0:0), %arg263: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_2027"} loc("LlamaForCausalLM":0:0), %arg264: tensor<1xf32> {ttir.name = "input_1_add_2048"} loc("LlamaForCausalLM":0:0), %arg265: tensor<1xf32> {ttir.name = "input_1_add_2068"} loc("LlamaForCausalLM":0:0), %arg266: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_2078"} loc("LlamaForCausalLM":0:0), %arg267: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_2088.1"} loc("LlamaForCausalLM":0:0), %arg268: tensor<1xf32> {ttir.name = "input_1_multiply_2089"} loc("LlamaForCausalLM":0:0), %arg269: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_2090.1"} loc("LlamaForCausalLM":0:0), %arg270: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_2102.1"} loc("LlamaForCausalLM":0:0), %arg271: tensor<1xf32> {ttir.name = "input_1_multiply_2103"} loc("LlamaForCausalLM":0:0), %arg272: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_2104.1"} loc("LlamaForCausalLM":0:0), %arg273: tensor<1xf32> {ttir.name = "input_1_multiply_2112"} loc("LlamaForCausalLM":0:0), %arg274: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_2113"} loc("LlamaForCausalLM":0:0), %arg275: tensor<1xf32> {ttir.name = "input_1_add_2134"} loc("LlamaForCausalLM":0:0), %arg276: tensor<1xf32> {ttir.name = "input_1_add_2154"} loc("LlamaForCausalLM":0:0), %arg277: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_2164"} loc("LlamaForCausalLM":0:0), %arg278: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_2174.1"} loc("LlamaForCausalLM":0:0), %arg279: tensor<1xf32> {ttir.name = "input_1_multiply_2175"} loc("LlamaForCausalLM":0:0), %arg280: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_2176.1"} loc("LlamaForCausalLM":0:0), %arg281: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_2188.1"} loc("LlamaForCausalLM":0:0), %arg282: tensor<1xf32> {ttir.name = "input_1_multiply_2189"} loc("LlamaForCausalLM":0:0), %arg283: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_2190.1"} loc("LlamaForCausalLM":0:0), %arg284: tensor<1xf32> {ttir.name = "input_1_multiply_2198"} loc("LlamaForCausalLM":0:0), %arg285: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_2199"} loc("LlamaForCausalLM":0:0), %arg286: tensor<1xf32> {ttir.name = "input_1_add_2220"} loc("LlamaForCausalLM":0:0), %arg287: tensor<1xf32> {ttir.name = "input_1_add_2240"} loc("LlamaForCausalLM":0:0), %arg288: tensor<3200xf32> {ttir.name = "model.norm.weight"} loc("LlamaForCausalLM":0:0), %arg289: tensor<32000x3200xf32> {ttir.name = "model.embed_tokens.weight"} loc("LlamaForCausalLM":0:0), %arg290: tensor<3200xf32> {ttir.name = "model.layers.0.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg291: tensor<3200x3200xf32> {ttir.name = "model.layers.0.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg292: tensor<3200x3200xf32> {ttir.name = "model.layers.0.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg293: tensor<3200x3200xf32> {ttir.name = "model.layers.0.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg294: tensor<3200x3200xf32> {ttir.name = "model.layers.0.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg295: tensor<3200xf32> {ttir.name = "model.layers.0.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg296: tensor<3200x8640xf32> {ttir.name = "model.layers.0.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg297: tensor<3200x8640xf32> {ttir.name = "model.layers.0.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg298: tensor<8640x3200xf32> {ttir.name = "model.layers.0.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg299: tensor<3200xf32> {ttir.name = "model.layers.1.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg300: tensor<3200x3200xf32> {ttir.name = "model.layers.1.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg301: tensor<3200x3200xf32> {ttir.name = "model.layers.1.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg302: tensor<3200x3200xf32> {ttir.name = "model.layers.1.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg303: tensor<3200x3200xf32> {ttir.name = "model.layers.1.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg304: tensor<3200xf32> {ttir.name = "model.layers.1.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg305: tensor<3200x8640xf32> {ttir.name = "model.layers.1.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg306: tensor<3200x8640xf32> {ttir.name = "model.layers.1.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg307: tensor<8640x3200xf32> {ttir.name = "model.layers.1.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg308: tensor<3200xf32> {ttir.name = "model.layers.2.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg309: tensor<3200x3200xf32> {ttir.name = "model.layers.2.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg310: tensor<3200x3200xf32> {ttir.name = "model.layers.2.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg311: tensor<3200x3200xf32> {ttir.name = "model.layers.2.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg312: tensor<3200x3200xf32> {ttir.name = "model.layers.2.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg313: tensor<3200xf32> {ttir.name = "model.layers.2.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg314: tensor<3200x8640xf32> {ttir.name = "model.layers.2.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg315: tensor<3200x8640xf32> {ttir.name = "model.layers.2.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg316: tensor<8640x3200xf32> {ttir.name = "model.layers.2.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg317: tensor<3200xf32> {ttir.name = "model.layers.3.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg318: tensor<3200x3200xf32> {ttir.name = "model.layers.3.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg319: tensor<3200x3200xf32> {ttir.name = "model.layers.3.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg320: tensor<3200x3200xf32> {ttir.name = "model.layers.3.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg321: tensor<3200x3200xf32> {ttir.name = "model.layers.3.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg322: tensor<3200xf32> {ttir.name = "model.layers.3.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg323: tensor<3200x8640xf32> {ttir.name = "model.layers.3.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg324: tensor<3200x8640xf32> {ttir.name = "model.layers.3.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg325: tensor<8640x3200xf32> {ttir.name = "model.layers.3.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg326: tensor<3200xf32> {ttir.name = "model.layers.4.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg327: tensor<3200x3200xf32> {ttir.name = "model.layers.4.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg328: tensor<3200x3200xf32> {ttir.name = "model.layers.4.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg329: tensor<3200x3200xf32> {ttir.name = "model.layers.4.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg330: tensor<3200x3200xf32> {ttir.name = "model.layers.4.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg331: tensor<3200xf32> {ttir.name = "model.layers.4.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg332: tensor<3200x8640xf32> {ttir.name = "model.layers.4.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg333: tensor<3200x8640xf32> {ttir.name = "model.layers.4.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg334: tensor<8640x3200xf32> {ttir.name = "model.layers.4.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg335: tensor<3200xf32> {ttir.name = "model.layers.5.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg336: tensor<3200x3200xf32> {ttir.name = "model.layers.5.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg337: tensor<3200x3200xf32> {ttir.name = "model.layers.5.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg338: tensor<3200x3200xf32> {ttir.name = "model.layers.5.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg339: tensor<3200x3200xf32> {ttir.name = "model.layers.5.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg340: tensor<3200xf32> {ttir.name = "model.layers.5.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg341: tensor<3200x8640xf32> {ttir.name = "model.layers.5.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg342: tensor<3200x8640xf32> {ttir.name = "model.layers.5.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg343: tensor<8640x3200xf32> {ttir.name = "model.layers.5.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg344: tensor<3200xf32> {ttir.name = "model.layers.6.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg345: tensor<3200x3200xf32> {ttir.name = "model.layers.6.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg346: tensor<3200x3200xf32> {ttir.name = "model.layers.6.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg347: tensor<3200x3200xf32> {ttir.name = "model.layers.6.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg348: tensor<3200x3200xf32> {ttir.name = "model.layers.6.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg349: tensor<3200xf32> {ttir.name = "model.layers.6.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg350: tensor<3200x8640xf32> {ttir.name = "model.layers.6.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg351: tensor<3200x8640xf32> {ttir.name = "model.layers.6.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg352: tensor<8640x3200xf32> {ttir.name = "model.layers.6.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg353: tensor<3200xf32> {ttir.name = "model.layers.7.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg354: tensor<3200x3200xf32> {ttir.name = "model.layers.7.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg355: tensor<3200x3200xf32> {ttir.name = "model.layers.7.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg356: tensor<3200x3200xf32> {ttir.name = "model.layers.7.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg357: tensor<3200x3200xf32> {ttir.name = "model.layers.7.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg358: tensor<3200xf32> {ttir.name = "model.layers.7.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg359: tensor<3200x8640xf32> {ttir.name = "model.layers.7.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg360: tensor<3200x8640xf32> {ttir.name = "model.layers.7.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg361: tensor<8640x3200xf32> {ttir.name = "model.layers.7.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg362: tensor<3200xf32> {ttir.name = "model.layers.8.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg363: tensor<3200x3200xf32> {ttir.name = "model.layers.8.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg364: tensor<3200x3200xf32> {ttir.name = "model.layers.8.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg365: tensor<3200x3200xf32> {ttir.name = "model.layers.8.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg366: tensor<3200x3200xf32> {ttir.name = "model.layers.8.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg367: tensor<3200xf32> {ttir.name = "model.layers.8.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg368: tensor<3200x8640xf32> {ttir.name = "model.layers.8.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg369: tensor<3200x8640xf32> {ttir.name = "model.layers.8.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg370: tensor<8640x3200xf32> {ttir.name = "model.layers.8.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg371: tensor<3200xf32> {ttir.name = "model.layers.9.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg372: tensor<3200x3200xf32> {ttir.name = "model.layers.9.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg373: tensor<3200x3200xf32> {ttir.name = "model.layers.9.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg374: tensor<3200x3200xf32> {ttir.name = "model.layers.9.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg375: tensor<3200x3200xf32> {ttir.name = "model.layers.9.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg376: tensor<3200xf32> {ttir.name = "model.layers.9.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg377: tensor<3200x8640xf32> {ttir.name = "model.layers.9.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg378: tensor<3200x8640xf32> {ttir.name = "model.layers.9.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg379: tensor<8640x3200xf32> {ttir.name = "model.layers.9.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg380: tensor<3200xf32> {ttir.name = "model.layers.10.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg381: tensor<3200x3200xf32> {ttir.name = "model.layers.10.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg382: tensor<3200x3200xf32> {ttir.name = "model.layers.10.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg383: tensor<3200x3200xf32> {ttir.name = "model.layers.10.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg384: tensor<3200x3200xf32> {ttir.name = "model.layers.10.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg385: tensor<3200xf32> {ttir.name = "model.layers.10.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg386: tensor<3200x8640xf32> {ttir.name = "model.layers.10.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg387: tensor<3200x8640xf32> {ttir.name = "model.layers.10.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg388: tensor<8640x3200xf32> {ttir.name = "model.layers.10.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg389: tensor<3200xf32> {ttir.name = "model.layers.11.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg390: tensor<3200x3200xf32> {ttir.name = "model.layers.11.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg391: tensor<3200x3200xf32> {ttir.name = "model.layers.11.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg392: tensor<3200x3200xf32> {ttir.name = "model.layers.11.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg393: tensor<3200x3200xf32> {ttir.name = "model.layers.11.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg394: tensor<3200xf32> {ttir.name = "model.layers.11.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg395: tensor<3200x8640xf32> {ttir.name = "model.layers.11.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg396: tensor<3200x8640xf32> {ttir.name = "model.layers.11.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg397: tensor<8640x3200xf32> {ttir.name = "model.layers.11.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg398: tensor<3200xf32> {ttir.name = "model.layers.12.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg399: tensor<3200x3200xf32> {ttir.name = "model.layers.12.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg400: tensor<3200x3200xf32> {ttir.name = "model.layers.12.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg401: tensor<3200x3200xf32> {ttir.name = "model.layers.12.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg402: tensor<3200x3200xf32> {ttir.name = "model.layers.12.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg403: tensor<3200xf32> {ttir.name = "model.layers.12.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg404: tensor<3200x8640xf32> {ttir.name = "model.layers.12.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg405: tensor<3200x8640xf32> {ttir.name = "model.layers.12.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg406: tensor<8640x3200xf32> {ttir.name = "model.layers.12.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg407: tensor<3200xf32> {ttir.name = "model.layers.13.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg408: tensor<3200x3200xf32> {ttir.name = "model.layers.13.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg409: tensor<3200x3200xf32> {ttir.name = "model.layers.13.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg410: tensor<3200x3200xf32> {ttir.name = "model.layers.13.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg411: tensor<3200x3200xf32> {ttir.name = "model.layers.13.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg412: tensor<3200xf32> {ttir.name = "model.layers.13.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg413: tensor<3200x8640xf32> {ttir.name = "model.layers.13.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg414: tensor<3200x8640xf32> {ttir.name = "model.layers.13.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg415: tensor<8640x3200xf32> {ttir.name = "model.layers.13.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg416: tensor<3200xf32> {ttir.name = "model.layers.14.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg417: tensor<3200x3200xf32> {ttir.name = "model.layers.14.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg418: tensor<3200x3200xf32> {ttir.name = "model.layers.14.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg419: tensor<3200x3200xf32> {ttir.name = "model.layers.14.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg420: tensor<3200x3200xf32> {ttir.name = "model.layers.14.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg421: tensor<3200xf32> {ttir.name = "model.layers.14.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg422: tensor<3200x8640xf32> {ttir.name = "model.layers.14.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg423: tensor<3200x8640xf32> {ttir.name = "model.layers.14.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg424: tensor<8640x3200xf32> {ttir.name = "model.layers.14.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg425: tensor<3200xf32> {ttir.name = "model.layers.15.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg426: tensor<3200x3200xf32> {ttir.name = "model.layers.15.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg427: tensor<3200x3200xf32> {ttir.name = "model.layers.15.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg428: tensor<3200x3200xf32> {ttir.name = "model.layers.15.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg429: tensor<3200x3200xf32> {ttir.name = "model.layers.15.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg430: tensor<3200xf32> {ttir.name = "model.layers.15.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg431: tensor<3200x8640xf32> {ttir.name = "model.layers.15.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg432: tensor<3200x8640xf32> {ttir.name = "model.layers.15.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg433: tensor<8640x3200xf32> {ttir.name = "model.layers.15.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg434: tensor<3200xf32> {ttir.name = "model.layers.16.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg435: tensor<3200x3200xf32> {ttir.name = "model.layers.16.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg436: tensor<3200x3200xf32> {ttir.name = "model.layers.16.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg437: tensor<3200x3200xf32> {ttir.name = "model.layers.16.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg438: tensor<3200x3200xf32> {ttir.name = "model.layers.16.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg439: tensor<3200xf32> {ttir.name = "model.layers.16.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg440: tensor<3200x8640xf32> {ttir.name = "model.layers.16.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg441: tensor<3200x8640xf32> {ttir.name = "model.layers.16.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg442: tensor<8640x3200xf32> {ttir.name = "model.layers.16.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg443: tensor<3200xf32> {ttir.name = "model.layers.17.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg444: tensor<3200x3200xf32> {ttir.name = "model.layers.17.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg445: tensor<3200x3200xf32> {ttir.name = "model.layers.17.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg446: tensor<3200x3200xf32> {ttir.name = "model.layers.17.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg447: tensor<3200x3200xf32> {ttir.name = "model.layers.17.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg448: tensor<3200xf32> {ttir.name = "model.layers.17.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg449: tensor<3200x8640xf32> {ttir.name = "model.layers.17.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg450: tensor<3200x8640xf32> {ttir.name = "model.layers.17.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg451: tensor<8640x3200xf32> {ttir.name = "model.layers.17.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg452: tensor<3200xf32> {ttir.name = "model.layers.18.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg453: tensor<3200x3200xf32> {ttir.name = "model.layers.18.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg454: tensor<3200x3200xf32> {ttir.name = "model.layers.18.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg455: tensor<3200x3200xf32> {ttir.name = "model.layers.18.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg456: tensor<3200x3200xf32> {ttir.name = "model.layers.18.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg457: tensor<3200xf32> {ttir.name = "model.layers.18.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg458: tensor<3200x8640xf32> {ttir.name = "model.layers.18.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg459: tensor<3200x8640xf32> {ttir.name = "model.layers.18.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg460: tensor<8640x3200xf32> {ttir.name = "model.layers.18.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg461: tensor<3200xf32> {ttir.name = "model.layers.19.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg462: tensor<3200x3200xf32> {ttir.name = "model.layers.19.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg463: tensor<3200x3200xf32> {ttir.name = "model.layers.19.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg464: tensor<3200x3200xf32> {ttir.name = "model.layers.19.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg465: tensor<3200x3200xf32> {ttir.name = "model.layers.19.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg466: tensor<3200xf32> {ttir.name = "model.layers.19.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg467: tensor<3200x8640xf32> {ttir.name = "model.layers.19.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg468: tensor<3200x8640xf32> {ttir.name = "model.layers.19.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg469: tensor<8640x3200xf32> {ttir.name = "model.layers.19.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg470: tensor<3200xf32> {ttir.name = "model.layers.20.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg471: tensor<3200x3200xf32> {ttir.name = "model.layers.20.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg472: tensor<3200x3200xf32> {ttir.name = "model.layers.20.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg473: tensor<3200x3200xf32> {ttir.name = "model.layers.20.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg474: tensor<3200x3200xf32> {ttir.name = "model.layers.20.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg475: tensor<3200xf32> {ttir.name = "model.layers.20.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg476: tensor<3200x8640xf32> {ttir.name = "model.layers.20.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg477: tensor<3200x8640xf32> {ttir.name = "model.layers.20.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg478: tensor<8640x3200xf32> {ttir.name = "model.layers.20.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg479: tensor<3200xf32> {ttir.name = "model.layers.21.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg480: tensor<3200x3200xf32> {ttir.name = "model.layers.21.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg481: tensor<3200x3200xf32> {ttir.name = "model.layers.21.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg482: tensor<3200x3200xf32> {ttir.name = "model.layers.21.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg483: tensor<3200x3200xf32> {ttir.name = "model.layers.21.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg484: tensor<3200xf32> {ttir.name = "model.layers.21.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg485: tensor<3200x8640xf32> {ttir.name = "model.layers.21.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg486: tensor<3200x8640xf32> {ttir.name = "model.layers.21.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg487: tensor<8640x3200xf32> {ttir.name = "model.layers.21.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg488: tensor<3200xf32> {ttir.name = "model.layers.22.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg489: tensor<3200x3200xf32> {ttir.name = "model.layers.22.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg490: tensor<3200x3200xf32> {ttir.name = "model.layers.22.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg491: tensor<3200x3200xf32> {ttir.name = "model.layers.22.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg492: tensor<3200x3200xf32> {ttir.name = "model.layers.22.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg493: tensor<3200xf32> {ttir.name = "model.layers.22.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg494: tensor<3200x8640xf32> {ttir.name = "model.layers.22.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg495: tensor<3200x8640xf32> {ttir.name = "model.layers.22.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg496: tensor<8640x3200xf32> {ttir.name = "model.layers.22.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg497: tensor<3200xf32> {ttir.name = "model.layers.23.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg498: tensor<3200x3200xf32> {ttir.name = "model.layers.23.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg499: tensor<3200x3200xf32> {ttir.name = "model.layers.23.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg500: tensor<3200x3200xf32> {ttir.name = "model.layers.23.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg501: tensor<3200x3200xf32> {ttir.name = "model.layers.23.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg502: tensor<3200xf32> {ttir.name = "model.layers.23.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg503: tensor<3200x8640xf32> {ttir.name = "model.layers.23.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg504: tensor<3200x8640xf32> {ttir.name = "model.layers.23.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg505: tensor<8640x3200xf32> {ttir.name = "model.layers.23.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg506: tensor<3200xf32> {ttir.name = "model.layers.24.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg507: tensor<3200x3200xf32> {ttir.name = "model.layers.24.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg508: tensor<3200x3200xf32> {ttir.name = "model.layers.24.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg509: tensor<3200x3200xf32> {ttir.name = "model.layers.24.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg510: tensor<3200x3200xf32> {ttir.name = "model.layers.24.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg511: tensor<3200xf32> {ttir.name = "model.layers.24.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg512: tensor<3200x8640xf32> {ttir.name = "model.layers.24.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg513: tensor<3200x8640xf32> {ttir.name = "model.layers.24.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg514: tensor<8640x3200xf32> {ttir.name = "model.layers.24.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg515: tensor<3200xf32> {ttir.name = "model.layers.25.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg516: tensor<3200x3200xf32> {ttir.name = "model.layers.25.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg517: tensor<3200x3200xf32> {ttir.name = "model.layers.25.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg518: tensor<3200x3200xf32> {ttir.name = "model.layers.25.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg519: tensor<3200x3200xf32> {ttir.name = "model.layers.25.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg520: tensor<3200xf32> {ttir.name = "model.layers.25.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg521: tensor<3200x8640xf32> {ttir.name = "model.layers.25.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg522: tensor<3200x8640xf32> {ttir.name = "model.layers.25.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg523: tensor<8640x3200xf32> {ttir.name = "model.layers.25.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg524: tensor<3200x32000xf32> {ttir.name = "lm_head.weight"} loc("LlamaForCausalLM":0:0)) -> (tensor<1x12x3200xf32> {ttir.name = "LlamaForCausalLM.output_matmul_2246"}) { %0 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2091) %1 = "ttir.embedding"(%arg0, %arg289, %0) <{operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12xi32>, tensor<32000x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2091) diff --git a/tools/explorer/test/models/open_llama_3b_ttir.mlir b/tools/explorer/test/models/open_llama_3b_ttir.mlir new file mode 100644 index 000000000..231432712 --- /dev/null +++ b/tools/explorer/test/models/open_llama_3b_ttir.mlir @@ -0,0 +1,8364 @@ +#any_device = #tt.operand_constraint +#loc = loc("LlamaForCausalLM":0:0) +module @LlamaForCausalLM attributes {} { + func.func @forward(%arg0: tensor<1x12xi32> {ttir.name = "input_1"} loc("LlamaForCausalLM":0:0), %arg1: tensor<1xf32> {ttir.name = "input_1_add_4"} loc("LlamaForCausalLM":0:0), %arg2: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_14"} loc("LlamaForCausalLM":0:0), %arg3: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_24.1"} loc("LlamaForCausalLM":0:0), %arg4: tensor<1xf32> {ttir.name = "input_1_multiply_25"} loc("LlamaForCausalLM":0:0), %arg5: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_26.1"} loc("LlamaForCausalLM":0:0), %arg6: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_38.1"} loc("LlamaForCausalLM":0:0), %arg7: tensor<1xf32> {ttir.name = "input_1_multiply_39"} loc("LlamaForCausalLM":0:0), %arg8: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_40.1"} loc("LlamaForCausalLM":0:0), %arg9: tensor<1xf32> {ttir.name = "input_1_multiply_48"} loc("LlamaForCausalLM":0:0), %arg10: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_49"} loc("LlamaForCausalLM":0:0), %arg11: tensor<1xf32> {ttir.name = "input_1_add_70"} loc("LlamaForCausalLM":0:0), %arg12: tensor<1xf32> {ttir.name = "input_1_add_90"} loc("LlamaForCausalLM":0:0), %arg13: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_100"} loc("LlamaForCausalLM":0:0), %arg14: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_110.1"} loc("LlamaForCausalLM":0:0), %arg15: tensor<1xf32> {ttir.name = "input_1_multiply_111"} loc("LlamaForCausalLM":0:0), %arg16: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_112.1"} loc("LlamaForCausalLM":0:0), %arg17: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_124.1"} loc("LlamaForCausalLM":0:0), %arg18: tensor<1xf32> {ttir.name = "input_1_multiply_125"} loc("LlamaForCausalLM":0:0), %arg19: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_126.1"} loc("LlamaForCausalLM":0:0), %arg20: tensor<1xf32> {ttir.name = "input_1_multiply_134"} loc("LlamaForCausalLM":0:0), %arg21: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_135"} loc("LlamaForCausalLM":0:0), %arg22: tensor<1xf32> {ttir.name = "input_1_add_156"} loc("LlamaForCausalLM":0:0), %arg23: tensor<1xf32> {ttir.name = "input_1_add_176"} loc("LlamaForCausalLM":0:0), %arg24: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_186"} loc("LlamaForCausalLM":0:0), %arg25: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_196.1"} loc("LlamaForCausalLM":0:0), %arg26: tensor<1xf32> {ttir.name = "input_1_multiply_197"} loc("LlamaForCausalLM":0:0), %arg27: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_198.1"} loc("LlamaForCausalLM":0:0), %arg28: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_210.1"} loc("LlamaForCausalLM":0:0), %arg29: tensor<1xf32> {ttir.name = "input_1_multiply_211"} loc("LlamaForCausalLM":0:0), %arg30: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_212.1"} loc("LlamaForCausalLM":0:0), %arg31: tensor<1xf32> {ttir.name = "input_1_multiply_220"} loc("LlamaForCausalLM":0:0), %arg32: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_221"} loc("LlamaForCausalLM":0:0), %arg33: tensor<1xf32> {ttir.name = "input_1_add_242"} loc("LlamaForCausalLM":0:0), %arg34: tensor<1xf32> {ttir.name = "input_1_add_262"} loc("LlamaForCausalLM":0:0), %arg35: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_272"} loc("LlamaForCausalLM":0:0), %arg36: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_282.1"} loc("LlamaForCausalLM":0:0), %arg37: tensor<1xf32> {ttir.name = "input_1_multiply_283"} loc("LlamaForCausalLM":0:0), %arg38: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_284.1"} loc("LlamaForCausalLM":0:0), %arg39: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_296.1"} loc("LlamaForCausalLM":0:0), %arg40: tensor<1xf32> {ttir.name = "input_1_multiply_297"} loc("LlamaForCausalLM":0:0), %arg41: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_298.1"} loc("LlamaForCausalLM":0:0), %arg42: tensor<1xf32> {ttir.name = "input_1_multiply_306"} loc("LlamaForCausalLM":0:0), %arg43: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_307"} loc("LlamaForCausalLM":0:0), %arg44: tensor<1xf32> {ttir.name = "input_1_add_328"} loc("LlamaForCausalLM":0:0), %arg45: tensor<1xf32> {ttir.name = "input_1_add_348"} loc("LlamaForCausalLM":0:0), %arg46: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_358"} loc("LlamaForCausalLM":0:0), %arg47: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_368.1"} loc("LlamaForCausalLM":0:0), %arg48: tensor<1xf32> {ttir.name = "input_1_multiply_369"} loc("LlamaForCausalLM":0:0), %arg49: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_370.1"} loc("LlamaForCausalLM":0:0), %arg50: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_382.1"} loc("LlamaForCausalLM":0:0), %arg51: tensor<1xf32> {ttir.name = "input_1_multiply_383"} loc("LlamaForCausalLM":0:0), %arg52: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_384.1"} loc("LlamaForCausalLM":0:0), %arg53: tensor<1xf32> {ttir.name = "input_1_multiply_392"} loc("LlamaForCausalLM":0:0), %arg54: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_393"} loc("LlamaForCausalLM":0:0), %arg55: tensor<1xf32> {ttir.name = "input_1_add_414"} loc("LlamaForCausalLM":0:0), %arg56: tensor<1xf32> {ttir.name = "input_1_add_434"} loc("LlamaForCausalLM":0:0), %arg57: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_444"} loc("LlamaForCausalLM":0:0), %arg58: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_454.1"} loc("LlamaForCausalLM":0:0), %arg59: tensor<1xf32> {ttir.name = "input_1_multiply_455"} loc("LlamaForCausalLM":0:0), %arg60: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_456.1"} loc("LlamaForCausalLM":0:0), %arg61: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_468.1"} loc("LlamaForCausalLM":0:0), %arg62: tensor<1xf32> {ttir.name = "input_1_multiply_469"} loc("LlamaForCausalLM":0:0), %arg63: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_470.1"} loc("LlamaForCausalLM":0:0), %arg64: tensor<1xf32> {ttir.name = "input_1_multiply_478"} loc("LlamaForCausalLM":0:0), %arg65: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_479"} loc("LlamaForCausalLM":0:0), %arg66: tensor<1xf32> {ttir.name = "input_1_add_500"} loc("LlamaForCausalLM":0:0), %arg67: tensor<1xf32> {ttir.name = "input_1_add_520"} loc("LlamaForCausalLM":0:0), %arg68: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_530"} loc("LlamaForCausalLM":0:0), %arg69: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_540.1"} loc("LlamaForCausalLM":0:0), %arg70: tensor<1xf32> {ttir.name = "input_1_multiply_541"} loc("LlamaForCausalLM":0:0), %arg71: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_542.1"} loc("LlamaForCausalLM":0:0), %arg72: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_554.1"} loc("LlamaForCausalLM":0:0), %arg73: tensor<1xf32> {ttir.name = "input_1_multiply_555"} loc("LlamaForCausalLM":0:0), %arg74: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_556.1"} loc("LlamaForCausalLM":0:0), %arg75: tensor<1xf32> {ttir.name = "input_1_multiply_564"} loc("LlamaForCausalLM":0:0), %arg76: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_565"} loc("LlamaForCausalLM":0:0), %arg77: tensor<1xf32> {ttir.name = "input_1_add_586"} loc("LlamaForCausalLM":0:0), %arg78: tensor<1xf32> {ttir.name = "input_1_add_606"} loc("LlamaForCausalLM":0:0), %arg79: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_616"} loc("LlamaForCausalLM":0:0), %arg80: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_626.1"} loc("LlamaForCausalLM":0:0), %arg81: tensor<1xf32> {ttir.name = "input_1_multiply_627"} loc("LlamaForCausalLM":0:0), %arg82: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_628.1"} loc("LlamaForCausalLM":0:0), %arg83: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_640.1"} loc("LlamaForCausalLM":0:0), %arg84: tensor<1xf32> {ttir.name = "input_1_multiply_641"} loc("LlamaForCausalLM":0:0), %arg85: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_642.1"} loc("LlamaForCausalLM":0:0), %arg86: tensor<1xf32> {ttir.name = "input_1_multiply_650"} loc("LlamaForCausalLM":0:0), %arg87: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_651"} loc("LlamaForCausalLM":0:0), %arg88: tensor<1xf32> {ttir.name = "input_1_add_672"} loc("LlamaForCausalLM":0:0), %arg89: tensor<1xf32> {ttir.name = "input_1_add_692"} loc("LlamaForCausalLM":0:0), %arg90: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_702"} loc("LlamaForCausalLM":0:0), %arg91: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_712.1"} loc("LlamaForCausalLM":0:0), %arg92: tensor<1xf32> {ttir.name = "input_1_multiply_713"} loc("LlamaForCausalLM":0:0), %arg93: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_714.1"} loc("LlamaForCausalLM":0:0), %arg94: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_726.1"} loc("LlamaForCausalLM":0:0), %arg95: tensor<1xf32> {ttir.name = "input_1_multiply_727"} loc("LlamaForCausalLM":0:0), %arg96: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_728.1"} loc("LlamaForCausalLM":0:0), %arg97: tensor<1xf32> {ttir.name = "input_1_multiply_736"} loc("LlamaForCausalLM":0:0), %arg98: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_737"} loc("LlamaForCausalLM":0:0), %arg99: tensor<1xf32> {ttir.name = "input_1_add_758"} loc("LlamaForCausalLM":0:0), %arg100: tensor<1xf32> {ttir.name = "input_1_add_778"} loc("LlamaForCausalLM":0:0), %arg101: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_788"} loc("LlamaForCausalLM":0:0), %arg102: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_798.1"} loc("LlamaForCausalLM":0:0), %arg103: tensor<1xf32> {ttir.name = "input_1_multiply_799"} loc("LlamaForCausalLM":0:0), %arg104: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_800.1"} loc("LlamaForCausalLM":0:0), %arg105: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_812.1"} loc("LlamaForCausalLM":0:0), %arg106: tensor<1xf32> {ttir.name = "input_1_multiply_813"} loc("LlamaForCausalLM":0:0), %arg107: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_814.1"} loc("LlamaForCausalLM":0:0), %arg108: tensor<1xf32> {ttir.name = "input_1_multiply_822"} loc("LlamaForCausalLM":0:0), %arg109: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_823"} loc("LlamaForCausalLM":0:0), %arg110: tensor<1xf32> {ttir.name = "input_1_add_844"} loc("LlamaForCausalLM":0:0), %arg111: tensor<1xf32> {ttir.name = "input_1_add_864"} loc("LlamaForCausalLM":0:0), %arg112: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_874"} loc("LlamaForCausalLM":0:0), %arg113: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_884.1"} loc("LlamaForCausalLM":0:0), %arg114: tensor<1xf32> {ttir.name = "input_1_multiply_885"} loc("LlamaForCausalLM":0:0), %arg115: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_886.1"} loc("LlamaForCausalLM":0:0), %arg116: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_898.1"} loc("LlamaForCausalLM":0:0), %arg117: tensor<1xf32> {ttir.name = "input_1_multiply_899"} loc("LlamaForCausalLM":0:0), %arg118: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_900.1"} loc("LlamaForCausalLM":0:0), %arg119: tensor<1xf32> {ttir.name = "input_1_multiply_908"} loc("LlamaForCausalLM":0:0), %arg120: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_909"} loc("LlamaForCausalLM":0:0), %arg121: tensor<1xf32> {ttir.name = "input_1_add_930"} loc("LlamaForCausalLM":0:0), %arg122: tensor<1xf32> {ttir.name = "input_1_add_950"} loc("LlamaForCausalLM":0:0), %arg123: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_960"} loc("LlamaForCausalLM":0:0), %arg124: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_970.1"} loc("LlamaForCausalLM":0:0), %arg125: tensor<1xf32> {ttir.name = "input_1_multiply_971"} loc("LlamaForCausalLM":0:0), %arg126: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_972.1"} loc("LlamaForCausalLM":0:0), %arg127: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_984.1"} loc("LlamaForCausalLM":0:0), %arg128: tensor<1xf32> {ttir.name = "input_1_multiply_985"} loc("LlamaForCausalLM":0:0), %arg129: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_986.1"} loc("LlamaForCausalLM":0:0), %arg130: tensor<1xf32> {ttir.name = "input_1_multiply_994"} loc("LlamaForCausalLM":0:0), %arg131: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_995"} loc("LlamaForCausalLM":0:0), %arg132: tensor<1xf32> {ttir.name = "input_1_add_1016"} loc("LlamaForCausalLM":0:0), %arg133: tensor<1xf32> {ttir.name = "input_1_add_1036"} loc("LlamaForCausalLM":0:0), %arg134: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_1046"} loc("LlamaForCausalLM":0:0), %arg135: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1056.1"} loc("LlamaForCausalLM":0:0), %arg136: tensor<1xf32> {ttir.name = "input_1_multiply_1057"} loc("LlamaForCausalLM":0:0), %arg137: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1058.1"} loc("LlamaForCausalLM":0:0), %arg138: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1070.1"} loc("LlamaForCausalLM":0:0), %arg139: tensor<1xf32> {ttir.name = "input_1_multiply_1071"} loc("LlamaForCausalLM":0:0), %arg140: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1072.1"} loc("LlamaForCausalLM":0:0), %arg141: tensor<1xf32> {ttir.name = "input_1_multiply_1080"} loc("LlamaForCausalLM":0:0), %arg142: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_1081"} loc("LlamaForCausalLM":0:0), %arg143: tensor<1xf32> {ttir.name = "input_1_add_1102"} loc("LlamaForCausalLM":0:0), %arg144: tensor<1xf32> {ttir.name = "input_1_add_1122"} loc("LlamaForCausalLM":0:0), %arg145: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_1132"} loc("LlamaForCausalLM":0:0), %arg146: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1142.1"} loc("LlamaForCausalLM":0:0), %arg147: tensor<1xf32> {ttir.name = "input_1_multiply_1143"} loc("LlamaForCausalLM":0:0), %arg148: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1144.1"} loc("LlamaForCausalLM":0:0), %arg149: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1156.1"} loc("LlamaForCausalLM":0:0), %arg150: tensor<1xf32> {ttir.name = "input_1_multiply_1157"} loc("LlamaForCausalLM":0:0), %arg151: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1158.1"} loc("LlamaForCausalLM":0:0), %arg152: tensor<1xf32> {ttir.name = "input_1_multiply_1166"} loc("LlamaForCausalLM":0:0), %arg153: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_1167"} loc("LlamaForCausalLM":0:0), %arg154: tensor<1xf32> {ttir.name = "input_1_add_1188"} loc("LlamaForCausalLM":0:0), %arg155: tensor<1xf32> {ttir.name = "input_1_add_1208"} loc("LlamaForCausalLM":0:0), %arg156: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_1218"} loc("LlamaForCausalLM":0:0), %arg157: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1228.1"} loc("LlamaForCausalLM":0:0), %arg158: tensor<1xf32> {ttir.name = "input_1_multiply_1229"} loc("LlamaForCausalLM":0:0), %arg159: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1230.1"} loc("LlamaForCausalLM":0:0), %arg160: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1242.1"} loc("LlamaForCausalLM":0:0), %arg161: tensor<1xf32> {ttir.name = "input_1_multiply_1243"} loc("LlamaForCausalLM":0:0), %arg162: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1244.1"} loc("LlamaForCausalLM":0:0), %arg163: tensor<1xf32> {ttir.name = "input_1_multiply_1252"} loc("LlamaForCausalLM":0:0), %arg164: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_1253"} loc("LlamaForCausalLM":0:0), %arg165: tensor<1xf32> {ttir.name = "input_1_add_1274"} loc("LlamaForCausalLM":0:0), %arg166: tensor<1xf32> {ttir.name = "input_1_add_1294"} loc("LlamaForCausalLM":0:0), %arg167: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_1304"} loc("LlamaForCausalLM":0:0), %arg168: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1314.1"} loc("LlamaForCausalLM":0:0), %arg169: tensor<1xf32> {ttir.name = "input_1_multiply_1315"} loc("LlamaForCausalLM":0:0), %arg170: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1316.1"} loc("LlamaForCausalLM":0:0), %arg171: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1328.1"} loc("LlamaForCausalLM":0:0), %arg172: tensor<1xf32> {ttir.name = "input_1_multiply_1329"} loc("LlamaForCausalLM":0:0), %arg173: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1330.1"} loc("LlamaForCausalLM":0:0), %arg174: tensor<1xf32> {ttir.name = "input_1_multiply_1338"} loc("LlamaForCausalLM":0:0), %arg175: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_1339"} loc("LlamaForCausalLM":0:0), %arg176: tensor<1xf32> {ttir.name = "input_1_add_1360"} loc("LlamaForCausalLM":0:0), %arg177: tensor<1xf32> {ttir.name = "input_1_add_1380"} loc("LlamaForCausalLM":0:0), %arg178: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_1390"} loc("LlamaForCausalLM":0:0), %arg179: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1400.1"} loc("LlamaForCausalLM":0:0), %arg180: tensor<1xf32> {ttir.name = "input_1_multiply_1401"} loc("LlamaForCausalLM":0:0), %arg181: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1402.1"} loc("LlamaForCausalLM":0:0), %arg182: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1414.1"} loc("LlamaForCausalLM":0:0), %arg183: tensor<1xf32> {ttir.name = "input_1_multiply_1415"} loc("LlamaForCausalLM":0:0), %arg184: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1416.1"} loc("LlamaForCausalLM":0:0), %arg185: tensor<1xf32> {ttir.name = "input_1_multiply_1424"} loc("LlamaForCausalLM":0:0), %arg186: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_1425"} loc("LlamaForCausalLM":0:0), %arg187: tensor<1xf32> {ttir.name = "input_1_add_1446"} loc("LlamaForCausalLM":0:0), %arg188: tensor<1xf32> {ttir.name = "input_1_add_1466"} loc("LlamaForCausalLM":0:0), %arg189: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_1476"} loc("LlamaForCausalLM":0:0), %arg190: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1486.1"} loc("LlamaForCausalLM":0:0), %arg191: tensor<1xf32> {ttir.name = "input_1_multiply_1487"} loc("LlamaForCausalLM":0:0), %arg192: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1488.1"} loc("LlamaForCausalLM":0:0), %arg193: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1500.1"} loc("LlamaForCausalLM":0:0), %arg194: tensor<1xf32> {ttir.name = "input_1_multiply_1501"} loc("LlamaForCausalLM":0:0), %arg195: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1502.1"} loc("LlamaForCausalLM":0:0), %arg196: tensor<1xf32> {ttir.name = "input_1_multiply_1510"} loc("LlamaForCausalLM":0:0), %arg197: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_1511"} loc("LlamaForCausalLM":0:0), %arg198: tensor<1xf32> {ttir.name = "input_1_add_1532"} loc("LlamaForCausalLM":0:0), %arg199: tensor<1xf32> {ttir.name = "input_1_add_1552"} loc("LlamaForCausalLM":0:0), %arg200: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_1562"} loc("LlamaForCausalLM":0:0), %arg201: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1572.1"} loc("LlamaForCausalLM":0:0), %arg202: tensor<1xf32> {ttir.name = "input_1_multiply_1573"} loc("LlamaForCausalLM":0:0), %arg203: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1574.1"} loc("LlamaForCausalLM":0:0), %arg204: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1586.1"} loc("LlamaForCausalLM":0:0), %arg205: tensor<1xf32> {ttir.name = "input_1_multiply_1587"} loc("LlamaForCausalLM":0:0), %arg206: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1588.1"} loc("LlamaForCausalLM":0:0), %arg207: tensor<1xf32> {ttir.name = "input_1_multiply_1596"} loc("LlamaForCausalLM":0:0), %arg208: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_1597"} loc("LlamaForCausalLM":0:0), %arg209: tensor<1xf32> {ttir.name = "input_1_add_1618"} loc("LlamaForCausalLM":0:0), %arg210: tensor<1xf32> {ttir.name = "input_1_add_1638"} loc("LlamaForCausalLM":0:0), %arg211: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_1648"} loc("LlamaForCausalLM":0:0), %arg212: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1658.1"} loc("LlamaForCausalLM":0:0), %arg213: tensor<1xf32> {ttir.name = "input_1_multiply_1659"} loc("LlamaForCausalLM":0:0), %arg214: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1660.1"} loc("LlamaForCausalLM":0:0), %arg215: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1672.1"} loc("LlamaForCausalLM":0:0), %arg216: tensor<1xf32> {ttir.name = "input_1_multiply_1673"} loc("LlamaForCausalLM":0:0), %arg217: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1674.1"} loc("LlamaForCausalLM":0:0), %arg218: tensor<1xf32> {ttir.name = "input_1_multiply_1682"} loc("LlamaForCausalLM":0:0), %arg219: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_1683"} loc("LlamaForCausalLM":0:0), %arg220: tensor<1xf32> {ttir.name = "input_1_add_1704"} loc("LlamaForCausalLM":0:0), %arg221: tensor<1xf32> {ttir.name = "input_1_add_1724"} loc("LlamaForCausalLM":0:0), %arg222: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_1734"} loc("LlamaForCausalLM":0:0), %arg223: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1744.1"} loc("LlamaForCausalLM":0:0), %arg224: tensor<1xf32> {ttir.name = "input_1_multiply_1745"} loc("LlamaForCausalLM":0:0), %arg225: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1746.1"} loc("LlamaForCausalLM":0:0), %arg226: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1758.1"} loc("LlamaForCausalLM":0:0), %arg227: tensor<1xf32> {ttir.name = "input_1_multiply_1759"} loc("LlamaForCausalLM":0:0), %arg228: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1760.1"} loc("LlamaForCausalLM":0:0), %arg229: tensor<1xf32> {ttir.name = "input_1_multiply_1768"} loc("LlamaForCausalLM":0:0), %arg230: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_1769"} loc("LlamaForCausalLM":0:0), %arg231: tensor<1xf32> {ttir.name = "input_1_add_1790"} loc("LlamaForCausalLM":0:0), %arg232: tensor<1xf32> {ttir.name = "input_1_add_1810"} loc("LlamaForCausalLM":0:0), %arg233: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_1820"} loc("LlamaForCausalLM":0:0), %arg234: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1830.1"} loc("LlamaForCausalLM":0:0), %arg235: tensor<1xf32> {ttir.name = "input_1_multiply_1831"} loc("LlamaForCausalLM":0:0), %arg236: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1832.1"} loc("LlamaForCausalLM":0:0), %arg237: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1844.1"} loc("LlamaForCausalLM":0:0), %arg238: tensor<1xf32> {ttir.name = "input_1_multiply_1845"} loc("LlamaForCausalLM":0:0), %arg239: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1846.1"} loc("LlamaForCausalLM":0:0), %arg240: tensor<1xf32> {ttir.name = "input_1_multiply_1854"} loc("LlamaForCausalLM":0:0), %arg241: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_1855"} loc("LlamaForCausalLM":0:0), %arg242: tensor<1xf32> {ttir.name = "input_1_add_1876"} loc("LlamaForCausalLM":0:0), %arg243: tensor<1xf32> {ttir.name = "input_1_add_1896"} loc("LlamaForCausalLM":0:0), %arg244: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_1906"} loc("LlamaForCausalLM":0:0), %arg245: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1916.1"} loc("LlamaForCausalLM":0:0), %arg246: tensor<1xf32> {ttir.name = "input_1_multiply_1917"} loc("LlamaForCausalLM":0:0), %arg247: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1918.1"} loc("LlamaForCausalLM":0:0), %arg248: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1930.1"} loc("LlamaForCausalLM":0:0), %arg249: tensor<1xf32> {ttir.name = "input_1_multiply_1931"} loc("LlamaForCausalLM":0:0), %arg250: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_1932.1"} loc("LlamaForCausalLM":0:0), %arg251: tensor<1xf32> {ttir.name = "input_1_multiply_1940"} loc("LlamaForCausalLM":0:0), %arg252: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_1941"} loc("LlamaForCausalLM":0:0), %arg253: tensor<1xf32> {ttir.name = "input_1_add_1962"} loc("LlamaForCausalLM":0:0), %arg254: tensor<1xf32> {ttir.name = "input_1_add_1982"} loc("LlamaForCausalLM":0:0), %arg255: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_1992"} loc("LlamaForCausalLM":0:0), %arg256: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_2002.1"} loc("LlamaForCausalLM":0:0), %arg257: tensor<1xf32> {ttir.name = "input_1_multiply_2003"} loc("LlamaForCausalLM":0:0), %arg258: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_2004.1"} loc("LlamaForCausalLM":0:0), %arg259: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_2016.1"} loc("LlamaForCausalLM":0:0), %arg260: tensor<1xf32> {ttir.name = "input_1_multiply_2017"} loc("LlamaForCausalLM":0:0), %arg261: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_2018.1"} loc("LlamaForCausalLM":0:0), %arg262: tensor<1xf32> {ttir.name = "input_1_multiply_2026"} loc("LlamaForCausalLM":0:0), %arg263: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_2027"} loc("LlamaForCausalLM":0:0), %arg264: tensor<1xf32> {ttir.name = "input_1_add_2048"} loc("LlamaForCausalLM":0:0), %arg265: tensor<1xf32> {ttir.name = "input_1_add_2068"} loc("LlamaForCausalLM":0:0), %arg266: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_2078"} loc("LlamaForCausalLM":0:0), %arg267: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_2088.1"} loc("LlamaForCausalLM":0:0), %arg268: tensor<1xf32> {ttir.name = "input_1_multiply_2089"} loc("LlamaForCausalLM":0:0), %arg269: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_2090.1"} loc("LlamaForCausalLM":0:0), %arg270: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_2102.1"} loc("LlamaForCausalLM":0:0), %arg271: tensor<1xf32> {ttir.name = "input_1_multiply_2103"} loc("LlamaForCausalLM":0:0), %arg272: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_2104.1"} loc("LlamaForCausalLM":0:0), %arg273: tensor<1xf32> {ttir.name = "input_1_multiply_2112"} loc("LlamaForCausalLM":0:0), %arg274: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_2113"} loc("LlamaForCausalLM":0:0), %arg275: tensor<1xf32> {ttir.name = "input_1_add_2134"} loc("LlamaForCausalLM":0:0), %arg276: tensor<1xf32> {ttir.name = "input_1_add_2154"} loc("LlamaForCausalLM":0:0), %arg277: tensor<1x12x50xf32> {ttir.name = "input_0_unsqueeze_2164"} loc("LlamaForCausalLM":0:0), %arg278: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_2174.1"} loc("LlamaForCausalLM":0:0), %arg279: tensor<1xf32> {ttir.name = "input_1_multiply_2175"} loc("LlamaForCausalLM":0:0), %arg280: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_2176.1"} loc("LlamaForCausalLM":0:0), %arg281: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_2188.1"} loc("LlamaForCausalLM":0:0), %arg282: tensor<1xf32> {ttir.name = "input_1_multiply_2189"} loc("LlamaForCausalLM":0:0), %arg283: tensor<1x32x50x100xf32> {ttir.name = "dc.input_tensor.index_2190.1"} loc("LlamaForCausalLM":0:0), %arg284: tensor<1xf32> {ttir.name = "input_1_multiply_2198"} loc("LlamaForCausalLM":0:0), %arg285: tensor<1x1x12x12xf32> {ttir.name = "input_1_add_2199"} loc("LlamaForCausalLM":0:0), %arg286: tensor<1xf32> {ttir.name = "input_1_add_2220"} loc("LlamaForCausalLM":0:0), %arg287: tensor<1xf32> {ttir.name = "input_1_add_2240"} loc("LlamaForCausalLM":0:0), %arg288: tensor<3200xf32> {ttir.name = "model.norm.weight"} loc("LlamaForCausalLM":0:0), %arg289: tensor<32000x3200xf32> {ttir.name = "model.embed_tokens.weight"} loc("LlamaForCausalLM":0:0), %arg290: tensor<3200xf32> {ttir.name = "model.layers.0.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg291: tensor<3200x3200xf32> {ttir.name = "model.layers.0.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg292: tensor<3200x3200xf32> {ttir.name = "model.layers.0.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg293: tensor<3200x3200xf32> {ttir.name = "model.layers.0.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg294: tensor<3200x3200xf32> {ttir.name = "model.layers.0.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg295: tensor<3200xf32> {ttir.name = "model.layers.0.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg296: tensor<3200x8640xf32> {ttir.name = "model.layers.0.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg297: tensor<3200x8640xf32> {ttir.name = "model.layers.0.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg298: tensor<8640x3200xf32> {ttir.name = "model.layers.0.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg299: tensor<3200xf32> {ttir.name = "model.layers.1.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg300: tensor<3200x3200xf32> {ttir.name = "model.layers.1.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg301: tensor<3200x3200xf32> {ttir.name = "model.layers.1.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg302: tensor<3200x3200xf32> {ttir.name = "model.layers.1.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg303: tensor<3200x3200xf32> {ttir.name = "model.layers.1.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg304: tensor<3200xf32> {ttir.name = "model.layers.1.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg305: tensor<3200x8640xf32> {ttir.name = "model.layers.1.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg306: tensor<3200x8640xf32> {ttir.name = "model.layers.1.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg307: tensor<8640x3200xf32> {ttir.name = "model.layers.1.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg308: tensor<3200xf32> {ttir.name = "model.layers.2.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg309: tensor<3200x3200xf32> {ttir.name = "model.layers.2.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg310: tensor<3200x3200xf32> {ttir.name = "model.layers.2.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg311: tensor<3200x3200xf32> {ttir.name = "model.layers.2.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg312: tensor<3200x3200xf32> {ttir.name = "model.layers.2.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg313: tensor<3200xf32> {ttir.name = "model.layers.2.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg314: tensor<3200x8640xf32> {ttir.name = "model.layers.2.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg315: tensor<3200x8640xf32> {ttir.name = "model.layers.2.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg316: tensor<8640x3200xf32> {ttir.name = "model.layers.2.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg317: tensor<3200xf32> {ttir.name = "model.layers.3.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg318: tensor<3200x3200xf32> {ttir.name = "model.layers.3.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg319: tensor<3200x3200xf32> {ttir.name = "model.layers.3.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg320: tensor<3200x3200xf32> {ttir.name = "model.layers.3.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg321: tensor<3200x3200xf32> {ttir.name = "model.layers.3.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg322: tensor<3200xf32> {ttir.name = "model.layers.3.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg323: tensor<3200x8640xf32> {ttir.name = "model.layers.3.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg324: tensor<3200x8640xf32> {ttir.name = "model.layers.3.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg325: tensor<8640x3200xf32> {ttir.name = "model.layers.3.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg326: tensor<3200xf32> {ttir.name = "model.layers.4.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg327: tensor<3200x3200xf32> {ttir.name = "model.layers.4.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg328: tensor<3200x3200xf32> {ttir.name = "model.layers.4.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg329: tensor<3200x3200xf32> {ttir.name = "model.layers.4.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg330: tensor<3200x3200xf32> {ttir.name = "model.layers.4.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg331: tensor<3200xf32> {ttir.name = "model.layers.4.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg332: tensor<3200x8640xf32> {ttir.name = "model.layers.4.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg333: tensor<3200x8640xf32> {ttir.name = "model.layers.4.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg334: tensor<8640x3200xf32> {ttir.name = "model.layers.4.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg335: tensor<3200xf32> {ttir.name = "model.layers.5.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg336: tensor<3200x3200xf32> {ttir.name = "model.layers.5.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg337: tensor<3200x3200xf32> {ttir.name = "model.layers.5.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg338: tensor<3200x3200xf32> {ttir.name = "model.layers.5.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg339: tensor<3200x3200xf32> {ttir.name = "model.layers.5.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg340: tensor<3200xf32> {ttir.name = "model.layers.5.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg341: tensor<3200x8640xf32> {ttir.name = "model.layers.5.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg342: tensor<3200x8640xf32> {ttir.name = "model.layers.5.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg343: tensor<8640x3200xf32> {ttir.name = "model.layers.5.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg344: tensor<3200xf32> {ttir.name = "model.layers.6.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg345: tensor<3200x3200xf32> {ttir.name = "model.layers.6.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg346: tensor<3200x3200xf32> {ttir.name = "model.layers.6.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg347: tensor<3200x3200xf32> {ttir.name = "model.layers.6.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg348: tensor<3200x3200xf32> {ttir.name = "model.layers.6.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg349: tensor<3200xf32> {ttir.name = "model.layers.6.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg350: tensor<3200x8640xf32> {ttir.name = "model.layers.6.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg351: tensor<3200x8640xf32> {ttir.name = "model.layers.6.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg352: tensor<8640x3200xf32> {ttir.name = "model.layers.6.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg353: tensor<3200xf32> {ttir.name = "model.layers.7.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg354: tensor<3200x3200xf32> {ttir.name = "model.layers.7.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg355: tensor<3200x3200xf32> {ttir.name = "model.layers.7.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg356: tensor<3200x3200xf32> {ttir.name = "model.layers.7.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg357: tensor<3200x3200xf32> {ttir.name = "model.layers.7.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg358: tensor<3200xf32> {ttir.name = "model.layers.7.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg359: tensor<3200x8640xf32> {ttir.name = "model.layers.7.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg360: tensor<3200x8640xf32> {ttir.name = "model.layers.7.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg361: tensor<8640x3200xf32> {ttir.name = "model.layers.7.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg362: tensor<3200xf32> {ttir.name = "model.layers.8.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg363: tensor<3200x3200xf32> {ttir.name = "model.layers.8.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg364: tensor<3200x3200xf32> {ttir.name = "model.layers.8.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg365: tensor<3200x3200xf32> {ttir.name = "model.layers.8.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg366: tensor<3200x3200xf32> {ttir.name = "model.layers.8.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg367: tensor<3200xf32> {ttir.name = "model.layers.8.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg368: tensor<3200x8640xf32> {ttir.name = "model.layers.8.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg369: tensor<3200x8640xf32> {ttir.name = "model.layers.8.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg370: tensor<8640x3200xf32> {ttir.name = "model.layers.8.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg371: tensor<3200xf32> {ttir.name = "model.layers.9.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg372: tensor<3200x3200xf32> {ttir.name = "model.layers.9.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg373: tensor<3200x3200xf32> {ttir.name = "model.layers.9.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg374: tensor<3200x3200xf32> {ttir.name = "model.layers.9.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg375: tensor<3200x3200xf32> {ttir.name = "model.layers.9.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg376: tensor<3200xf32> {ttir.name = "model.layers.9.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg377: tensor<3200x8640xf32> {ttir.name = "model.layers.9.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg378: tensor<3200x8640xf32> {ttir.name = "model.layers.9.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg379: tensor<8640x3200xf32> {ttir.name = "model.layers.9.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg380: tensor<3200xf32> {ttir.name = "model.layers.10.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg381: tensor<3200x3200xf32> {ttir.name = "model.layers.10.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg382: tensor<3200x3200xf32> {ttir.name = "model.layers.10.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg383: tensor<3200x3200xf32> {ttir.name = "model.layers.10.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg384: tensor<3200x3200xf32> {ttir.name = "model.layers.10.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg385: tensor<3200xf32> {ttir.name = "model.layers.10.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg386: tensor<3200x8640xf32> {ttir.name = "model.layers.10.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg387: tensor<3200x8640xf32> {ttir.name = "model.layers.10.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg388: tensor<8640x3200xf32> {ttir.name = "model.layers.10.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg389: tensor<3200xf32> {ttir.name = "model.layers.11.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg390: tensor<3200x3200xf32> {ttir.name = "model.layers.11.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg391: tensor<3200x3200xf32> {ttir.name = "model.layers.11.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg392: tensor<3200x3200xf32> {ttir.name = "model.layers.11.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg393: tensor<3200x3200xf32> {ttir.name = "model.layers.11.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg394: tensor<3200xf32> {ttir.name = "model.layers.11.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg395: tensor<3200x8640xf32> {ttir.name = "model.layers.11.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg396: tensor<3200x8640xf32> {ttir.name = "model.layers.11.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg397: tensor<8640x3200xf32> {ttir.name = "model.layers.11.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg398: tensor<3200xf32> {ttir.name = "model.layers.12.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg399: tensor<3200x3200xf32> {ttir.name = "model.layers.12.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg400: tensor<3200x3200xf32> {ttir.name = "model.layers.12.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg401: tensor<3200x3200xf32> {ttir.name = "model.layers.12.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg402: tensor<3200x3200xf32> {ttir.name = "model.layers.12.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg403: tensor<3200xf32> {ttir.name = "model.layers.12.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg404: tensor<3200x8640xf32> {ttir.name = "model.layers.12.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg405: tensor<3200x8640xf32> {ttir.name = "model.layers.12.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg406: tensor<8640x3200xf32> {ttir.name = "model.layers.12.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg407: tensor<3200xf32> {ttir.name = "model.layers.13.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg408: tensor<3200x3200xf32> {ttir.name = "model.layers.13.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg409: tensor<3200x3200xf32> {ttir.name = "model.layers.13.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg410: tensor<3200x3200xf32> {ttir.name = "model.layers.13.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg411: tensor<3200x3200xf32> {ttir.name = "model.layers.13.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg412: tensor<3200xf32> {ttir.name = "model.layers.13.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg413: tensor<3200x8640xf32> {ttir.name = "model.layers.13.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg414: tensor<3200x8640xf32> {ttir.name = "model.layers.13.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg415: tensor<8640x3200xf32> {ttir.name = "model.layers.13.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg416: tensor<3200xf32> {ttir.name = "model.layers.14.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg417: tensor<3200x3200xf32> {ttir.name = "model.layers.14.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg418: tensor<3200x3200xf32> {ttir.name = "model.layers.14.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg419: tensor<3200x3200xf32> {ttir.name = "model.layers.14.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg420: tensor<3200x3200xf32> {ttir.name = "model.layers.14.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg421: tensor<3200xf32> {ttir.name = "model.layers.14.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg422: tensor<3200x8640xf32> {ttir.name = "model.layers.14.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg423: tensor<3200x8640xf32> {ttir.name = "model.layers.14.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg424: tensor<8640x3200xf32> {ttir.name = "model.layers.14.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg425: tensor<3200xf32> {ttir.name = "model.layers.15.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg426: tensor<3200x3200xf32> {ttir.name = "model.layers.15.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg427: tensor<3200x3200xf32> {ttir.name = "model.layers.15.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg428: tensor<3200x3200xf32> {ttir.name = "model.layers.15.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg429: tensor<3200x3200xf32> {ttir.name = "model.layers.15.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg430: tensor<3200xf32> {ttir.name = "model.layers.15.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg431: tensor<3200x8640xf32> {ttir.name = "model.layers.15.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg432: tensor<3200x8640xf32> {ttir.name = "model.layers.15.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg433: tensor<8640x3200xf32> {ttir.name = "model.layers.15.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg434: tensor<3200xf32> {ttir.name = "model.layers.16.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg435: tensor<3200x3200xf32> {ttir.name = "model.layers.16.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg436: tensor<3200x3200xf32> {ttir.name = "model.layers.16.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg437: tensor<3200x3200xf32> {ttir.name = "model.layers.16.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg438: tensor<3200x3200xf32> {ttir.name = "model.layers.16.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg439: tensor<3200xf32> {ttir.name = "model.layers.16.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg440: tensor<3200x8640xf32> {ttir.name = "model.layers.16.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg441: tensor<3200x8640xf32> {ttir.name = "model.layers.16.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg442: tensor<8640x3200xf32> {ttir.name = "model.layers.16.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg443: tensor<3200xf32> {ttir.name = "model.layers.17.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg444: tensor<3200x3200xf32> {ttir.name = "model.layers.17.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg445: tensor<3200x3200xf32> {ttir.name = "model.layers.17.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg446: tensor<3200x3200xf32> {ttir.name = "model.layers.17.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg447: tensor<3200x3200xf32> {ttir.name = "model.layers.17.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg448: tensor<3200xf32> {ttir.name = "model.layers.17.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg449: tensor<3200x8640xf32> {ttir.name = "model.layers.17.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg450: tensor<3200x8640xf32> {ttir.name = "model.layers.17.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg451: tensor<8640x3200xf32> {ttir.name = "model.layers.17.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg452: tensor<3200xf32> {ttir.name = "model.layers.18.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg453: tensor<3200x3200xf32> {ttir.name = "model.layers.18.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg454: tensor<3200x3200xf32> {ttir.name = "model.layers.18.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg455: tensor<3200x3200xf32> {ttir.name = "model.layers.18.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg456: tensor<3200x3200xf32> {ttir.name = "model.layers.18.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg457: tensor<3200xf32> {ttir.name = "model.layers.18.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg458: tensor<3200x8640xf32> {ttir.name = "model.layers.18.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg459: tensor<3200x8640xf32> {ttir.name = "model.layers.18.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg460: tensor<8640x3200xf32> {ttir.name = "model.layers.18.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg461: tensor<3200xf32> {ttir.name = "model.layers.19.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg462: tensor<3200x3200xf32> {ttir.name = "model.layers.19.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg463: tensor<3200x3200xf32> {ttir.name = "model.layers.19.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg464: tensor<3200x3200xf32> {ttir.name = "model.layers.19.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg465: tensor<3200x3200xf32> {ttir.name = "model.layers.19.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg466: tensor<3200xf32> {ttir.name = "model.layers.19.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg467: tensor<3200x8640xf32> {ttir.name = "model.layers.19.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg468: tensor<3200x8640xf32> {ttir.name = "model.layers.19.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg469: tensor<8640x3200xf32> {ttir.name = "model.layers.19.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg470: tensor<3200xf32> {ttir.name = "model.layers.20.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg471: tensor<3200x3200xf32> {ttir.name = "model.layers.20.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg472: tensor<3200x3200xf32> {ttir.name = "model.layers.20.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg473: tensor<3200x3200xf32> {ttir.name = "model.layers.20.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg474: tensor<3200x3200xf32> {ttir.name = "model.layers.20.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg475: tensor<3200xf32> {ttir.name = "model.layers.20.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg476: tensor<3200x8640xf32> {ttir.name = "model.layers.20.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg477: tensor<3200x8640xf32> {ttir.name = "model.layers.20.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg478: tensor<8640x3200xf32> {ttir.name = "model.layers.20.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg479: tensor<3200xf32> {ttir.name = "model.layers.21.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg480: tensor<3200x3200xf32> {ttir.name = "model.layers.21.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg481: tensor<3200x3200xf32> {ttir.name = "model.layers.21.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg482: tensor<3200x3200xf32> {ttir.name = "model.layers.21.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg483: tensor<3200x3200xf32> {ttir.name = "model.layers.21.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg484: tensor<3200xf32> {ttir.name = "model.layers.21.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg485: tensor<3200x8640xf32> {ttir.name = "model.layers.21.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg486: tensor<3200x8640xf32> {ttir.name = "model.layers.21.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg487: tensor<8640x3200xf32> {ttir.name = "model.layers.21.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg488: tensor<3200xf32> {ttir.name = "model.layers.22.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg489: tensor<3200x3200xf32> {ttir.name = "model.layers.22.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg490: tensor<3200x3200xf32> {ttir.name = "model.layers.22.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg491: tensor<3200x3200xf32> {ttir.name = "model.layers.22.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg492: tensor<3200x3200xf32> {ttir.name = "model.layers.22.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg493: tensor<3200xf32> {ttir.name = "model.layers.22.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg494: tensor<3200x8640xf32> {ttir.name = "model.layers.22.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg495: tensor<3200x8640xf32> {ttir.name = "model.layers.22.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg496: tensor<8640x3200xf32> {ttir.name = "model.layers.22.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg497: tensor<3200xf32> {ttir.name = "model.layers.23.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg498: tensor<3200x3200xf32> {ttir.name = "model.layers.23.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg499: tensor<3200x3200xf32> {ttir.name = "model.layers.23.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg500: tensor<3200x3200xf32> {ttir.name = "model.layers.23.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg501: tensor<3200x3200xf32> {ttir.name = "model.layers.23.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg502: tensor<3200xf32> {ttir.name = "model.layers.23.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg503: tensor<3200x8640xf32> {ttir.name = "model.layers.23.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg504: tensor<3200x8640xf32> {ttir.name = "model.layers.23.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg505: tensor<8640x3200xf32> {ttir.name = "model.layers.23.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg506: tensor<3200xf32> {ttir.name = "model.layers.24.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg507: tensor<3200x3200xf32> {ttir.name = "model.layers.24.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg508: tensor<3200x3200xf32> {ttir.name = "model.layers.24.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg509: tensor<3200x3200xf32> {ttir.name = "model.layers.24.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg510: tensor<3200x3200xf32> {ttir.name = "model.layers.24.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg511: tensor<3200xf32> {ttir.name = "model.layers.24.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg512: tensor<3200x8640xf32> {ttir.name = "model.layers.24.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg513: tensor<3200x8640xf32> {ttir.name = "model.layers.24.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg514: tensor<8640x3200xf32> {ttir.name = "model.layers.24.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg515: tensor<3200xf32> {ttir.name = "model.layers.25.input_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg516: tensor<3200x3200xf32> {ttir.name = "model.layers.25.self_attn.q_proj.weight"} loc("LlamaForCausalLM":0:0), %arg517: tensor<3200x3200xf32> {ttir.name = "model.layers.25.self_attn.k_proj.weight"} loc("LlamaForCausalLM":0:0), %arg518: tensor<3200x3200xf32> {ttir.name = "model.layers.25.self_attn.v_proj.weight"} loc("LlamaForCausalLM":0:0), %arg519: tensor<3200x3200xf32> {ttir.name = "model.layers.25.self_attn.o_proj.weight"} loc("LlamaForCausalLM":0:0), %arg520: tensor<3200xf32> {ttir.name = "model.layers.25.post_attention_layernorm.weight"} loc("LlamaForCausalLM":0:0), %arg521: tensor<3200x8640xf32> {ttir.name = "model.layers.25.mlp.gate_proj.weight"} loc("LlamaForCausalLM":0:0), %arg522: tensor<3200x8640xf32> {ttir.name = "model.layers.25.mlp.up_proj.weight"} loc("LlamaForCausalLM":0:0), %arg523: tensor<8640x3200xf32> {ttir.name = "model.layers.25.mlp.down_proj.weight"} loc("LlamaForCausalLM":0:0), %arg524: tensor<3200x32000xf32> {ttir.name = "lm_head.weight"} loc("LlamaForCausalLM":0:0)) -> (tensor<1x12x32000xf32> {ttir.name = "LlamaForCausalLM.output_matmul_2246"}) { + %0 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2091) + %1 = "ttir.embedding"(%arg0, %arg289, %0) <{operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12xi32>, tensor<32000x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2091) + %2 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2092) + %3 = "ttir.multiply"(%1, %1, %2) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2092) + %4 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2093) + %5 = "ttir.mean"(%3, %4) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2093) + %6 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2094) + %7 = "ttir.add"(%5, %arg1, %6) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2094) + %8 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2095) + %9 = "ttir.sqrt"(%7, %8) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2095) + %10 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2096) + %11 = "ttir.reciprocal"(%9, %10) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2096) + %12 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2097) + %13 = "ttir.multiply"(%1, %11, %12) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2097) + %14 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2098) + %15 = "ttir.multiply"(%arg290, %13, %14) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2098) + %16 = tensor.empty() : tensor<12x3200xf32> loc(#loc2099) + %17 = "ttir.squeeze"(%15, %16) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2099) + %18 = tensor.empty() : tensor<12x3200xf32> loc(#loc2100) + %19 = "ttir.matmul"(%17, %arg291, %18) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2100) + %20 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2101) + %21 = "ttir.reshape"(%19, %20) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2101) + %22 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2102) + %23 = "ttir.transpose"(%21, %22) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2102) + %24 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2103) + %25 = "ttir.concat"(%arg2, %arg2, %24) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2103) + %26 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2104) + %27 = "ttir.sin"(%25, %26) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2104) + %28 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc2105) + %29 = "ttir.unsqueeze"(%27, %28) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc2105) + %30 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2106) + %31 = "ttir.multiply"(%23, %29, %30) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2106) + %32 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2107) + %33 = "ttir.transpose"(%23, %32) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2107) + %34 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2108) + %35 = "ttir.matmul"(%arg3, %33, %34) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2108) + %36 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2109) + %37 = "ttir.transpose"(%35, %36) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2109) + %38 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2110) + %39 = "ttir.multiply"(%37, %arg4, %38) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2110) + %40 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2111) + %41 = "ttir.transpose"(%23, %40) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2111) + %42 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2112) + %43 = "ttir.matmul"(%arg5, %41, %42) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2112) + %44 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2113) + %45 = "ttir.transpose"(%43, %44) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2113) + %46 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2114) + %47 = "ttir.concat"(%39, %45, %46) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2114) + %48 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2115) + %49 = "ttir.cos"(%25, %48) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2115) + %50 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc2116) + %51 = "ttir.unsqueeze"(%49, %50) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc2116) + %52 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2117) + %53 = "ttir.multiply"(%47, %51, %52) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2117) + %54 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2118) + %55 = "ttir.add"(%31, %53, %54) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2118) + %56 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2119) + %57 = "ttir.squeeze"(%55, %56) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2119) + %58 = tensor.empty() : tensor<12x3200xf32> loc(#loc2120) + %59 = "ttir.matmul"(%17, %arg292, %58) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2120) + %60 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2121) + %61 = "ttir.reshape"(%59, %60) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2121) + %62 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2122) + %63 = "ttir.transpose"(%61, %62) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2122) + %64 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2123) + %65 = "ttir.multiply"(%63, %29, %64) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2123) + %66 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2124) + %67 = "ttir.transpose"(%63, %66) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2124) + %68 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2125) + %69 = "ttir.matmul"(%arg6, %67, %68) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2125) + %70 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2126) + %71 = "ttir.transpose"(%69, %70) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2126) + %72 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2127) + %73 = "ttir.multiply"(%71, %arg7, %72) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2127) + %74 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2128) + %75 = "ttir.transpose"(%63, %74) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2128) + %76 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2129) + %77 = "ttir.matmul"(%arg8, %75, %76) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2129) + %78 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2130) + %79 = "ttir.transpose"(%77, %78) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2130) + %80 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2131) + %81 = "ttir.concat"(%73, %79, %80) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2131) + %82 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2132) + %83 = "ttir.multiply"(%81, %51, %82) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2132) + %84 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2133) + %85 = "ttir.add"(%65, %83, %84) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2133) + %86 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2134) + %87 = "ttir.squeeze"(%85, %86) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2134) + %88 = tensor.empty() : tensor<32x100x12xf32> loc(#loc2135) + %89 = "ttir.transpose"(%87, %88) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc2135) + %90 = tensor.empty() : tensor<32x12x12xf32> loc(#loc2136) + %91 = "ttir.matmul"(%57, %89, %90) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc2136) + %92 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2137) + %93 = "ttir.unsqueeze"(%91, %92) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2137) + %94 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2138) + %95 = "ttir.multiply"(%93, %arg9, %94) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2138) + %96 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2139) + %97 = "ttir.add"(%95, %arg10, %96) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2139) + %98 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2140) + %99 = "ttir.softmax"(%97, %98) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2140) + %100 = tensor.empty() : tensor<32x12x12xf32> loc(#loc2141) + %101 = "ttir.squeeze"(%99, %100) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc2141) + %102 = tensor.empty() : tensor<12x3200xf32> loc(#loc2142) + %103 = "ttir.matmul"(%17, %arg293, %102) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2142) + %104 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2143) + %105 = "ttir.reshape"(%103, %104) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2143) + %106 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2144) + %107 = "ttir.transpose"(%105, %106) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2144) + %108 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2145) + %109 = "ttir.transpose"(%107, %108) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2145) + %110 = tensor.empty() : tensor<32x100x12xf32> loc(#loc2146) + %111 = "ttir.squeeze"(%109, %110) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc2146) + %112 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2147) + %113 = "ttir.transpose"(%111, %112) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2147) + %114 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2148) + %115 = "ttir.matmul"(%101, %113, %114) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2148) + %116 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2149) + %117 = "ttir.unsqueeze"(%115, %116) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2149) + %118 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2150) + %119 = "ttir.transpose"(%117, %118) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2150) + %120 = tensor.empty() : tensor<12x3200xf32> loc(#loc2151) + %121 = "ttir.reshape"(%119, %120) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2151) + %122 = tensor.empty() : tensor<12x3200xf32> loc(#loc2152) + %123 = "ttir.matmul"(%121, %arg294, %122) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2152) + %124 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2153) + %125 = "ttir.unsqueeze"(%123, %124) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2153) + %126 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2154) + %127 = "ttir.add"(%1, %125, %126) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2154) + %128 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2155) + %129 = "ttir.multiply"(%127, %127, %128) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2155) + %130 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2156) + %131 = "ttir.mean"(%129, %130) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2156) + %132 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2157) + %133 = "ttir.add"(%131, %arg11, %132) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2157) + %134 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2158) + %135 = "ttir.sqrt"(%133, %134) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2158) + %136 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2159) + %137 = "ttir.reciprocal"(%135, %136) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2159) + %138 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2160) + %139 = "ttir.multiply"(%127, %137, %138) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2160) + %140 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2161) + %141 = "ttir.multiply"(%arg295, %139, %140) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2161) + %142 = tensor.empty() : tensor<12x3200xf32> loc(#loc2162) + %143 = "ttir.squeeze"(%141, %142) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2162) + %144 = tensor.empty() : tensor<12x8640xf32> loc(#loc2163) + %145 = "ttir.matmul"(%143, %arg296, %144) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc2163) + %146 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2164) + %147 = "ttir.unsqueeze"(%145, %146) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2164) + %148 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2165) + %149 = "ttir.sigmoid"(%147, %148) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2165) + %150 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2166) + %151 = "ttir.multiply"(%147, %149, %150) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2166) + %152 = tensor.empty() : tensor<12x8640xf32> loc(#loc2167) + %153 = "ttir.matmul"(%143, %arg297, %152) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc2167) + %154 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2168) + %155 = "ttir.unsqueeze"(%153, %154) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2168) + %156 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2169) + %157 = "ttir.multiply"(%151, %155, %156) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2169) + %158 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2170) + %159 = "ttir.matmul"(%157, %arg298, %158) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2170) + %160 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2171) + %161 = "ttir.add"(%127, %159, %160) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2171) + %162 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2172) + %163 = "ttir.multiply"(%161, %161, %162) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2172) + %164 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2173) + %165 = "ttir.mean"(%163, %164) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2173) + %166 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2174) + %167 = "ttir.add"(%165, %arg12, %166) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2174) + %168 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2175) + %169 = "ttir.sqrt"(%167, %168) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2175) + %170 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2176) + %171 = "ttir.reciprocal"(%169, %170) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2176) + %172 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2177) + %173 = "ttir.multiply"(%161, %171, %172) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2177) + %174 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2178) + %175 = "ttir.multiply"(%arg299, %173, %174) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2178) + %176 = tensor.empty() : tensor<12x3200xf32> loc(#loc2179) + %177 = "ttir.squeeze"(%175, %176) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2179) + %178 = tensor.empty() : tensor<12x3200xf32> loc(#loc2180) + %179 = "ttir.matmul"(%177, %arg300, %178) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2180) + %180 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2181) + %181 = "ttir.reshape"(%179, %180) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2181) + %182 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2182) + %183 = "ttir.transpose"(%181, %182) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2182) + %184 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2183) + %185 = "ttir.concat"(%arg13, %arg13, %184) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2183) + %186 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2184) + %187 = "ttir.sin"(%185, %186) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2184) + %188 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc2185) + %189 = "ttir.unsqueeze"(%187, %188) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc2185) + %190 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2186) + %191 = "ttir.multiply"(%183, %189, %190) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2186) + %192 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2187) + %193 = "ttir.transpose"(%183, %192) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2187) + %194 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2188) + %195 = "ttir.matmul"(%arg14, %193, %194) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2188) + %196 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2189) + %197 = "ttir.transpose"(%195, %196) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2189) + %198 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2190) + %199 = "ttir.multiply"(%197, %arg15, %198) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2190) + %200 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2191) + %201 = "ttir.transpose"(%183, %200) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2191) + %202 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2192) + %203 = "ttir.matmul"(%arg16, %201, %202) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2192) + %204 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2193) + %205 = "ttir.transpose"(%203, %204) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2193) + %206 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2194) + %207 = "ttir.concat"(%199, %205, %206) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2194) + %208 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2195) + %209 = "ttir.cos"(%185, %208) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2195) + %210 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc2196) + %211 = "ttir.unsqueeze"(%209, %210) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc2196) + %212 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2197) + %213 = "ttir.multiply"(%207, %211, %212) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2197) + %214 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2198) + %215 = "ttir.add"(%191, %213, %214) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2198) + %216 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2199) + %217 = "ttir.squeeze"(%215, %216) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2199) + %218 = tensor.empty() : tensor<12x3200xf32> loc(#loc2200) + %219 = "ttir.matmul"(%177, %arg301, %218) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2200) + %220 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2201) + %221 = "ttir.reshape"(%219, %220) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2201) + %222 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2202) + %223 = "ttir.transpose"(%221, %222) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2202) + %224 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2203) + %225 = "ttir.multiply"(%223, %189, %224) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2203) + %226 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2204) + %227 = "ttir.transpose"(%223, %226) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2204) + %228 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2205) + %229 = "ttir.matmul"(%arg17, %227, %228) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2205) + %230 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2206) + %231 = "ttir.transpose"(%229, %230) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2206) + %232 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2207) + %233 = "ttir.multiply"(%231, %arg18, %232) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2207) + %234 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2208) + %235 = "ttir.transpose"(%223, %234) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2208) + %236 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2209) + %237 = "ttir.matmul"(%arg19, %235, %236) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2209) + %238 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2210) + %239 = "ttir.transpose"(%237, %238) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2210) + %240 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2211) + %241 = "ttir.concat"(%233, %239, %240) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2211) + %242 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2212) + %243 = "ttir.multiply"(%241, %211, %242) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2212) + %244 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2213) + %245 = "ttir.add"(%225, %243, %244) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2213) + %246 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2214) + %247 = "ttir.squeeze"(%245, %246) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2214) + %248 = tensor.empty() : tensor<32x100x12xf32> loc(#loc2215) + %249 = "ttir.transpose"(%247, %248) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc2215) + %250 = tensor.empty() : tensor<32x12x12xf32> loc(#loc2216) + %251 = "ttir.matmul"(%217, %249, %250) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc2216) + %252 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2217) + %253 = "ttir.unsqueeze"(%251, %252) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2217) + %254 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2218) + %255 = "ttir.multiply"(%253, %arg20, %254) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2218) + %256 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2219) + %257 = "ttir.add"(%255, %arg21, %256) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2219) + %258 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2220) + %259 = "ttir.softmax"(%257, %258) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2220) + %260 = tensor.empty() : tensor<32x12x12xf32> loc(#loc2221) + %261 = "ttir.squeeze"(%259, %260) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc2221) + %262 = tensor.empty() : tensor<12x3200xf32> loc(#loc2222) + %263 = "ttir.matmul"(%177, %arg302, %262) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2222) + %264 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2223) + %265 = "ttir.reshape"(%263, %264) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2223) + %266 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2224) + %267 = "ttir.transpose"(%265, %266) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2224) + %268 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2225) + %269 = "ttir.transpose"(%267, %268) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2225) + %270 = tensor.empty() : tensor<32x100x12xf32> loc(#loc2226) + %271 = "ttir.squeeze"(%269, %270) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc2226) + %272 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2227) + %273 = "ttir.transpose"(%271, %272) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2227) + %274 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2228) + %275 = "ttir.matmul"(%261, %273, %274) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2228) + %276 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2229) + %277 = "ttir.unsqueeze"(%275, %276) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2229) + %278 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2230) + %279 = "ttir.transpose"(%277, %278) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2230) + %280 = tensor.empty() : tensor<12x3200xf32> loc(#loc2231) + %281 = "ttir.reshape"(%279, %280) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2231) + %282 = tensor.empty() : tensor<12x3200xf32> loc(#loc2232) + %283 = "ttir.matmul"(%281, %arg303, %282) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2232) + %284 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2233) + %285 = "ttir.unsqueeze"(%283, %284) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2233) + %286 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2234) + %287 = "ttir.add"(%161, %285, %286) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2234) + %288 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2235) + %289 = "ttir.multiply"(%287, %287, %288) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2235) + %290 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2236) + %291 = "ttir.mean"(%289, %290) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2236) + %292 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2237) + %293 = "ttir.add"(%291, %arg22, %292) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2237) + %294 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2238) + %295 = "ttir.sqrt"(%293, %294) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2238) + %296 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2239) + %297 = "ttir.reciprocal"(%295, %296) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2239) + %298 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2240) + %299 = "ttir.multiply"(%287, %297, %298) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2240) + %300 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2241) + %301 = "ttir.multiply"(%arg304, %299, %300) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2241) + %302 = tensor.empty() : tensor<12x3200xf32> loc(#loc2242) + %303 = "ttir.squeeze"(%301, %302) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2242) + %304 = tensor.empty() : tensor<12x8640xf32> loc(#loc2243) + %305 = "ttir.matmul"(%303, %arg305, %304) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc2243) + %306 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2244) + %307 = "ttir.unsqueeze"(%305, %306) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2244) + %308 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2245) + %309 = "ttir.sigmoid"(%307, %308) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2245) + %310 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2246) + %311 = "ttir.multiply"(%307, %309, %310) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2246) + %312 = tensor.empty() : tensor<12x8640xf32> loc(#loc2247) + %313 = "ttir.matmul"(%303, %arg306, %312) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc2247) + %314 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2248) + %315 = "ttir.unsqueeze"(%313, %314) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2248) + %316 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2249) + %317 = "ttir.multiply"(%311, %315, %316) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2249) + %318 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2250) + %319 = "ttir.matmul"(%317, %arg307, %318) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2250) + %320 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2251) + %321 = "ttir.add"(%287, %319, %320) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2251) + %322 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2252) + %323 = "ttir.multiply"(%321, %321, %322) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2252) + %324 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2253) + %325 = "ttir.mean"(%323, %324) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2253) + %326 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2254) + %327 = "ttir.add"(%325, %arg23, %326) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2254) + %328 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2255) + %329 = "ttir.sqrt"(%327, %328) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2255) + %330 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2256) + %331 = "ttir.reciprocal"(%329, %330) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2256) + %332 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2257) + %333 = "ttir.multiply"(%321, %331, %332) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2257) + %334 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2258) + %335 = "ttir.multiply"(%arg308, %333, %334) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2258) + %336 = tensor.empty() : tensor<12x3200xf32> loc(#loc2259) + %337 = "ttir.squeeze"(%335, %336) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2259) + %338 = tensor.empty() : tensor<12x3200xf32> loc(#loc2260) + %339 = "ttir.matmul"(%337, %arg309, %338) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2260) + %340 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2261) + %341 = "ttir.reshape"(%339, %340) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2261) + %342 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2262) + %343 = "ttir.transpose"(%341, %342) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2262) + %344 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2263) + %345 = "ttir.concat"(%arg24, %arg24, %344) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2263) + %346 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2264) + %347 = "ttir.sin"(%345, %346) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2264) + %348 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc2265) + %349 = "ttir.unsqueeze"(%347, %348) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc2265) + %350 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2266) + %351 = "ttir.multiply"(%343, %349, %350) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2266) + %352 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2267) + %353 = "ttir.transpose"(%343, %352) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2267) + %354 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2268) + %355 = "ttir.matmul"(%arg25, %353, %354) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2268) + %356 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2269) + %357 = "ttir.transpose"(%355, %356) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2269) + %358 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2270) + %359 = "ttir.multiply"(%357, %arg26, %358) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2270) + %360 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2271) + %361 = "ttir.transpose"(%343, %360) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2271) + %362 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2272) + %363 = "ttir.matmul"(%arg27, %361, %362) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2272) + %364 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2273) + %365 = "ttir.transpose"(%363, %364) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2273) + %366 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2274) + %367 = "ttir.concat"(%359, %365, %366) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2274) + %368 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2275) + %369 = "ttir.cos"(%345, %368) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2275) + %370 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc2276) + %371 = "ttir.unsqueeze"(%369, %370) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc2276) + %372 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2277) + %373 = "ttir.multiply"(%367, %371, %372) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2277) + %374 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2278) + %375 = "ttir.add"(%351, %373, %374) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2278) + %376 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2279) + %377 = "ttir.squeeze"(%375, %376) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2279) + %378 = tensor.empty() : tensor<12x3200xf32> loc(#loc2280) + %379 = "ttir.matmul"(%337, %arg310, %378) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2280) + %380 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2281) + %381 = "ttir.reshape"(%379, %380) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2281) + %382 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2282) + %383 = "ttir.transpose"(%381, %382) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2282) + %384 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2283) + %385 = "ttir.multiply"(%383, %349, %384) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2283) + %386 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2284) + %387 = "ttir.transpose"(%383, %386) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2284) + %388 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2285) + %389 = "ttir.matmul"(%arg28, %387, %388) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2285) + %390 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2286) + %391 = "ttir.transpose"(%389, %390) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2286) + %392 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2287) + %393 = "ttir.multiply"(%391, %arg29, %392) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2287) + %394 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2288) + %395 = "ttir.transpose"(%383, %394) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2288) + %396 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2289) + %397 = "ttir.matmul"(%arg30, %395, %396) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2289) + %398 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2290) + %399 = "ttir.transpose"(%397, %398) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2290) + %400 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2291) + %401 = "ttir.concat"(%393, %399, %400) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2291) + %402 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2292) + %403 = "ttir.multiply"(%401, %371, %402) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2292) + %404 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2293) + %405 = "ttir.add"(%385, %403, %404) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2293) + %406 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2294) + %407 = "ttir.squeeze"(%405, %406) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2294) + %408 = tensor.empty() : tensor<32x100x12xf32> loc(#loc2295) + %409 = "ttir.transpose"(%407, %408) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc2295) + %410 = tensor.empty() : tensor<32x12x12xf32> loc(#loc2296) + %411 = "ttir.matmul"(%377, %409, %410) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc2296) + %412 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2297) + %413 = "ttir.unsqueeze"(%411, %412) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2297) + %414 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2298) + %415 = "ttir.multiply"(%413, %arg31, %414) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2298) + %416 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2299) + %417 = "ttir.add"(%415, %arg32, %416) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2299) + %418 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2300) + %419 = "ttir.softmax"(%417, %418) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2300) + %420 = tensor.empty() : tensor<32x12x12xf32> loc(#loc2301) + %421 = "ttir.squeeze"(%419, %420) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc2301) + %422 = tensor.empty() : tensor<12x3200xf32> loc(#loc2302) + %423 = "ttir.matmul"(%337, %arg311, %422) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2302) + %424 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2303) + %425 = "ttir.reshape"(%423, %424) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2303) + %426 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2304) + %427 = "ttir.transpose"(%425, %426) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2304) + %428 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2305) + %429 = "ttir.transpose"(%427, %428) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2305) + %430 = tensor.empty() : tensor<32x100x12xf32> loc(#loc2306) + %431 = "ttir.squeeze"(%429, %430) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc2306) + %432 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2307) + %433 = "ttir.transpose"(%431, %432) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2307) + %434 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2308) + %435 = "ttir.matmul"(%421, %433, %434) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2308) + %436 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2309) + %437 = "ttir.unsqueeze"(%435, %436) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2309) + %438 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2310) + %439 = "ttir.transpose"(%437, %438) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2310) + %440 = tensor.empty() : tensor<12x3200xf32> loc(#loc2311) + %441 = "ttir.reshape"(%439, %440) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2311) + %442 = tensor.empty() : tensor<12x3200xf32> loc(#loc2312) + %443 = "ttir.matmul"(%441, %arg312, %442) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2312) + %444 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2313) + %445 = "ttir.unsqueeze"(%443, %444) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2313) + %446 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2314) + %447 = "ttir.add"(%321, %445, %446) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2314) + %448 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2315) + %449 = "ttir.multiply"(%447, %447, %448) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2315) + %450 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2316) + %451 = "ttir.mean"(%449, %450) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2316) + %452 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2317) + %453 = "ttir.add"(%451, %arg33, %452) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2317) + %454 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2318) + %455 = "ttir.sqrt"(%453, %454) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2318) + %456 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2319) + %457 = "ttir.reciprocal"(%455, %456) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2319) + %458 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2320) + %459 = "ttir.multiply"(%447, %457, %458) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2320) + %460 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2321) + %461 = "ttir.multiply"(%arg313, %459, %460) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2321) + %462 = tensor.empty() : tensor<12x3200xf32> loc(#loc2322) + %463 = "ttir.squeeze"(%461, %462) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2322) + %464 = tensor.empty() : tensor<12x8640xf32> loc(#loc2323) + %465 = "ttir.matmul"(%463, %arg314, %464) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc2323) + %466 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2324) + %467 = "ttir.unsqueeze"(%465, %466) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2324) + %468 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2325) + %469 = "ttir.sigmoid"(%467, %468) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2325) + %470 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2326) + %471 = "ttir.multiply"(%467, %469, %470) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2326) + %472 = tensor.empty() : tensor<12x8640xf32> loc(#loc2327) + %473 = "ttir.matmul"(%463, %arg315, %472) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc2327) + %474 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2328) + %475 = "ttir.unsqueeze"(%473, %474) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2328) + %476 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2329) + %477 = "ttir.multiply"(%471, %475, %476) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2329) + %478 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2330) + %479 = "ttir.matmul"(%477, %arg316, %478) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2330) + %480 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2331) + %481 = "ttir.add"(%447, %479, %480) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2331) + %482 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2332) + %483 = "ttir.multiply"(%481, %481, %482) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2332) + %484 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2333) + %485 = "ttir.mean"(%483, %484) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2333) + %486 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2334) + %487 = "ttir.add"(%485, %arg34, %486) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2334) + %488 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2335) + %489 = "ttir.sqrt"(%487, %488) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2335) + %490 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2336) + %491 = "ttir.reciprocal"(%489, %490) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2336) + %492 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2337) + %493 = "ttir.multiply"(%481, %491, %492) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2337) + %494 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2338) + %495 = "ttir.multiply"(%arg317, %493, %494) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2338) + %496 = tensor.empty() : tensor<12x3200xf32> loc(#loc2339) + %497 = "ttir.squeeze"(%495, %496) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2339) + %498 = tensor.empty() : tensor<12x3200xf32> loc(#loc2340) + %499 = "ttir.matmul"(%497, %arg318, %498) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2340) + %500 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2341) + %501 = "ttir.reshape"(%499, %500) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2341) + %502 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2342) + %503 = "ttir.transpose"(%501, %502) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2342) + %504 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2343) + %505 = "ttir.concat"(%arg35, %arg35, %504) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2343) + %506 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2344) + %507 = "ttir.sin"(%505, %506) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2344) + %508 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc2345) + %509 = "ttir.unsqueeze"(%507, %508) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc2345) + %510 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2346) + %511 = "ttir.multiply"(%503, %509, %510) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2346) + %512 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2347) + %513 = "ttir.transpose"(%503, %512) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2347) + %514 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2348) + %515 = "ttir.matmul"(%arg36, %513, %514) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2348) + %516 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2349) + %517 = "ttir.transpose"(%515, %516) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2349) + %518 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2350) + %519 = "ttir.multiply"(%517, %arg37, %518) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2350) + %520 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2351) + %521 = "ttir.transpose"(%503, %520) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2351) + %522 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2352) + %523 = "ttir.matmul"(%arg38, %521, %522) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2352) + %524 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2353) + %525 = "ttir.transpose"(%523, %524) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2353) + %526 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2354) + %527 = "ttir.concat"(%519, %525, %526) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2354) + %528 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2355) + %529 = "ttir.cos"(%505, %528) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2355) + %530 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc2356) + %531 = "ttir.unsqueeze"(%529, %530) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc2356) + %532 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2357) + %533 = "ttir.multiply"(%527, %531, %532) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2357) + %534 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2358) + %535 = "ttir.add"(%511, %533, %534) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2358) + %536 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2359) + %537 = "ttir.squeeze"(%535, %536) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2359) + %538 = tensor.empty() : tensor<12x3200xf32> loc(#loc2360) + %539 = "ttir.matmul"(%497, %arg319, %538) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2360) + %540 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2361) + %541 = "ttir.reshape"(%539, %540) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2361) + %542 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2362) + %543 = "ttir.transpose"(%541, %542) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2362) + %544 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2363) + %545 = "ttir.multiply"(%543, %509, %544) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2363) + %546 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2364) + %547 = "ttir.transpose"(%543, %546) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2364) + %548 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2365) + %549 = "ttir.matmul"(%arg39, %547, %548) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2365) + %550 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2366) + %551 = "ttir.transpose"(%549, %550) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2366) + %552 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2367) + %553 = "ttir.multiply"(%551, %arg40, %552) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2367) + %554 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2368) + %555 = "ttir.transpose"(%543, %554) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2368) + %556 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2369) + %557 = "ttir.matmul"(%arg41, %555, %556) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2369) + %558 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2370) + %559 = "ttir.transpose"(%557, %558) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2370) + %560 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2371) + %561 = "ttir.concat"(%553, %559, %560) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2371) + %562 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2372) + %563 = "ttir.multiply"(%561, %531, %562) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2372) + %564 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2373) + %565 = "ttir.add"(%545, %563, %564) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2373) + %566 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2374) + %567 = "ttir.squeeze"(%565, %566) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2374) + %568 = tensor.empty() : tensor<32x100x12xf32> loc(#loc2375) + %569 = "ttir.transpose"(%567, %568) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc2375) + %570 = tensor.empty() : tensor<32x12x12xf32> loc(#loc2376) + %571 = "ttir.matmul"(%537, %569, %570) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc2376) + %572 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2377) + %573 = "ttir.unsqueeze"(%571, %572) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2377) + %574 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2378) + %575 = "ttir.multiply"(%573, %arg42, %574) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2378) + %576 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2379) + %577 = "ttir.add"(%575, %arg43, %576) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2379) + %578 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2380) + %579 = "ttir.softmax"(%577, %578) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2380) + %580 = tensor.empty() : tensor<32x12x12xf32> loc(#loc2381) + %581 = "ttir.squeeze"(%579, %580) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc2381) + %582 = tensor.empty() : tensor<12x3200xf32> loc(#loc2382) + %583 = "ttir.matmul"(%497, %arg320, %582) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2382) + %584 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2383) + %585 = "ttir.reshape"(%583, %584) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2383) + %586 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2384) + %587 = "ttir.transpose"(%585, %586) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2384) + %588 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2385) + %589 = "ttir.transpose"(%587, %588) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2385) + %590 = tensor.empty() : tensor<32x100x12xf32> loc(#loc2386) + %591 = "ttir.squeeze"(%589, %590) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc2386) + %592 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2387) + %593 = "ttir.transpose"(%591, %592) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2387) + %594 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2388) + %595 = "ttir.matmul"(%581, %593, %594) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2388) + %596 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2389) + %597 = "ttir.unsqueeze"(%595, %596) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2389) + %598 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2390) + %599 = "ttir.transpose"(%597, %598) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2390) + %600 = tensor.empty() : tensor<12x3200xf32> loc(#loc2391) + %601 = "ttir.reshape"(%599, %600) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2391) + %602 = tensor.empty() : tensor<12x3200xf32> loc(#loc2392) + %603 = "ttir.matmul"(%601, %arg321, %602) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2392) + %604 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2393) + %605 = "ttir.unsqueeze"(%603, %604) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2393) + %606 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2394) + %607 = "ttir.add"(%481, %605, %606) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2394) + %608 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2395) + %609 = "ttir.multiply"(%607, %607, %608) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2395) + %610 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2396) + %611 = "ttir.mean"(%609, %610) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2396) + %612 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2397) + %613 = "ttir.add"(%611, %arg44, %612) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2397) + %614 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2398) + %615 = "ttir.sqrt"(%613, %614) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2398) + %616 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2399) + %617 = "ttir.reciprocal"(%615, %616) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2399) + %618 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2400) + %619 = "ttir.multiply"(%607, %617, %618) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2400) + %620 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2401) + %621 = "ttir.multiply"(%arg322, %619, %620) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2401) + %622 = tensor.empty() : tensor<12x3200xf32> loc(#loc2402) + %623 = "ttir.squeeze"(%621, %622) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2402) + %624 = tensor.empty() : tensor<12x8640xf32> loc(#loc2403) + %625 = "ttir.matmul"(%623, %arg323, %624) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc2403) + %626 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2404) + %627 = "ttir.unsqueeze"(%625, %626) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2404) + %628 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2405) + %629 = "ttir.sigmoid"(%627, %628) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2405) + %630 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2406) + %631 = "ttir.multiply"(%627, %629, %630) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2406) + %632 = tensor.empty() : tensor<12x8640xf32> loc(#loc2407) + %633 = "ttir.matmul"(%623, %arg324, %632) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc2407) + %634 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2408) + %635 = "ttir.unsqueeze"(%633, %634) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2408) + %636 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2409) + %637 = "ttir.multiply"(%631, %635, %636) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2409) + %638 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2410) + %639 = "ttir.matmul"(%637, %arg325, %638) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2410) + %640 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2411) + %641 = "ttir.add"(%607, %639, %640) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2411) + %642 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2412) + %643 = "ttir.multiply"(%641, %641, %642) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2412) + %644 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2413) + %645 = "ttir.mean"(%643, %644) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2413) + %646 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2414) + %647 = "ttir.add"(%645, %arg45, %646) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2414) + %648 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2415) + %649 = "ttir.sqrt"(%647, %648) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2415) + %650 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2416) + %651 = "ttir.reciprocal"(%649, %650) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2416) + %652 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2417) + %653 = "ttir.multiply"(%641, %651, %652) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2417) + %654 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2418) + %655 = "ttir.multiply"(%arg326, %653, %654) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2418) + %656 = tensor.empty() : tensor<12x3200xf32> loc(#loc2419) + %657 = "ttir.squeeze"(%655, %656) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2419) + %658 = tensor.empty() : tensor<12x3200xf32> loc(#loc2420) + %659 = "ttir.matmul"(%657, %arg327, %658) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2420) + %660 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2421) + %661 = "ttir.reshape"(%659, %660) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2421) + %662 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2422) + %663 = "ttir.transpose"(%661, %662) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2422) + %664 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2423) + %665 = "ttir.concat"(%arg46, %arg46, %664) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2423) + %666 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2424) + %667 = "ttir.sin"(%665, %666) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2424) + %668 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc2425) + %669 = "ttir.unsqueeze"(%667, %668) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc2425) + %670 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2426) + %671 = "ttir.multiply"(%663, %669, %670) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2426) + %672 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2427) + %673 = "ttir.transpose"(%663, %672) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2427) + %674 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2428) + %675 = "ttir.matmul"(%arg47, %673, %674) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2428) + %676 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2429) + %677 = "ttir.transpose"(%675, %676) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2429) + %678 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2430) + %679 = "ttir.multiply"(%677, %arg48, %678) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2430) + %680 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2431) + %681 = "ttir.transpose"(%663, %680) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2431) + %682 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2432) + %683 = "ttir.matmul"(%arg49, %681, %682) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2432) + %684 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2433) + %685 = "ttir.transpose"(%683, %684) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2433) + %686 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2434) + %687 = "ttir.concat"(%679, %685, %686) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2434) + %688 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2435) + %689 = "ttir.cos"(%665, %688) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2435) + %690 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc2436) + %691 = "ttir.unsqueeze"(%689, %690) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc2436) + %692 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2437) + %693 = "ttir.multiply"(%687, %691, %692) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2437) + %694 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2438) + %695 = "ttir.add"(%671, %693, %694) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2438) + %696 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2439) + %697 = "ttir.squeeze"(%695, %696) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2439) + %698 = tensor.empty() : tensor<12x3200xf32> loc(#loc2440) + %699 = "ttir.matmul"(%657, %arg328, %698) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2440) + %700 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2441) + %701 = "ttir.reshape"(%699, %700) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2441) + %702 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2442) + %703 = "ttir.transpose"(%701, %702) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2442) + %704 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2443) + %705 = "ttir.multiply"(%703, %669, %704) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2443) + %706 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2444) + %707 = "ttir.transpose"(%703, %706) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2444) + %708 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2445) + %709 = "ttir.matmul"(%arg50, %707, %708) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2445) + %710 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2446) + %711 = "ttir.transpose"(%709, %710) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2446) + %712 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2447) + %713 = "ttir.multiply"(%711, %arg51, %712) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2447) + %714 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2448) + %715 = "ttir.transpose"(%703, %714) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2448) + %716 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2449) + %717 = "ttir.matmul"(%arg52, %715, %716) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2449) + %718 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2450) + %719 = "ttir.transpose"(%717, %718) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2450) + %720 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2451) + %721 = "ttir.concat"(%713, %719, %720) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2451) + %722 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2452) + %723 = "ttir.multiply"(%721, %691, %722) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2452) + %724 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2453) + %725 = "ttir.add"(%705, %723, %724) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2453) + %726 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2454) + %727 = "ttir.squeeze"(%725, %726) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2454) + %728 = tensor.empty() : tensor<32x100x12xf32> loc(#loc2455) + %729 = "ttir.transpose"(%727, %728) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc2455) + %730 = tensor.empty() : tensor<32x12x12xf32> loc(#loc2456) + %731 = "ttir.matmul"(%697, %729, %730) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc2456) + %732 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2457) + %733 = "ttir.unsqueeze"(%731, %732) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2457) + %734 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2458) + %735 = "ttir.multiply"(%733, %arg53, %734) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2458) + %736 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2459) + %737 = "ttir.add"(%735, %arg54, %736) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2459) + %738 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2460) + %739 = "ttir.softmax"(%737, %738) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2460) + %740 = tensor.empty() : tensor<32x12x12xf32> loc(#loc2461) + %741 = "ttir.squeeze"(%739, %740) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc2461) + %742 = tensor.empty() : tensor<12x3200xf32> loc(#loc2462) + %743 = "ttir.matmul"(%657, %arg329, %742) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2462) + %744 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2463) + %745 = "ttir.reshape"(%743, %744) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2463) + %746 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2464) + %747 = "ttir.transpose"(%745, %746) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2464) + %748 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2465) + %749 = "ttir.transpose"(%747, %748) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2465) + %750 = tensor.empty() : tensor<32x100x12xf32> loc(#loc2466) + %751 = "ttir.squeeze"(%749, %750) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc2466) + %752 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2467) + %753 = "ttir.transpose"(%751, %752) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2467) + %754 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2468) + %755 = "ttir.matmul"(%741, %753, %754) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2468) + %756 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2469) + %757 = "ttir.unsqueeze"(%755, %756) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2469) + %758 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2470) + %759 = "ttir.transpose"(%757, %758) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2470) + %760 = tensor.empty() : tensor<12x3200xf32> loc(#loc2471) + %761 = "ttir.reshape"(%759, %760) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2471) + %762 = tensor.empty() : tensor<12x3200xf32> loc(#loc2472) + %763 = "ttir.matmul"(%761, %arg330, %762) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2472) + %764 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2473) + %765 = "ttir.unsqueeze"(%763, %764) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2473) + %766 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2474) + %767 = "ttir.add"(%641, %765, %766) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2474) + %768 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2475) + %769 = "ttir.multiply"(%767, %767, %768) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2475) + %770 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2476) + %771 = "ttir.mean"(%769, %770) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2476) + %772 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2477) + %773 = "ttir.add"(%771, %arg55, %772) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2477) + %774 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2478) + %775 = "ttir.sqrt"(%773, %774) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2478) + %776 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2479) + %777 = "ttir.reciprocal"(%775, %776) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2479) + %778 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2480) + %779 = "ttir.multiply"(%767, %777, %778) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2480) + %780 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2481) + %781 = "ttir.multiply"(%arg331, %779, %780) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2481) + %782 = tensor.empty() : tensor<12x3200xf32> loc(#loc2482) + %783 = "ttir.squeeze"(%781, %782) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2482) + %784 = tensor.empty() : tensor<12x8640xf32> loc(#loc2483) + %785 = "ttir.matmul"(%783, %arg332, %784) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc2483) + %786 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2484) + %787 = "ttir.unsqueeze"(%785, %786) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2484) + %788 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2485) + %789 = "ttir.sigmoid"(%787, %788) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2485) + %790 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2486) + %791 = "ttir.multiply"(%787, %789, %790) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2486) + %792 = tensor.empty() : tensor<12x8640xf32> loc(#loc2487) + %793 = "ttir.matmul"(%783, %arg333, %792) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc2487) + %794 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2488) + %795 = "ttir.unsqueeze"(%793, %794) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2488) + %796 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2489) + %797 = "ttir.multiply"(%791, %795, %796) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2489) + %798 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2490) + %799 = "ttir.matmul"(%797, %arg334, %798) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2490) + %800 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2491) + %801 = "ttir.add"(%767, %799, %800) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2491) + %802 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2492) + %803 = "ttir.multiply"(%801, %801, %802) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2492) + %804 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2493) + %805 = "ttir.mean"(%803, %804) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2493) + %806 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2494) + %807 = "ttir.add"(%805, %arg56, %806) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2494) + %808 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2495) + %809 = "ttir.sqrt"(%807, %808) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2495) + %810 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2496) + %811 = "ttir.reciprocal"(%809, %810) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2496) + %812 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2497) + %813 = "ttir.multiply"(%801, %811, %812) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2497) + %814 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2498) + %815 = "ttir.multiply"(%arg335, %813, %814) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2498) + %816 = tensor.empty() : tensor<12x3200xf32> loc(#loc2499) + %817 = "ttir.squeeze"(%815, %816) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2499) + %818 = tensor.empty() : tensor<12x3200xf32> loc(#loc2500) + %819 = "ttir.matmul"(%817, %arg336, %818) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2500) + %820 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2501) + %821 = "ttir.reshape"(%819, %820) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2501) + %822 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2502) + %823 = "ttir.transpose"(%821, %822) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2502) + %824 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2503) + %825 = "ttir.concat"(%arg57, %arg57, %824) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2503) + %826 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2504) + %827 = "ttir.sin"(%825, %826) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2504) + %828 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc2505) + %829 = "ttir.unsqueeze"(%827, %828) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc2505) + %830 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2506) + %831 = "ttir.multiply"(%823, %829, %830) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2506) + %832 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2507) + %833 = "ttir.transpose"(%823, %832) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2507) + %834 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2508) + %835 = "ttir.matmul"(%arg58, %833, %834) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2508) + %836 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2509) + %837 = "ttir.transpose"(%835, %836) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2509) + %838 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2510) + %839 = "ttir.multiply"(%837, %arg59, %838) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2510) + %840 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2511) + %841 = "ttir.transpose"(%823, %840) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2511) + %842 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2512) + %843 = "ttir.matmul"(%arg60, %841, %842) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2512) + %844 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2513) + %845 = "ttir.transpose"(%843, %844) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2513) + %846 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2514) + %847 = "ttir.concat"(%839, %845, %846) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2514) + %848 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2515) + %849 = "ttir.cos"(%825, %848) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2515) + %850 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc2516) + %851 = "ttir.unsqueeze"(%849, %850) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc2516) + %852 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2517) + %853 = "ttir.multiply"(%847, %851, %852) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2517) + %854 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2518) + %855 = "ttir.add"(%831, %853, %854) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2518) + %856 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2519) + %857 = "ttir.squeeze"(%855, %856) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2519) + %858 = tensor.empty() : tensor<12x3200xf32> loc(#loc2520) + %859 = "ttir.matmul"(%817, %arg337, %858) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2520) + %860 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2521) + %861 = "ttir.reshape"(%859, %860) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2521) + %862 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2522) + %863 = "ttir.transpose"(%861, %862) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2522) + %864 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2523) + %865 = "ttir.multiply"(%863, %829, %864) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2523) + %866 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2524) + %867 = "ttir.transpose"(%863, %866) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2524) + %868 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2525) + %869 = "ttir.matmul"(%arg61, %867, %868) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2525) + %870 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2526) + %871 = "ttir.transpose"(%869, %870) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2526) + %872 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2527) + %873 = "ttir.multiply"(%871, %arg62, %872) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2527) + %874 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2528) + %875 = "ttir.transpose"(%863, %874) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2528) + %876 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2529) + %877 = "ttir.matmul"(%arg63, %875, %876) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2529) + %878 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2530) + %879 = "ttir.transpose"(%877, %878) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2530) + %880 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2531) + %881 = "ttir.concat"(%873, %879, %880) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2531) + %882 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2532) + %883 = "ttir.multiply"(%881, %851, %882) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2532) + %884 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2533) + %885 = "ttir.add"(%865, %883, %884) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2533) + %886 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2534) + %887 = "ttir.squeeze"(%885, %886) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2534) + %888 = tensor.empty() : tensor<32x100x12xf32> loc(#loc2535) + %889 = "ttir.transpose"(%887, %888) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc2535) + %890 = tensor.empty() : tensor<32x12x12xf32> loc(#loc2536) + %891 = "ttir.matmul"(%857, %889, %890) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc2536) + %892 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2537) + %893 = "ttir.unsqueeze"(%891, %892) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2537) + %894 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2538) + %895 = "ttir.multiply"(%893, %arg64, %894) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2538) + %896 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2539) + %897 = "ttir.add"(%895, %arg65, %896) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2539) + %898 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2540) + %899 = "ttir.softmax"(%897, %898) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2540) + %900 = tensor.empty() : tensor<32x12x12xf32> loc(#loc2541) + %901 = "ttir.squeeze"(%899, %900) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc2541) + %902 = tensor.empty() : tensor<12x3200xf32> loc(#loc2542) + %903 = "ttir.matmul"(%817, %arg338, %902) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2542) + %904 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2543) + %905 = "ttir.reshape"(%903, %904) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2543) + %906 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2544) + %907 = "ttir.transpose"(%905, %906) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2544) + %908 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2545) + %909 = "ttir.transpose"(%907, %908) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2545) + %910 = tensor.empty() : tensor<32x100x12xf32> loc(#loc2546) + %911 = "ttir.squeeze"(%909, %910) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc2546) + %912 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2547) + %913 = "ttir.transpose"(%911, %912) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2547) + %914 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2548) + %915 = "ttir.matmul"(%901, %913, %914) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2548) + %916 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2549) + %917 = "ttir.unsqueeze"(%915, %916) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2549) + %918 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2550) + %919 = "ttir.transpose"(%917, %918) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2550) + %920 = tensor.empty() : tensor<12x3200xf32> loc(#loc2551) + %921 = "ttir.reshape"(%919, %920) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2551) + %922 = tensor.empty() : tensor<12x3200xf32> loc(#loc2552) + %923 = "ttir.matmul"(%921, %arg339, %922) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2552) + %924 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2553) + %925 = "ttir.unsqueeze"(%923, %924) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2553) + %926 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2554) + %927 = "ttir.add"(%801, %925, %926) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2554) + %928 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2555) + %929 = "ttir.multiply"(%927, %927, %928) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2555) + %930 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2556) + %931 = "ttir.mean"(%929, %930) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2556) + %932 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2557) + %933 = "ttir.add"(%931, %arg66, %932) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2557) + %934 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2558) + %935 = "ttir.sqrt"(%933, %934) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2558) + %936 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2559) + %937 = "ttir.reciprocal"(%935, %936) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2559) + %938 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2560) + %939 = "ttir.multiply"(%927, %937, %938) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2560) + %940 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2561) + %941 = "ttir.multiply"(%arg340, %939, %940) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2561) + %942 = tensor.empty() : tensor<12x3200xf32> loc(#loc2562) + %943 = "ttir.squeeze"(%941, %942) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2562) + %944 = tensor.empty() : tensor<12x8640xf32> loc(#loc2563) + %945 = "ttir.matmul"(%943, %arg341, %944) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc2563) + %946 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2564) + %947 = "ttir.unsqueeze"(%945, %946) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2564) + %948 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2565) + %949 = "ttir.sigmoid"(%947, %948) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2565) + %950 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2566) + %951 = "ttir.multiply"(%947, %949, %950) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2566) + %952 = tensor.empty() : tensor<12x8640xf32> loc(#loc2567) + %953 = "ttir.matmul"(%943, %arg342, %952) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc2567) + %954 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2568) + %955 = "ttir.unsqueeze"(%953, %954) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2568) + %956 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2569) + %957 = "ttir.multiply"(%951, %955, %956) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2569) + %958 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2570) + %959 = "ttir.matmul"(%957, %arg343, %958) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2570) + %960 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2571) + %961 = "ttir.add"(%927, %959, %960) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2571) + %962 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2572) + %963 = "ttir.multiply"(%961, %961, %962) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2572) + %964 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2573) + %965 = "ttir.mean"(%963, %964) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2573) + %966 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2574) + %967 = "ttir.add"(%965, %arg67, %966) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2574) + %968 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2575) + %969 = "ttir.sqrt"(%967, %968) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2575) + %970 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2576) + %971 = "ttir.reciprocal"(%969, %970) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2576) + %972 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2577) + %973 = "ttir.multiply"(%961, %971, %972) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2577) + %974 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2578) + %975 = "ttir.multiply"(%arg344, %973, %974) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2578) + %976 = tensor.empty() : tensor<12x3200xf32> loc(#loc2579) + %977 = "ttir.squeeze"(%975, %976) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2579) + %978 = tensor.empty() : tensor<12x3200xf32> loc(#loc2580) + %979 = "ttir.matmul"(%977, %arg345, %978) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2580) + %980 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2581) + %981 = "ttir.reshape"(%979, %980) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2581) + %982 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2582) + %983 = "ttir.transpose"(%981, %982) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2582) + %984 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2583) + %985 = "ttir.concat"(%arg68, %arg68, %984) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2583) + %986 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2584) + %987 = "ttir.sin"(%985, %986) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2584) + %988 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc2585) + %989 = "ttir.unsqueeze"(%987, %988) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc2585) + %990 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2586) + %991 = "ttir.multiply"(%983, %989, %990) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2586) + %992 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2587) + %993 = "ttir.transpose"(%983, %992) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2587) + %994 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2588) + %995 = "ttir.matmul"(%arg69, %993, %994) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2588) + %996 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2589) + %997 = "ttir.transpose"(%995, %996) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2589) + %998 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2590) + %999 = "ttir.multiply"(%997, %arg70, %998) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2590) + %1000 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2591) + %1001 = "ttir.transpose"(%983, %1000) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2591) + %1002 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2592) + %1003 = "ttir.matmul"(%arg71, %1001, %1002) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2592) + %1004 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2593) + %1005 = "ttir.transpose"(%1003, %1004) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2593) + %1006 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2594) + %1007 = "ttir.concat"(%999, %1005, %1006) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2594) + %1008 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2595) + %1009 = "ttir.cos"(%985, %1008) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2595) + %1010 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc2596) + %1011 = "ttir.unsqueeze"(%1009, %1010) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc2596) + %1012 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2597) + %1013 = "ttir.multiply"(%1007, %1011, %1012) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2597) + %1014 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2598) + %1015 = "ttir.add"(%991, %1013, %1014) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2598) + %1016 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2599) + %1017 = "ttir.squeeze"(%1015, %1016) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2599) + %1018 = tensor.empty() : tensor<12x3200xf32> loc(#loc2600) + %1019 = "ttir.matmul"(%977, %arg346, %1018) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2600) + %1020 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2601) + %1021 = "ttir.reshape"(%1019, %1020) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2601) + %1022 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2602) + %1023 = "ttir.transpose"(%1021, %1022) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2602) + %1024 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2603) + %1025 = "ttir.multiply"(%1023, %989, %1024) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2603) + %1026 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2604) + %1027 = "ttir.transpose"(%1023, %1026) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2604) + %1028 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2605) + %1029 = "ttir.matmul"(%arg72, %1027, %1028) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2605) + %1030 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2606) + %1031 = "ttir.transpose"(%1029, %1030) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2606) + %1032 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2607) + %1033 = "ttir.multiply"(%1031, %arg73, %1032) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2607) + %1034 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2608) + %1035 = "ttir.transpose"(%1023, %1034) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2608) + %1036 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2609) + %1037 = "ttir.matmul"(%arg74, %1035, %1036) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2609) + %1038 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2610) + %1039 = "ttir.transpose"(%1037, %1038) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2610) + %1040 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2611) + %1041 = "ttir.concat"(%1033, %1039, %1040) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2611) + %1042 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2612) + %1043 = "ttir.multiply"(%1041, %1011, %1042) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2612) + %1044 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2613) + %1045 = "ttir.add"(%1025, %1043, %1044) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2613) + %1046 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2614) + %1047 = "ttir.squeeze"(%1045, %1046) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2614) + %1048 = tensor.empty() : tensor<32x100x12xf32> loc(#loc2615) + %1049 = "ttir.transpose"(%1047, %1048) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc2615) + %1050 = tensor.empty() : tensor<32x12x12xf32> loc(#loc2616) + %1051 = "ttir.matmul"(%1017, %1049, %1050) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc2616) + %1052 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2617) + %1053 = "ttir.unsqueeze"(%1051, %1052) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2617) + %1054 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2618) + %1055 = "ttir.multiply"(%1053, %arg75, %1054) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2618) + %1056 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2619) + %1057 = "ttir.add"(%1055, %arg76, %1056) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2619) + %1058 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2620) + %1059 = "ttir.softmax"(%1057, %1058) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2620) + %1060 = tensor.empty() : tensor<32x12x12xf32> loc(#loc2621) + %1061 = "ttir.squeeze"(%1059, %1060) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc2621) + %1062 = tensor.empty() : tensor<12x3200xf32> loc(#loc2622) + %1063 = "ttir.matmul"(%977, %arg347, %1062) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2622) + %1064 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2623) + %1065 = "ttir.reshape"(%1063, %1064) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2623) + %1066 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2624) + %1067 = "ttir.transpose"(%1065, %1066) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2624) + %1068 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2625) + %1069 = "ttir.transpose"(%1067, %1068) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2625) + %1070 = tensor.empty() : tensor<32x100x12xf32> loc(#loc2626) + %1071 = "ttir.squeeze"(%1069, %1070) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc2626) + %1072 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2627) + %1073 = "ttir.transpose"(%1071, %1072) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2627) + %1074 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2628) + %1075 = "ttir.matmul"(%1061, %1073, %1074) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2628) + %1076 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2629) + %1077 = "ttir.unsqueeze"(%1075, %1076) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2629) + %1078 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2630) + %1079 = "ttir.transpose"(%1077, %1078) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2630) + %1080 = tensor.empty() : tensor<12x3200xf32> loc(#loc2631) + %1081 = "ttir.reshape"(%1079, %1080) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2631) + %1082 = tensor.empty() : tensor<12x3200xf32> loc(#loc2632) + %1083 = "ttir.matmul"(%1081, %arg348, %1082) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2632) + %1084 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2633) + %1085 = "ttir.unsqueeze"(%1083, %1084) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2633) + %1086 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2634) + %1087 = "ttir.add"(%961, %1085, %1086) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2634) + %1088 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2635) + %1089 = "ttir.multiply"(%1087, %1087, %1088) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2635) + %1090 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2636) + %1091 = "ttir.mean"(%1089, %1090) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2636) + %1092 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2637) + %1093 = "ttir.add"(%1091, %arg77, %1092) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2637) + %1094 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2638) + %1095 = "ttir.sqrt"(%1093, %1094) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2638) + %1096 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2639) + %1097 = "ttir.reciprocal"(%1095, %1096) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2639) + %1098 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2640) + %1099 = "ttir.multiply"(%1087, %1097, %1098) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2640) + %1100 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2641) + %1101 = "ttir.multiply"(%arg349, %1099, %1100) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2641) + %1102 = tensor.empty() : tensor<12x3200xf32> loc(#loc2642) + %1103 = "ttir.squeeze"(%1101, %1102) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2642) + %1104 = tensor.empty() : tensor<12x8640xf32> loc(#loc2643) + %1105 = "ttir.matmul"(%1103, %arg350, %1104) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc2643) + %1106 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2644) + %1107 = "ttir.unsqueeze"(%1105, %1106) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2644) + %1108 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2645) + %1109 = "ttir.sigmoid"(%1107, %1108) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2645) + %1110 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2646) + %1111 = "ttir.multiply"(%1107, %1109, %1110) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2646) + %1112 = tensor.empty() : tensor<12x8640xf32> loc(#loc2647) + %1113 = "ttir.matmul"(%1103, %arg351, %1112) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc2647) + %1114 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2648) + %1115 = "ttir.unsqueeze"(%1113, %1114) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2648) + %1116 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2649) + %1117 = "ttir.multiply"(%1111, %1115, %1116) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2649) + %1118 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2650) + %1119 = "ttir.matmul"(%1117, %arg352, %1118) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2650) + %1120 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2651) + %1121 = "ttir.add"(%1087, %1119, %1120) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2651) + %1122 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2652) + %1123 = "ttir.multiply"(%1121, %1121, %1122) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2652) + %1124 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2653) + %1125 = "ttir.mean"(%1123, %1124) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2653) + %1126 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2654) + %1127 = "ttir.add"(%1125, %arg78, %1126) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2654) + %1128 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2655) + %1129 = "ttir.sqrt"(%1127, %1128) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2655) + %1130 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2656) + %1131 = "ttir.reciprocal"(%1129, %1130) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2656) + %1132 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2657) + %1133 = "ttir.multiply"(%1121, %1131, %1132) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2657) + %1134 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2658) + %1135 = "ttir.multiply"(%arg353, %1133, %1134) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2658) + %1136 = tensor.empty() : tensor<12x3200xf32> loc(#loc2659) + %1137 = "ttir.squeeze"(%1135, %1136) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2659) + %1138 = tensor.empty() : tensor<12x3200xf32> loc(#loc2660) + %1139 = "ttir.matmul"(%1137, %arg354, %1138) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2660) + %1140 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2661) + %1141 = "ttir.reshape"(%1139, %1140) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2661) + %1142 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2662) + %1143 = "ttir.transpose"(%1141, %1142) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2662) + %1144 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2663) + %1145 = "ttir.concat"(%arg79, %arg79, %1144) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2663) + %1146 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2664) + %1147 = "ttir.sin"(%1145, %1146) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2664) + %1148 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc2665) + %1149 = "ttir.unsqueeze"(%1147, %1148) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc2665) + %1150 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2666) + %1151 = "ttir.multiply"(%1143, %1149, %1150) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2666) + %1152 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2667) + %1153 = "ttir.transpose"(%1143, %1152) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2667) + %1154 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2668) + %1155 = "ttir.matmul"(%arg80, %1153, %1154) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2668) + %1156 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2669) + %1157 = "ttir.transpose"(%1155, %1156) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2669) + %1158 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2670) + %1159 = "ttir.multiply"(%1157, %arg81, %1158) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2670) + %1160 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2671) + %1161 = "ttir.transpose"(%1143, %1160) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2671) + %1162 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2672) + %1163 = "ttir.matmul"(%arg82, %1161, %1162) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2672) + %1164 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2673) + %1165 = "ttir.transpose"(%1163, %1164) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2673) + %1166 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2674) + %1167 = "ttir.concat"(%1159, %1165, %1166) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2674) + %1168 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2675) + %1169 = "ttir.cos"(%1145, %1168) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2675) + %1170 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc2676) + %1171 = "ttir.unsqueeze"(%1169, %1170) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc2676) + %1172 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2677) + %1173 = "ttir.multiply"(%1167, %1171, %1172) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2677) + %1174 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2678) + %1175 = "ttir.add"(%1151, %1173, %1174) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2678) + %1176 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2679) + %1177 = "ttir.squeeze"(%1175, %1176) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2679) + %1178 = tensor.empty() : tensor<12x3200xf32> loc(#loc2680) + %1179 = "ttir.matmul"(%1137, %arg355, %1178) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2680) + %1180 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2681) + %1181 = "ttir.reshape"(%1179, %1180) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2681) + %1182 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2682) + %1183 = "ttir.transpose"(%1181, %1182) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2682) + %1184 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2683) + %1185 = "ttir.multiply"(%1183, %1149, %1184) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2683) + %1186 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2684) + %1187 = "ttir.transpose"(%1183, %1186) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2684) + %1188 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2685) + %1189 = "ttir.matmul"(%arg83, %1187, %1188) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2685) + %1190 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2686) + %1191 = "ttir.transpose"(%1189, %1190) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2686) + %1192 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2687) + %1193 = "ttir.multiply"(%1191, %arg84, %1192) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2687) + %1194 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2688) + %1195 = "ttir.transpose"(%1183, %1194) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2688) + %1196 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2689) + %1197 = "ttir.matmul"(%arg85, %1195, %1196) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2689) + %1198 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2690) + %1199 = "ttir.transpose"(%1197, %1198) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2690) + %1200 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2691) + %1201 = "ttir.concat"(%1193, %1199, %1200) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2691) + %1202 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2692) + %1203 = "ttir.multiply"(%1201, %1171, %1202) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2692) + %1204 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2693) + %1205 = "ttir.add"(%1185, %1203, %1204) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2693) + %1206 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2694) + %1207 = "ttir.squeeze"(%1205, %1206) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2694) + %1208 = tensor.empty() : tensor<32x100x12xf32> loc(#loc2695) + %1209 = "ttir.transpose"(%1207, %1208) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc2695) + %1210 = tensor.empty() : tensor<32x12x12xf32> loc(#loc2696) + %1211 = "ttir.matmul"(%1177, %1209, %1210) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc2696) + %1212 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2697) + %1213 = "ttir.unsqueeze"(%1211, %1212) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2697) + %1214 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2698) + %1215 = "ttir.multiply"(%1213, %arg86, %1214) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2698) + %1216 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2699) + %1217 = "ttir.add"(%1215, %arg87, %1216) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2699) + %1218 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2700) + %1219 = "ttir.softmax"(%1217, %1218) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2700) + %1220 = tensor.empty() : tensor<32x12x12xf32> loc(#loc2701) + %1221 = "ttir.squeeze"(%1219, %1220) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc2701) + %1222 = tensor.empty() : tensor<12x3200xf32> loc(#loc2702) + %1223 = "ttir.matmul"(%1137, %arg356, %1222) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2702) + %1224 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2703) + %1225 = "ttir.reshape"(%1223, %1224) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2703) + %1226 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2704) + %1227 = "ttir.transpose"(%1225, %1226) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2704) + %1228 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2705) + %1229 = "ttir.transpose"(%1227, %1228) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2705) + %1230 = tensor.empty() : tensor<32x100x12xf32> loc(#loc2706) + %1231 = "ttir.squeeze"(%1229, %1230) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc2706) + %1232 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2707) + %1233 = "ttir.transpose"(%1231, %1232) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2707) + %1234 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2708) + %1235 = "ttir.matmul"(%1221, %1233, %1234) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2708) + %1236 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2709) + %1237 = "ttir.unsqueeze"(%1235, %1236) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2709) + %1238 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2710) + %1239 = "ttir.transpose"(%1237, %1238) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2710) + %1240 = tensor.empty() : tensor<12x3200xf32> loc(#loc2711) + %1241 = "ttir.reshape"(%1239, %1240) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2711) + %1242 = tensor.empty() : tensor<12x3200xf32> loc(#loc2712) + %1243 = "ttir.matmul"(%1241, %arg357, %1242) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2712) + %1244 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2713) + %1245 = "ttir.unsqueeze"(%1243, %1244) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2713) + %1246 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2714) + %1247 = "ttir.add"(%1121, %1245, %1246) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2714) + %1248 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2715) + %1249 = "ttir.multiply"(%1247, %1247, %1248) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2715) + %1250 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2716) + %1251 = "ttir.mean"(%1249, %1250) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2716) + %1252 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2717) + %1253 = "ttir.add"(%1251, %arg88, %1252) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2717) + %1254 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2718) + %1255 = "ttir.sqrt"(%1253, %1254) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2718) + %1256 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2719) + %1257 = "ttir.reciprocal"(%1255, %1256) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2719) + %1258 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2720) + %1259 = "ttir.multiply"(%1247, %1257, %1258) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2720) + %1260 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2721) + %1261 = "ttir.multiply"(%arg358, %1259, %1260) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2721) + %1262 = tensor.empty() : tensor<12x3200xf32> loc(#loc2722) + %1263 = "ttir.squeeze"(%1261, %1262) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2722) + %1264 = tensor.empty() : tensor<12x8640xf32> loc(#loc2723) + %1265 = "ttir.matmul"(%1263, %arg359, %1264) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc2723) + %1266 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2724) + %1267 = "ttir.unsqueeze"(%1265, %1266) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2724) + %1268 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2725) + %1269 = "ttir.sigmoid"(%1267, %1268) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2725) + %1270 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2726) + %1271 = "ttir.multiply"(%1267, %1269, %1270) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2726) + %1272 = tensor.empty() : tensor<12x8640xf32> loc(#loc2727) + %1273 = "ttir.matmul"(%1263, %arg360, %1272) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc2727) + %1274 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2728) + %1275 = "ttir.unsqueeze"(%1273, %1274) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2728) + %1276 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2729) + %1277 = "ttir.multiply"(%1271, %1275, %1276) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2729) + %1278 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2730) + %1279 = "ttir.matmul"(%1277, %arg361, %1278) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2730) + %1280 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2731) + %1281 = "ttir.add"(%1247, %1279, %1280) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2731) + %1282 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2732) + %1283 = "ttir.multiply"(%1281, %1281, %1282) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2732) + %1284 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2733) + %1285 = "ttir.mean"(%1283, %1284) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2733) + %1286 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2734) + %1287 = "ttir.add"(%1285, %arg89, %1286) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2734) + %1288 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2735) + %1289 = "ttir.sqrt"(%1287, %1288) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2735) + %1290 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2736) + %1291 = "ttir.reciprocal"(%1289, %1290) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2736) + %1292 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2737) + %1293 = "ttir.multiply"(%1281, %1291, %1292) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2737) + %1294 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2738) + %1295 = "ttir.multiply"(%arg362, %1293, %1294) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2738) + %1296 = tensor.empty() : tensor<12x3200xf32> loc(#loc2739) + %1297 = "ttir.squeeze"(%1295, %1296) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2739) + %1298 = tensor.empty() : tensor<12x3200xf32> loc(#loc2740) + %1299 = "ttir.matmul"(%1297, %arg363, %1298) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2740) + %1300 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2741) + %1301 = "ttir.reshape"(%1299, %1300) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2741) + %1302 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2742) + %1303 = "ttir.transpose"(%1301, %1302) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2742) + %1304 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2743) + %1305 = "ttir.concat"(%arg90, %arg90, %1304) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2743) + %1306 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2744) + %1307 = "ttir.sin"(%1305, %1306) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2744) + %1308 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc2745) + %1309 = "ttir.unsqueeze"(%1307, %1308) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc2745) + %1310 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2746) + %1311 = "ttir.multiply"(%1303, %1309, %1310) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2746) + %1312 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2747) + %1313 = "ttir.transpose"(%1303, %1312) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2747) + %1314 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2748) + %1315 = "ttir.matmul"(%arg91, %1313, %1314) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2748) + %1316 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2749) + %1317 = "ttir.transpose"(%1315, %1316) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2749) + %1318 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2750) + %1319 = "ttir.multiply"(%1317, %arg92, %1318) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2750) + %1320 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2751) + %1321 = "ttir.transpose"(%1303, %1320) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2751) + %1322 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2752) + %1323 = "ttir.matmul"(%arg93, %1321, %1322) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2752) + %1324 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2753) + %1325 = "ttir.transpose"(%1323, %1324) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2753) + %1326 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2754) + %1327 = "ttir.concat"(%1319, %1325, %1326) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2754) + %1328 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2755) + %1329 = "ttir.cos"(%1305, %1328) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2755) + %1330 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc2756) + %1331 = "ttir.unsqueeze"(%1329, %1330) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc2756) + %1332 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2757) + %1333 = "ttir.multiply"(%1327, %1331, %1332) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2757) + %1334 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2758) + %1335 = "ttir.add"(%1311, %1333, %1334) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2758) + %1336 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2759) + %1337 = "ttir.squeeze"(%1335, %1336) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2759) + %1338 = tensor.empty() : tensor<12x3200xf32> loc(#loc2760) + %1339 = "ttir.matmul"(%1297, %arg364, %1338) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2760) + %1340 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2761) + %1341 = "ttir.reshape"(%1339, %1340) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2761) + %1342 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2762) + %1343 = "ttir.transpose"(%1341, %1342) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2762) + %1344 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2763) + %1345 = "ttir.multiply"(%1343, %1309, %1344) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2763) + %1346 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2764) + %1347 = "ttir.transpose"(%1343, %1346) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2764) + %1348 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2765) + %1349 = "ttir.matmul"(%arg94, %1347, %1348) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2765) + %1350 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2766) + %1351 = "ttir.transpose"(%1349, %1350) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2766) + %1352 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2767) + %1353 = "ttir.multiply"(%1351, %arg95, %1352) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2767) + %1354 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2768) + %1355 = "ttir.transpose"(%1343, %1354) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2768) + %1356 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2769) + %1357 = "ttir.matmul"(%arg96, %1355, %1356) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2769) + %1358 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2770) + %1359 = "ttir.transpose"(%1357, %1358) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2770) + %1360 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2771) + %1361 = "ttir.concat"(%1353, %1359, %1360) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2771) + %1362 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2772) + %1363 = "ttir.multiply"(%1361, %1331, %1362) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2772) + %1364 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2773) + %1365 = "ttir.add"(%1345, %1363, %1364) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2773) + %1366 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2774) + %1367 = "ttir.squeeze"(%1365, %1366) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2774) + %1368 = tensor.empty() : tensor<32x100x12xf32> loc(#loc2775) + %1369 = "ttir.transpose"(%1367, %1368) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc2775) + %1370 = tensor.empty() : tensor<32x12x12xf32> loc(#loc2776) + %1371 = "ttir.matmul"(%1337, %1369, %1370) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc2776) + %1372 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2777) + %1373 = "ttir.unsqueeze"(%1371, %1372) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2777) + %1374 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2778) + %1375 = "ttir.multiply"(%1373, %arg97, %1374) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2778) + %1376 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2779) + %1377 = "ttir.add"(%1375, %arg98, %1376) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2779) + %1378 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2780) + %1379 = "ttir.softmax"(%1377, %1378) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2780) + %1380 = tensor.empty() : tensor<32x12x12xf32> loc(#loc2781) + %1381 = "ttir.squeeze"(%1379, %1380) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc2781) + %1382 = tensor.empty() : tensor<12x3200xf32> loc(#loc2782) + %1383 = "ttir.matmul"(%1297, %arg365, %1382) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2782) + %1384 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2783) + %1385 = "ttir.reshape"(%1383, %1384) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2783) + %1386 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2784) + %1387 = "ttir.transpose"(%1385, %1386) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2784) + %1388 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2785) + %1389 = "ttir.transpose"(%1387, %1388) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2785) + %1390 = tensor.empty() : tensor<32x100x12xf32> loc(#loc2786) + %1391 = "ttir.squeeze"(%1389, %1390) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc2786) + %1392 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2787) + %1393 = "ttir.transpose"(%1391, %1392) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2787) + %1394 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2788) + %1395 = "ttir.matmul"(%1381, %1393, %1394) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2788) + %1396 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2789) + %1397 = "ttir.unsqueeze"(%1395, %1396) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2789) + %1398 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2790) + %1399 = "ttir.transpose"(%1397, %1398) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2790) + %1400 = tensor.empty() : tensor<12x3200xf32> loc(#loc2791) + %1401 = "ttir.reshape"(%1399, %1400) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2791) + %1402 = tensor.empty() : tensor<12x3200xf32> loc(#loc2792) + %1403 = "ttir.matmul"(%1401, %arg366, %1402) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2792) + %1404 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2793) + %1405 = "ttir.unsqueeze"(%1403, %1404) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2793) + %1406 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2794) + %1407 = "ttir.add"(%1281, %1405, %1406) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2794) + %1408 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2795) + %1409 = "ttir.multiply"(%1407, %1407, %1408) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2795) + %1410 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2796) + %1411 = "ttir.mean"(%1409, %1410) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2796) + %1412 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2797) + %1413 = "ttir.add"(%1411, %arg99, %1412) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2797) + %1414 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2798) + %1415 = "ttir.sqrt"(%1413, %1414) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2798) + %1416 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2799) + %1417 = "ttir.reciprocal"(%1415, %1416) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2799) + %1418 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2800) + %1419 = "ttir.multiply"(%1407, %1417, %1418) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2800) + %1420 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2801) + %1421 = "ttir.multiply"(%arg367, %1419, %1420) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2801) + %1422 = tensor.empty() : tensor<12x3200xf32> loc(#loc2802) + %1423 = "ttir.squeeze"(%1421, %1422) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2802) + %1424 = tensor.empty() : tensor<12x8640xf32> loc(#loc2803) + %1425 = "ttir.matmul"(%1423, %arg368, %1424) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc2803) + %1426 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2804) + %1427 = "ttir.unsqueeze"(%1425, %1426) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2804) + %1428 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2805) + %1429 = "ttir.sigmoid"(%1427, %1428) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2805) + %1430 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2806) + %1431 = "ttir.multiply"(%1427, %1429, %1430) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2806) + %1432 = tensor.empty() : tensor<12x8640xf32> loc(#loc2807) + %1433 = "ttir.matmul"(%1423, %arg369, %1432) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc2807) + %1434 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2808) + %1435 = "ttir.unsqueeze"(%1433, %1434) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2808) + %1436 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2809) + %1437 = "ttir.multiply"(%1431, %1435, %1436) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2809) + %1438 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2810) + %1439 = "ttir.matmul"(%1437, %arg370, %1438) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2810) + %1440 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2811) + %1441 = "ttir.add"(%1407, %1439, %1440) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2811) + %1442 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2812) + %1443 = "ttir.multiply"(%1441, %1441, %1442) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2812) + %1444 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2813) + %1445 = "ttir.mean"(%1443, %1444) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2813) + %1446 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2814) + %1447 = "ttir.add"(%1445, %arg100, %1446) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2814) + %1448 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2815) + %1449 = "ttir.sqrt"(%1447, %1448) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2815) + %1450 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2816) + %1451 = "ttir.reciprocal"(%1449, %1450) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2816) + %1452 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2817) + %1453 = "ttir.multiply"(%1441, %1451, %1452) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2817) + %1454 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2818) + %1455 = "ttir.multiply"(%arg371, %1453, %1454) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2818) + %1456 = tensor.empty() : tensor<12x3200xf32> loc(#loc2819) + %1457 = "ttir.squeeze"(%1455, %1456) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2819) + %1458 = tensor.empty() : tensor<12x3200xf32> loc(#loc2820) + %1459 = "ttir.matmul"(%1457, %arg372, %1458) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2820) + %1460 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2821) + %1461 = "ttir.reshape"(%1459, %1460) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2821) + %1462 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2822) + %1463 = "ttir.transpose"(%1461, %1462) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2822) + %1464 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2823) + %1465 = "ttir.concat"(%arg101, %arg101, %1464) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2823) + %1466 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2824) + %1467 = "ttir.sin"(%1465, %1466) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2824) + %1468 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc2825) + %1469 = "ttir.unsqueeze"(%1467, %1468) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc2825) + %1470 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2826) + %1471 = "ttir.multiply"(%1463, %1469, %1470) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2826) + %1472 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2827) + %1473 = "ttir.transpose"(%1463, %1472) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2827) + %1474 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2828) + %1475 = "ttir.matmul"(%arg102, %1473, %1474) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2828) + %1476 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2829) + %1477 = "ttir.transpose"(%1475, %1476) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2829) + %1478 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2830) + %1479 = "ttir.multiply"(%1477, %arg103, %1478) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2830) + %1480 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2831) + %1481 = "ttir.transpose"(%1463, %1480) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2831) + %1482 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2832) + %1483 = "ttir.matmul"(%arg104, %1481, %1482) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2832) + %1484 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2833) + %1485 = "ttir.transpose"(%1483, %1484) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2833) + %1486 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2834) + %1487 = "ttir.concat"(%1479, %1485, %1486) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2834) + %1488 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2835) + %1489 = "ttir.cos"(%1465, %1488) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2835) + %1490 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc2836) + %1491 = "ttir.unsqueeze"(%1489, %1490) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc2836) + %1492 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2837) + %1493 = "ttir.multiply"(%1487, %1491, %1492) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2837) + %1494 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2838) + %1495 = "ttir.add"(%1471, %1493, %1494) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2838) + %1496 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2839) + %1497 = "ttir.squeeze"(%1495, %1496) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2839) + %1498 = tensor.empty() : tensor<12x3200xf32> loc(#loc2840) + %1499 = "ttir.matmul"(%1457, %arg373, %1498) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2840) + %1500 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2841) + %1501 = "ttir.reshape"(%1499, %1500) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2841) + %1502 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2842) + %1503 = "ttir.transpose"(%1501, %1502) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2842) + %1504 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2843) + %1505 = "ttir.multiply"(%1503, %1469, %1504) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2843) + %1506 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2844) + %1507 = "ttir.transpose"(%1503, %1506) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2844) + %1508 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2845) + %1509 = "ttir.matmul"(%arg105, %1507, %1508) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2845) + %1510 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2846) + %1511 = "ttir.transpose"(%1509, %1510) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2846) + %1512 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2847) + %1513 = "ttir.multiply"(%1511, %arg106, %1512) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2847) + %1514 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2848) + %1515 = "ttir.transpose"(%1503, %1514) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2848) + %1516 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2849) + %1517 = "ttir.matmul"(%arg107, %1515, %1516) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2849) + %1518 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2850) + %1519 = "ttir.transpose"(%1517, %1518) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2850) + %1520 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2851) + %1521 = "ttir.concat"(%1513, %1519, %1520) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2851) + %1522 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2852) + %1523 = "ttir.multiply"(%1521, %1491, %1522) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2852) + %1524 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2853) + %1525 = "ttir.add"(%1505, %1523, %1524) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2853) + %1526 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2854) + %1527 = "ttir.squeeze"(%1525, %1526) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2854) + %1528 = tensor.empty() : tensor<32x100x12xf32> loc(#loc2855) + %1529 = "ttir.transpose"(%1527, %1528) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc2855) + %1530 = tensor.empty() : tensor<32x12x12xf32> loc(#loc2856) + %1531 = "ttir.matmul"(%1497, %1529, %1530) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc2856) + %1532 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2857) + %1533 = "ttir.unsqueeze"(%1531, %1532) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2857) + %1534 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2858) + %1535 = "ttir.multiply"(%1533, %arg108, %1534) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2858) + %1536 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2859) + %1537 = "ttir.add"(%1535, %arg109, %1536) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2859) + %1538 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2860) + %1539 = "ttir.softmax"(%1537, %1538) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2860) + %1540 = tensor.empty() : tensor<32x12x12xf32> loc(#loc2861) + %1541 = "ttir.squeeze"(%1539, %1540) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc2861) + %1542 = tensor.empty() : tensor<12x3200xf32> loc(#loc2862) + %1543 = "ttir.matmul"(%1457, %arg374, %1542) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2862) + %1544 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2863) + %1545 = "ttir.reshape"(%1543, %1544) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2863) + %1546 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2864) + %1547 = "ttir.transpose"(%1545, %1546) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2864) + %1548 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2865) + %1549 = "ttir.transpose"(%1547, %1548) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2865) + %1550 = tensor.empty() : tensor<32x100x12xf32> loc(#loc2866) + %1551 = "ttir.squeeze"(%1549, %1550) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc2866) + %1552 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2867) + %1553 = "ttir.transpose"(%1551, %1552) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2867) + %1554 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2868) + %1555 = "ttir.matmul"(%1541, %1553, %1554) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2868) + %1556 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2869) + %1557 = "ttir.unsqueeze"(%1555, %1556) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2869) + %1558 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2870) + %1559 = "ttir.transpose"(%1557, %1558) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2870) + %1560 = tensor.empty() : tensor<12x3200xf32> loc(#loc2871) + %1561 = "ttir.reshape"(%1559, %1560) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2871) + %1562 = tensor.empty() : tensor<12x3200xf32> loc(#loc2872) + %1563 = "ttir.matmul"(%1561, %arg375, %1562) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2872) + %1564 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2873) + %1565 = "ttir.unsqueeze"(%1563, %1564) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2873) + %1566 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2874) + %1567 = "ttir.add"(%1441, %1565, %1566) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2874) + %1568 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2875) + %1569 = "ttir.multiply"(%1567, %1567, %1568) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2875) + %1570 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2876) + %1571 = "ttir.mean"(%1569, %1570) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2876) + %1572 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2877) + %1573 = "ttir.add"(%1571, %arg110, %1572) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2877) + %1574 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2878) + %1575 = "ttir.sqrt"(%1573, %1574) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2878) + %1576 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2879) + %1577 = "ttir.reciprocal"(%1575, %1576) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2879) + %1578 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2880) + %1579 = "ttir.multiply"(%1567, %1577, %1578) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2880) + %1580 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2881) + %1581 = "ttir.multiply"(%arg376, %1579, %1580) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2881) + %1582 = tensor.empty() : tensor<12x3200xf32> loc(#loc2882) + %1583 = "ttir.squeeze"(%1581, %1582) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2882) + %1584 = tensor.empty() : tensor<12x8640xf32> loc(#loc2883) + %1585 = "ttir.matmul"(%1583, %arg377, %1584) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc2883) + %1586 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2884) + %1587 = "ttir.unsqueeze"(%1585, %1586) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2884) + %1588 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2885) + %1589 = "ttir.sigmoid"(%1587, %1588) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2885) + %1590 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2886) + %1591 = "ttir.multiply"(%1587, %1589, %1590) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2886) + %1592 = tensor.empty() : tensor<12x8640xf32> loc(#loc2887) + %1593 = "ttir.matmul"(%1583, %arg378, %1592) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc2887) + %1594 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2888) + %1595 = "ttir.unsqueeze"(%1593, %1594) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2888) + %1596 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2889) + %1597 = "ttir.multiply"(%1591, %1595, %1596) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2889) + %1598 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2890) + %1599 = "ttir.matmul"(%1597, %arg379, %1598) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2890) + %1600 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2891) + %1601 = "ttir.add"(%1567, %1599, %1600) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2891) + %1602 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2892) + %1603 = "ttir.multiply"(%1601, %1601, %1602) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2892) + %1604 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2893) + %1605 = "ttir.mean"(%1603, %1604) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2893) + %1606 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2894) + %1607 = "ttir.add"(%1605, %arg111, %1606) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2894) + %1608 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2895) + %1609 = "ttir.sqrt"(%1607, %1608) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2895) + %1610 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2896) + %1611 = "ttir.reciprocal"(%1609, %1610) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2896) + %1612 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2897) + %1613 = "ttir.multiply"(%1601, %1611, %1612) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2897) + %1614 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2898) + %1615 = "ttir.multiply"(%arg380, %1613, %1614) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2898) + %1616 = tensor.empty() : tensor<12x3200xf32> loc(#loc2899) + %1617 = "ttir.squeeze"(%1615, %1616) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2899) + %1618 = tensor.empty() : tensor<12x3200xf32> loc(#loc2900) + %1619 = "ttir.matmul"(%1617, %arg381, %1618) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2900) + %1620 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2901) + %1621 = "ttir.reshape"(%1619, %1620) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2901) + %1622 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2902) + %1623 = "ttir.transpose"(%1621, %1622) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2902) + %1624 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2903) + %1625 = "ttir.concat"(%arg112, %arg112, %1624) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2903) + %1626 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2904) + %1627 = "ttir.sin"(%1625, %1626) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2904) + %1628 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc2905) + %1629 = "ttir.unsqueeze"(%1627, %1628) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc2905) + %1630 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2906) + %1631 = "ttir.multiply"(%1623, %1629, %1630) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2906) + %1632 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2907) + %1633 = "ttir.transpose"(%1623, %1632) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2907) + %1634 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2908) + %1635 = "ttir.matmul"(%arg113, %1633, %1634) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2908) + %1636 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2909) + %1637 = "ttir.transpose"(%1635, %1636) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2909) + %1638 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2910) + %1639 = "ttir.multiply"(%1637, %arg114, %1638) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2910) + %1640 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2911) + %1641 = "ttir.transpose"(%1623, %1640) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2911) + %1642 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2912) + %1643 = "ttir.matmul"(%arg115, %1641, %1642) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2912) + %1644 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2913) + %1645 = "ttir.transpose"(%1643, %1644) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2913) + %1646 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2914) + %1647 = "ttir.concat"(%1639, %1645, %1646) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2914) + %1648 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2915) + %1649 = "ttir.cos"(%1625, %1648) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2915) + %1650 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc2916) + %1651 = "ttir.unsqueeze"(%1649, %1650) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc2916) + %1652 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2917) + %1653 = "ttir.multiply"(%1647, %1651, %1652) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2917) + %1654 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2918) + %1655 = "ttir.add"(%1631, %1653, %1654) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2918) + %1656 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2919) + %1657 = "ttir.squeeze"(%1655, %1656) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2919) + %1658 = tensor.empty() : tensor<12x3200xf32> loc(#loc2920) + %1659 = "ttir.matmul"(%1617, %arg382, %1658) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2920) + %1660 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2921) + %1661 = "ttir.reshape"(%1659, %1660) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2921) + %1662 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2922) + %1663 = "ttir.transpose"(%1661, %1662) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2922) + %1664 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2923) + %1665 = "ttir.multiply"(%1663, %1629, %1664) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2923) + %1666 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2924) + %1667 = "ttir.transpose"(%1663, %1666) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2924) + %1668 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2925) + %1669 = "ttir.matmul"(%arg116, %1667, %1668) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2925) + %1670 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2926) + %1671 = "ttir.transpose"(%1669, %1670) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2926) + %1672 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2927) + %1673 = "ttir.multiply"(%1671, %arg117, %1672) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2927) + %1674 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2928) + %1675 = "ttir.transpose"(%1663, %1674) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2928) + %1676 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2929) + %1677 = "ttir.matmul"(%arg118, %1675, %1676) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2929) + %1678 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2930) + %1679 = "ttir.transpose"(%1677, %1678) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2930) + %1680 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2931) + %1681 = "ttir.concat"(%1673, %1679, %1680) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2931) + %1682 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2932) + %1683 = "ttir.multiply"(%1681, %1651, %1682) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2932) + %1684 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2933) + %1685 = "ttir.add"(%1665, %1683, %1684) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2933) + %1686 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2934) + %1687 = "ttir.squeeze"(%1685, %1686) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2934) + %1688 = tensor.empty() : tensor<32x100x12xf32> loc(#loc2935) + %1689 = "ttir.transpose"(%1687, %1688) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc2935) + %1690 = tensor.empty() : tensor<32x12x12xf32> loc(#loc2936) + %1691 = "ttir.matmul"(%1657, %1689, %1690) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc2936) + %1692 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2937) + %1693 = "ttir.unsqueeze"(%1691, %1692) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2937) + %1694 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2938) + %1695 = "ttir.multiply"(%1693, %arg119, %1694) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2938) + %1696 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2939) + %1697 = "ttir.add"(%1695, %arg120, %1696) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2939) + %1698 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc2940) + %1699 = "ttir.softmax"(%1697, %1698) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc2940) + %1700 = tensor.empty() : tensor<32x12x12xf32> loc(#loc2941) + %1701 = "ttir.squeeze"(%1699, %1700) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc2941) + %1702 = tensor.empty() : tensor<12x3200xf32> loc(#loc2942) + %1703 = "ttir.matmul"(%1617, %arg383, %1702) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2942) + %1704 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2943) + %1705 = "ttir.reshape"(%1703, %1704) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2943) + %1706 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2944) + %1707 = "ttir.transpose"(%1705, %1706) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2944) + %1708 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2945) + %1709 = "ttir.transpose"(%1707, %1708) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2945) + %1710 = tensor.empty() : tensor<32x100x12xf32> loc(#loc2946) + %1711 = "ttir.squeeze"(%1709, %1710) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc2946) + %1712 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2947) + %1713 = "ttir.transpose"(%1711, %1712) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2947) + %1714 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2948) + %1715 = "ttir.matmul"(%1701, %1713, %1714) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2948) + %1716 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2949) + %1717 = "ttir.unsqueeze"(%1715, %1716) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2949) + %1718 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2950) + %1719 = "ttir.transpose"(%1717, %1718) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2950) + %1720 = tensor.empty() : tensor<12x3200xf32> loc(#loc2951) + %1721 = "ttir.reshape"(%1719, %1720) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2951) + %1722 = tensor.empty() : tensor<12x3200xf32> loc(#loc2952) + %1723 = "ttir.matmul"(%1721, %arg384, %1722) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2952) + %1724 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2953) + %1725 = "ttir.unsqueeze"(%1723, %1724) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2953) + %1726 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2954) + %1727 = "ttir.add"(%1601, %1725, %1726) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2954) + %1728 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2955) + %1729 = "ttir.multiply"(%1727, %1727, %1728) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2955) + %1730 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2956) + %1731 = "ttir.mean"(%1729, %1730) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2956) + %1732 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2957) + %1733 = "ttir.add"(%1731, %arg121, %1732) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2957) + %1734 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2958) + %1735 = "ttir.sqrt"(%1733, %1734) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2958) + %1736 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2959) + %1737 = "ttir.reciprocal"(%1735, %1736) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2959) + %1738 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2960) + %1739 = "ttir.multiply"(%1727, %1737, %1738) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2960) + %1740 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2961) + %1741 = "ttir.multiply"(%arg385, %1739, %1740) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2961) + %1742 = tensor.empty() : tensor<12x3200xf32> loc(#loc2962) + %1743 = "ttir.squeeze"(%1741, %1742) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2962) + %1744 = tensor.empty() : tensor<12x8640xf32> loc(#loc2963) + %1745 = "ttir.matmul"(%1743, %arg386, %1744) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc2963) + %1746 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2964) + %1747 = "ttir.unsqueeze"(%1745, %1746) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2964) + %1748 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2965) + %1749 = "ttir.sigmoid"(%1747, %1748) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2965) + %1750 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2966) + %1751 = "ttir.multiply"(%1747, %1749, %1750) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2966) + %1752 = tensor.empty() : tensor<12x8640xf32> loc(#loc2967) + %1753 = "ttir.matmul"(%1743, %arg387, %1752) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc2967) + %1754 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2968) + %1755 = "ttir.unsqueeze"(%1753, %1754) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2968) + %1756 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc2969) + %1757 = "ttir.multiply"(%1751, %1755, %1756) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc2969) + %1758 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2970) + %1759 = "ttir.matmul"(%1757, %arg388, %1758) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2970) + %1760 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2971) + %1761 = "ttir.add"(%1727, %1759, %1760) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2971) + %1762 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2972) + %1763 = "ttir.multiply"(%1761, %1761, %1762) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2972) + %1764 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2973) + %1765 = "ttir.mean"(%1763, %1764) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2973) + %1766 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2974) + %1767 = "ttir.add"(%1765, %arg122, %1766) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2974) + %1768 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2975) + %1769 = "ttir.sqrt"(%1767, %1768) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2975) + %1770 = tensor.empty() : tensor<1x12x1xf32> loc(#loc2976) + %1771 = "ttir.reciprocal"(%1769, %1770) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc2976) + %1772 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2977) + %1773 = "ttir.multiply"(%1761, %1771, %1772) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2977) + %1774 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc2978) + %1775 = "ttir.multiply"(%arg389, %1773, %1774) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc2978) + %1776 = tensor.empty() : tensor<12x3200xf32> loc(#loc2979) + %1777 = "ttir.squeeze"(%1775, %1776) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2979) + %1778 = tensor.empty() : tensor<12x3200xf32> loc(#loc2980) + %1779 = "ttir.matmul"(%1777, %arg390, %1778) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc2980) + %1780 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc2981) + %1781 = "ttir.reshape"(%1779, %1780) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc2981) + %1782 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2982) + %1783 = "ttir.transpose"(%1781, %1782) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2982) + %1784 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2983) + %1785 = "ttir.concat"(%arg123, %arg123, %1784) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2983) + %1786 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2984) + %1787 = "ttir.sin"(%1785, %1786) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2984) + %1788 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc2985) + %1789 = "ttir.unsqueeze"(%1787, %1788) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc2985) + %1790 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2986) + %1791 = "ttir.multiply"(%1783, %1789, %1790) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2986) + %1792 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2987) + %1793 = "ttir.transpose"(%1783, %1792) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2987) + %1794 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2988) + %1795 = "ttir.matmul"(%arg124, %1793, %1794) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2988) + %1796 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2989) + %1797 = "ttir.transpose"(%1795, %1796) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2989) + %1798 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2990) + %1799 = "ttir.multiply"(%1797, %arg125, %1798) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2990) + %1800 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc2991) + %1801 = "ttir.transpose"(%1783, %1800) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc2991) + %1802 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc2992) + %1803 = "ttir.matmul"(%arg126, %1801, %1802) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc2992) + %1804 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc2993) + %1805 = "ttir.transpose"(%1803, %1804) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc2993) + %1806 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2994) + %1807 = "ttir.concat"(%1799, %1805, %1806) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2994) + %1808 = tensor.empty() : tensor<1x12x100xf32> loc(#loc2995) + %1809 = "ttir.cos"(%1785, %1808) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc2995) + %1810 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc2996) + %1811 = "ttir.unsqueeze"(%1809, %1810) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc2996) + %1812 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2997) + %1813 = "ttir.multiply"(%1807, %1811, %1812) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2997) + %1814 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc2998) + %1815 = "ttir.add"(%1791, %1813, %1814) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc2998) + %1816 = tensor.empty() : tensor<32x12x100xf32> loc(#loc2999) + %1817 = "ttir.squeeze"(%1815, %1816) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc2999) + %1818 = tensor.empty() : tensor<12x3200xf32> loc(#loc3000) + %1819 = "ttir.matmul"(%1777, %arg391, %1818) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3000) + %1820 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3001) + %1821 = "ttir.reshape"(%1819, %1820) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3001) + %1822 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3002) + %1823 = "ttir.transpose"(%1821, %1822) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3002) + %1824 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3003) + %1825 = "ttir.multiply"(%1823, %1789, %1824) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3003) + %1826 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3004) + %1827 = "ttir.transpose"(%1823, %1826) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3004) + %1828 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3005) + %1829 = "ttir.matmul"(%arg127, %1827, %1828) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3005) + %1830 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3006) + %1831 = "ttir.transpose"(%1829, %1830) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3006) + %1832 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3007) + %1833 = "ttir.multiply"(%1831, %arg128, %1832) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3007) + %1834 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3008) + %1835 = "ttir.transpose"(%1823, %1834) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3008) + %1836 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3009) + %1837 = "ttir.matmul"(%arg129, %1835, %1836) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3009) + %1838 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3010) + %1839 = "ttir.transpose"(%1837, %1838) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3010) + %1840 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3011) + %1841 = "ttir.concat"(%1833, %1839, %1840) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3011) + %1842 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3012) + %1843 = "ttir.multiply"(%1841, %1811, %1842) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3012) + %1844 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3013) + %1845 = "ttir.add"(%1825, %1843, %1844) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3013) + %1846 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3014) + %1847 = "ttir.squeeze"(%1845, %1846) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3014) + %1848 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3015) + %1849 = "ttir.transpose"(%1847, %1848) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3015) + %1850 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3016) + %1851 = "ttir.matmul"(%1817, %1849, %1850) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3016) + %1852 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3017) + %1853 = "ttir.unsqueeze"(%1851, %1852) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3017) + %1854 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3018) + %1855 = "ttir.multiply"(%1853, %arg130, %1854) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3018) + %1856 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3019) + %1857 = "ttir.add"(%1855, %arg131, %1856) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3019) + %1858 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3020) + %1859 = "ttir.softmax"(%1857, %1858) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3020) + %1860 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3021) + %1861 = "ttir.squeeze"(%1859, %1860) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3021) + %1862 = tensor.empty() : tensor<12x3200xf32> loc(#loc3022) + %1863 = "ttir.matmul"(%1777, %arg392, %1862) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3022) + %1864 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3023) + %1865 = "ttir.reshape"(%1863, %1864) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3023) + %1866 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3024) + %1867 = "ttir.transpose"(%1865, %1866) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3024) + %1868 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3025) + %1869 = "ttir.transpose"(%1867, %1868) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3025) + %1870 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3026) + %1871 = "ttir.squeeze"(%1869, %1870) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3026) + %1872 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3027) + %1873 = "ttir.transpose"(%1871, %1872) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3027) + %1874 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3028) + %1875 = "ttir.matmul"(%1861, %1873, %1874) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3028) + %1876 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3029) + %1877 = "ttir.unsqueeze"(%1875, %1876) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3029) + %1878 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3030) + %1879 = "ttir.transpose"(%1877, %1878) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3030) + %1880 = tensor.empty() : tensor<12x3200xf32> loc(#loc3031) + %1881 = "ttir.reshape"(%1879, %1880) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3031) + %1882 = tensor.empty() : tensor<12x3200xf32> loc(#loc3032) + %1883 = "ttir.matmul"(%1881, %arg393, %1882) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3032) + %1884 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3033) + %1885 = "ttir.unsqueeze"(%1883, %1884) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3033) + %1886 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3034) + %1887 = "ttir.add"(%1761, %1885, %1886) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3034) + %1888 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3035) + %1889 = "ttir.multiply"(%1887, %1887, %1888) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3035) + %1890 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3036) + %1891 = "ttir.mean"(%1889, %1890) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3036) + %1892 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3037) + %1893 = "ttir.add"(%1891, %arg132, %1892) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3037) + %1894 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3038) + %1895 = "ttir.sqrt"(%1893, %1894) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3038) + %1896 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3039) + %1897 = "ttir.reciprocal"(%1895, %1896) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3039) + %1898 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3040) + %1899 = "ttir.multiply"(%1887, %1897, %1898) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3040) + %1900 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3041) + %1901 = "ttir.multiply"(%arg394, %1899, %1900) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3041) + %1902 = tensor.empty() : tensor<12x3200xf32> loc(#loc3042) + %1903 = "ttir.squeeze"(%1901, %1902) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3042) + %1904 = tensor.empty() : tensor<12x8640xf32> loc(#loc3043) + %1905 = "ttir.matmul"(%1903, %arg395, %1904) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc3043) + %1906 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3044) + %1907 = "ttir.unsqueeze"(%1905, %1906) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3044) + %1908 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3045) + %1909 = "ttir.sigmoid"(%1907, %1908) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3045) + %1910 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3046) + %1911 = "ttir.multiply"(%1907, %1909, %1910) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3046) + %1912 = tensor.empty() : tensor<12x8640xf32> loc(#loc3047) + %1913 = "ttir.matmul"(%1903, %arg396, %1912) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc3047) + %1914 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3048) + %1915 = "ttir.unsqueeze"(%1913, %1914) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3048) + %1916 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3049) + %1917 = "ttir.multiply"(%1911, %1915, %1916) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3049) + %1918 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3050) + %1919 = "ttir.matmul"(%1917, %arg397, %1918) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3050) + %1920 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3051) + %1921 = "ttir.add"(%1887, %1919, %1920) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3051) + %1922 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3052) + %1923 = "ttir.multiply"(%1921, %1921, %1922) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3052) + %1924 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3053) + %1925 = "ttir.mean"(%1923, %1924) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3053) + %1926 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3054) + %1927 = "ttir.add"(%1925, %arg133, %1926) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3054) + %1928 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3055) + %1929 = "ttir.sqrt"(%1927, %1928) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3055) + %1930 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3056) + %1931 = "ttir.reciprocal"(%1929, %1930) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3056) + %1932 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3057) + %1933 = "ttir.multiply"(%1921, %1931, %1932) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3057) + %1934 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3058) + %1935 = "ttir.multiply"(%arg398, %1933, %1934) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3058) + %1936 = tensor.empty() : tensor<12x3200xf32> loc(#loc3059) + %1937 = "ttir.squeeze"(%1935, %1936) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3059) + %1938 = tensor.empty() : tensor<12x3200xf32> loc(#loc3060) + %1939 = "ttir.matmul"(%1937, %arg399, %1938) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3060) + %1940 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3061) + %1941 = "ttir.reshape"(%1939, %1940) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3061) + %1942 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3062) + %1943 = "ttir.transpose"(%1941, %1942) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3062) + %1944 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3063) + %1945 = "ttir.concat"(%arg134, %arg134, %1944) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3063) + %1946 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3064) + %1947 = "ttir.sin"(%1945, %1946) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3064) + %1948 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc3065) + %1949 = "ttir.unsqueeze"(%1947, %1948) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc3065) + %1950 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3066) + %1951 = "ttir.multiply"(%1943, %1949, %1950) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3066) + %1952 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3067) + %1953 = "ttir.transpose"(%1943, %1952) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3067) + %1954 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3068) + %1955 = "ttir.matmul"(%arg135, %1953, %1954) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3068) + %1956 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3069) + %1957 = "ttir.transpose"(%1955, %1956) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3069) + %1958 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3070) + %1959 = "ttir.multiply"(%1957, %arg136, %1958) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3070) + %1960 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3071) + %1961 = "ttir.transpose"(%1943, %1960) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3071) + %1962 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3072) + %1963 = "ttir.matmul"(%arg137, %1961, %1962) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3072) + %1964 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3073) + %1965 = "ttir.transpose"(%1963, %1964) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3073) + %1966 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3074) + %1967 = "ttir.concat"(%1959, %1965, %1966) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3074) + %1968 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3075) + %1969 = "ttir.cos"(%1945, %1968) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3075) + %1970 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc3076) + %1971 = "ttir.unsqueeze"(%1969, %1970) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc3076) + %1972 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3077) + %1973 = "ttir.multiply"(%1967, %1971, %1972) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3077) + %1974 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3078) + %1975 = "ttir.add"(%1951, %1973, %1974) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3078) + %1976 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3079) + %1977 = "ttir.squeeze"(%1975, %1976) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3079) + %1978 = tensor.empty() : tensor<12x3200xf32> loc(#loc3080) + %1979 = "ttir.matmul"(%1937, %arg400, %1978) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3080) + %1980 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3081) + %1981 = "ttir.reshape"(%1979, %1980) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3081) + %1982 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3082) + %1983 = "ttir.transpose"(%1981, %1982) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3082) + %1984 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3083) + %1985 = "ttir.multiply"(%1983, %1949, %1984) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3083) + %1986 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3084) + %1987 = "ttir.transpose"(%1983, %1986) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3084) + %1988 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3085) + %1989 = "ttir.matmul"(%arg138, %1987, %1988) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3085) + %1990 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3086) + %1991 = "ttir.transpose"(%1989, %1990) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3086) + %1992 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3087) + %1993 = "ttir.multiply"(%1991, %arg139, %1992) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3087) + %1994 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3088) + %1995 = "ttir.transpose"(%1983, %1994) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3088) + %1996 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3089) + %1997 = "ttir.matmul"(%arg140, %1995, %1996) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3089) + %1998 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3090) + %1999 = "ttir.transpose"(%1997, %1998) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3090) + %2000 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3091) + %2001 = "ttir.concat"(%1993, %1999, %2000) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3091) + %2002 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3092) + %2003 = "ttir.multiply"(%2001, %1971, %2002) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3092) + %2004 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3093) + %2005 = "ttir.add"(%1985, %2003, %2004) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3093) + %2006 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3094) + %2007 = "ttir.squeeze"(%2005, %2006) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3094) + %2008 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3095) + %2009 = "ttir.transpose"(%2007, %2008) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3095) + %2010 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3096) + %2011 = "ttir.matmul"(%1977, %2009, %2010) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3096) + %2012 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3097) + %2013 = "ttir.unsqueeze"(%2011, %2012) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3097) + %2014 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3098) + %2015 = "ttir.multiply"(%2013, %arg141, %2014) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3098) + %2016 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3099) + %2017 = "ttir.add"(%2015, %arg142, %2016) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3099) + %2018 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3100) + %2019 = "ttir.softmax"(%2017, %2018) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3100) + %2020 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3101) + %2021 = "ttir.squeeze"(%2019, %2020) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3101) + %2022 = tensor.empty() : tensor<12x3200xf32> loc(#loc3102) + %2023 = "ttir.matmul"(%1937, %arg401, %2022) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3102) + %2024 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3103) + %2025 = "ttir.reshape"(%2023, %2024) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3103) + %2026 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3104) + %2027 = "ttir.transpose"(%2025, %2026) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3104) + %2028 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3105) + %2029 = "ttir.transpose"(%2027, %2028) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3105) + %2030 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3106) + %2031 = "ttir.squeeze"(%2029, %2030) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3106) + %2032 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3107) + %2033 = "ttir.transpose"(%2031, %2032) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3107) + %2034 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3108) + %2035 = "ttir.matmul"(%2021, %2033, %2034) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3108) + %2036 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3109) + %2037 = "ttir.unsqueeze"(%2035, %2036) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3109) + %2038 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3110) + %2039 = "ttir.transpose"(%2037, %2038) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3110) + %2040 = tensor.empty() : tensor<12x3200xf32> loc(#loc3111) + %2041 = "ttir.reshape"(%2039, %2040) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3111) + %2042 = tensor.empty() : tensor<12x3200xf32> loc(#loc3112) + %2043 = "ttir.matmul"(%2041, %arg402, %2042) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3112) + %2044 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3113) + %2045 = "ttir.unsqueeze"(%2043, %2044) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3113) + %2046 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3114) + %2047 = "ttir.add"(%1921, %2045, %2046) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3114) + %2048 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3115) + %2049 = "ttir.multiply"(%2047, %2047, %2048) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3115) + %2050 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3116) + %2051 = "ttir.mean"(%2049, %2050) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3116) + %2052 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3117) + %2053 = "ttir.add"(%2051, %arg143, %2052) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3117) + %2054 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3118) + %2055 = "ttir.sqrt"(%2053, %2054) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3118) + %2056 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3119) + %2057 = "ttir.reciprocal"(%2055, %2056) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3119) + %2058 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3120) + %2059 = "ttir.multiply"(%2047, %2057, %2058) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3120) + %2060 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3121) + %2061 = "ttir.multiply"(%arg403, %2059, %2060) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3121) + %2062 = tensor.empty() : tensor<12x3200xf32> loc(#loc3122) + %2063 = "ttir.squeeze"(%2061, %2062) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3122) + %2064 = tensor.empty() : tensor<12x8640xf32> loc(#loc3123) + %2065 = "ttir.matmul"(%2063, %arg404, %2064) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc3123) + %2066 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3124) + %2067 = "ttir.unsqueeze"(%2065, %2066) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3124) + %2068 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3125) + %2069 = "ttir.sigmoid"(%2067, %2068) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3125) + %2070 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3126) + %2071 = "ttir.multiply"(%2067, %2069, %2070) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3126) + %2072 = tensor.empty() : tensor<12x8640xf32> loc(#loc3127) + %2073 = "ttir.matmul"(%2063, %arg405, %2072) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc3127) + %2074 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3128) + %2075 = "ttir.unsqueeze"(%2073, %2074) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3128) + %2076 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3129) + %2077 = "ttir.multiply"(%2071, %2075, %2076) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3129) + %2078 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3130) + %2079 = "ttir.matmul"(%2077, %arg406, %2078) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3130) + %2080 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3131) + %2081 = "ttir.add"(%2047, %2079, %2080) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3131) + %2082 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3132) + %2083 = "ttir.multiply"(%2081, %2081, %2082) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3132) + %2084 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3133) + %2085 = "ttir.mean"(%2083, %2084) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3133) + %2086 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3134) + %2087 = "ttir.add"(%2085, %arg144, %2086) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3134) + %2088 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3135) + %2089 = "ttir.sqrt"(%2087, %2088) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3135) + %2090 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3136) + %2091 = "ttir.reciprocal"(%2089, %2090) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3136) + %2092 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3137) + %2093 = "ttir.multiply"(%2081, %2091, %2092) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3137) + %2094 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3138) + %2095 = "ttir.multiply"(%arg407, %2093, %2094) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3138) + %2096 = tensor.empty() : tensor<12x3200xf32> loc(#loc3139) + %2097 = "ttir.squeeze"(%2095, %2096) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3139) + %2098 = tensor.empty() : tensor<12x3200xf32> loc(#loc3140) + %2099 = "ttir.matmul"(%2097, %arg408, %2098) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3140) + %2100 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3141) + %2101 = "ttir.reshape"(%2099, %2100) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3141) + %2102 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3142) + %2103 = "ttir.transpose"(%2101, %2102) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3142) + %2104 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3143) + %2105 = "ttir.concat"(%arg145, %arg145, %2104) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3143) + %2106 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3144) + %2107 = "ttir.sin"(%2105, %2106) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3144) + %2108 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc3145) + %2109 = "ttir.unsqueeze"(%2107, %2108) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc3145) + %2110 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3146) + %2111 = "ttir.multiply"(%2103, %2109, %2110) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3146) + %2112 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3147) + %2113 = "ttir.transpose"(%2103, %2112) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3147) + %2114 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3148) + %2115 = "ttir.matmul"(%arg146, %2113, %2114) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3148) + %2116 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3149) + %2117 = "ttir.transpose"(%2115, %2116) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3149) + %2118 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3150) + %2119 = "ttir.multiply"(%2117, %arg147, %2118) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3150) + %2120 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3151) + %2121 = "ttir.transpose"(%2103, %2120) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3151) + %2122 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3152) + %2123 = "ttir.matmul"(%arg148, %2121, %2122) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3152) + %2124 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3153) + %2125 = "ttir.transpose"(%2123, %2124) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3153) + %2126 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3154) + %2127 = "ttir.concat"(%2119, %2125, %2126) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3154) + %2128 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3155) + %2129 = "ttir.cos"(%2105, %2128) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3155) + %2130 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc3156) + %2131 = "ttir.unsqueeze"(%2129, %2130) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc3156) + %2132 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3157) + %2133 = "ttir.multiply"(%2127, %2131, %2132) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3157) + %2134 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3158) + %2135 = "ttir.add"(%2111, %2133, %2134) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3158) + %2136 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3159) + %2137 = "ttir.squeeze"(%2135, %2136) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3159) + %2138 = tensor.empty() : tensor<12x3200xf32> loc(#loc3160) + %2139 = "ttir.matmul"(%2097, %arg409, %2138) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3160) + %2140 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3161) + %2141 = "ttir.reshape"(%2139, %2140) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3161) + %2142 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3162) + %2143 = "ttir.transpose"(%2141, %2142) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3162) + %2144 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3163) + %2145 = "ttir.multiply"(%2143, %2109, %2144) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3163) + %2146 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3164) + %2147 = "ttir.transpose"(%2143, %2146) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3164) + %2148 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3165) + %2149 = "ttir.matmul"(%arg149, %2147, %2148) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3165) + %2150 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3166) + %2151 = "ttir.transpose"(%2149, %2150) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3166) + %2152 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3167) + %2153 = "ttir.multiply"(%2151, %arg150, %2152) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3167) + %2154 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3168) + %2155 = "ttir.transpose"(%2143, %2154) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3168) + %2156 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3169) + %2157 = "ttir.matmul"(%arg151, %2155, %2156) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3169) + %2158 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3170) + %2159 = "ttir.transpose"(%2157, %2158) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3170) + %2160 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3171) + %2161 = "ttir.concat"(%2153, %2159, %2160) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3171) + %2162 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3172) + %2163 = "ttir.multiply"(%2161, %2131, %2162) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3172) + %2164 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3173) + %2165 = "ttir.add"(%2145, %2163, %2164) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3173) + %2166 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3174) + %2167 = "ttir.squeeze"(%2165, %2166) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3174) + %2168 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3175) + %2169 = "ttir.transpose"(%2167, %2168) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3175) + %2170 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3176) + %2171 = "ttir.matmul"(%2137, %2169, %2170) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3176) + %2172 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3177) + %2173 = "ttir.unsqueeze"(%2171, %2172) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3177) + %2174 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3178) + %2175 = "ttir.multiply"(%2173, %arg152, %2174) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3178) + %2176 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3179) + %2177 = "ttir.add"(%2175, %arg153, %2176) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3179) + %2178 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3180) + %2179 = "ttir.softmax"(%2177, %2178) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3180) + %2180 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3181) + %2181 = "ttir.squeeze"(%2179, %2180) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3181) + %2182 = tensor.empty() : tensor<12x3200xf32> loc(#loc3182) + %2183 = "ttir.matmul"(%2097, %arg410, %2182) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3182) + %2184 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3183) + %2185 = "ttir.reshape"(%2183, %2184) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3183) + %2186 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3184) + %2187 = "ttir.transpose"(%2185, %2186) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3184) + %2188 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3185) + %2189 = "ttir.transpose"(%2187, %2188) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3185) + %2190 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3186) + %2191 = "ttir.squeeze"(%2189, %2190) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3186) + %2192 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3187) + %2193 = "ttir.transpose"(%2191, %2192) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3187) + %2194 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3188) + %2195 = "ttir.matmul"(%2181, %2193, %2194) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3188) + %2196 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3189) + %2197 = "ttir.unsqueeze"(%2195, %2196) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3189) + %2198 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3190) + %2199 = "ttir.transpose"(%2197, %2198) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3190) + %2200 = tensor.empty() : tensor<12x3200xf32> loc(#loc3191) + %2201 = "ttir.reshape"(%2199, %2200) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3191) + %2202 = tensor.empty() : tensor<12x3200xf32> loc(#loc3192) + %2203 = "ttir.matmul"(%2201, %arg411, %2202) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3192) + %2204 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3193) + %2205 = "ttir.unsqueeze"(%2203, %2204) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3193) + %2206 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3194) + %2207 = "ttir.add"(%2081, %2205, %2206) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3194) + %2208 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3195) + %2209 = "ttir.multiply"(%2207, %2207, %2208) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3195) + %2210 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3196) + %2211 = "ttir.mean"(%2209, %2210) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3196) + %2212 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3197) + %2213 = "ttir.add"(%2211, %arg154, %2212) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3197) + %2214 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3198) + %2215 = "ttir.sqrt"(%2213, %2214) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3198) + %2216 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3199) + %2217 = "ttir.reciprocal"(%2215, %2216) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3199) + %2218 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3200) + %2219 = "ttir.multiply"(%2207, %2217, %2218) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3200) + %2220 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3201) + %2221 = "ttir.multiply"(%arg412, %2219, %2220) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3201) + %2222 = tensor.empty() : tensor<12x3200xf32> loc(#loc3202) + %2223 = "ttir.squeeze"(%2221, %2222) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3202) + %2224 = tensor.empty() : tensor<12x8640xf32> loc(#loc3203) + %2225 = "ttir.matmul"(%2223, %arg413, %2224) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc3203) + %2226 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3204) + %2227 = "ttir.unsqueeze"(%2225, %2226) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3204) + %2228 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3205) + %2229 = "ttir.sigmoid"(%2227, %2228) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3205) + %2230 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3206) + %2231 = "ttir.multiply"(%2227, %2229, %2230) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3206) + %2232 = tensor.empty() : tensor<12x8640xf32> loc(#loc3207) + %2233 = "ttir.matmul"(%2223, %arg414, %2232) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc3207) + %2234 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3208) + %2235 = "ttir.unsqueeze"(%2233, %2234) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3208) + %2236 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3209) + %2237 = "ttir.multiply"(%2231, %2235, %2236) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3209) + %2238 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3210) + %2239 = "ttir.matmul"(%2237, %arg415, %2238) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3210) + %2240 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3211) + %2241 = "ttir.add"(%2207, %2239, %2240) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3211) + %2242 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3212) + %2243 = "ttir.multiply"(%2241, %2241, %2242) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3212) + %2244 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3213) + %2245 = "ttir.mean"(%2243, %2244) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3213) + %2246 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3214) + %2247 = "ttir.add"(%2245, %arg155, %2246) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3214) + %2248 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3215) + %2249 = "ttir.sqrt"(%2247, %2248) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3215) + %2250 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3216) + %2251 = "ttir.reciprocal"(%2249, %2250) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3216) + %2252 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3217) + %2253 = "ttir.multiply"(%2241, %2251, %2252) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3217) + %2254 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3218) + %2255 = "ttir.multiply"(%arg416, %2253, %2254) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3218) + %2256 = tensor.empty() : tensor<12x3200xf32> loc(#loc3219) + %2257 = "ttir.squeeze"(%2255, %2256) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3219) + %2258 = tensor.empty() : tensor<12x3200xf32> loc(#loc3220) + %2259 = "ttir.matmul"(%2257, %arg417, %2258) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3220) + %2260 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3221) + %2261 = "ttir.reshape"(%2259, %2260) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3221) + %2262 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3222) + %2263 = "ttir.transpose"(%2261, %2262) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3222) + %2264 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3223) + %2265 = "ttir.concat"(%arg156, %arg156, %2264) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3223) + %2266 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3224) + %2267 = "ttir.sin"(%2265, %2266) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3224) + %2268 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc3225) + %2269 = "ttir.unsqueeze"(%2267, %2268) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc3225) + %2270 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3226) + %2271 = "ttir.multiply"(%2263, %2269, %2270) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3226) + %2272 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3227) + %2273 = "ttir.transpose"(%2263, %2272) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3227) + %2274 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3228) + %2275 = "ttir.matmul"(%arg157, %2273, %2274) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3228) + %2276 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3229) + %2277 = "ttir.transpose"(%2275, %2276) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3229) + %2278 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3230) + %2279 = "ttir.multiply"(%2277, %arg158, %2278) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3230) + %2280 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3231) + %2281 = "ttir.transpose"(%2263, %2280) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3231) + %2282 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3232) + %2283 = "ttir.matmul"(%arg159, %2281, %2282) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3232) + %2284 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3233) + %2285 = "ttir.transpose"(%2283, %2284) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3233) + %2286 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3234) + %2287 = "ttir.concat"(%2279, %2285, %2286) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3234) + %2288 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3235) + %2289 = "ttir.cos"(%2265, %2288) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3235) + %2290 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc3236) + %2291 = "ttir.unsqueeze"(%2289, %2290) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc3236) + %2292 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3237) + %2293 = "ttir.multiply"(%2287, %2291, %2292) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3237) + %2294 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3238) + %2295 = "ttir.add"(%2271, %2293, %2294) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3238) + %2296 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3239) + %2297 = "ttir.squeeze"(%2295, %2296) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3239) + %2298 = tensor.empty() : tensor<12x3200xf32> loc(#loc3240) + %2299 = "ttir.matmul"(%2257, %arg418, %2298) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3240) + %2300 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3241) + %2301 = "ttir.reshape"(%2299, %2300) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3241) + %2302 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3242) + %2303 = "ttir.transpose"(%2301, %2302) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3242) + %2304 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3243) + %2305 = "ttir.multiply"(%2303, %2269, %2304) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3243) + %2306 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3244) + %2307 = "ttir.transpose"(%2303, %2306) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3244) + %2308 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3245) + %2309 = "ttir.matmul"(%arg160, %2307, %2308) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3245) + %2310 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3246) + %2311 = "ttir.transpose"(%2309, %2310) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3246) + %2312 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3247) + %2313 = "ttir.multiply"(%2311, %arg161, %2312) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3247) + %2314 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3248) + %2315 = "ttir.transpose"(%2303, %2314) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3248) + %2316 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3249) + %2317 = "ttir.matmul"(%arg162, %2315, %2316) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3249) + %2318 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3250) + %2319 = "ttir.transpose"(%2317, %2318) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3250) + %2320 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3251) + %2321 = "ttir.concat"(%2313, %2319, %2320) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3251) + %2322 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3252) + %2323 = "ttir.multiply"(%2321, %2291, %2322) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3252) + %2324 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3253) + %2325 = "ttir.add"(%2305, %2323, %2324) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3253) + %2326 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3254) + %2327 = "ttir.squeeze"(%2325, %2326) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3254) + %2328 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3255) + %2329 = "ttir.transpose"(%2327, %2328) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3255) + %2330 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3256) + %2331 = "ttir.matmul"(%2297, %2329, %2330) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3256) + %2332 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3257) + %2333 = "ttir.unsqueeze"(%2331, %2332) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3257) + %2334 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3258) + %2335 = "ttir.multiply"(%2333, %arg163, %2334) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3258) + %2336 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3259) + %2337 = "ttir.add"(%2335, %arg164, %2336) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3259) + %2338 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3260) + %2339 = "ttir.softmax"(%2337, %2338) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3260) + %2340 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3261) + %2341 = "ttir.squeeze"(%2339, %2340) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3261) + %2342 = tensor.empty() : tensor<12x3200xf32> loc(#loc3262) + %2343 = "ttir.matmul"(%2257, %arg419, %2342) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3262) + %2344 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3263) + %2345 = "ttir.reshape"(%2343, %2344) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3263) + %2346 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3264) + %2347 = "ttir.transpose"(%2345, %2346) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3264) + %2348 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3265) + %2349 = "ttir.transpose"(%2347, %2348) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3265) + %2350 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3266) + %2351 = "ttir.squeeze"(%2349, %2350) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3266) + %2352 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3267) + %2353 = "ttir.transpose"(%2351, %2352) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3267) + %2354 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3268) + %2355 = "ttir.matmul"(%2341, %2353, %2354) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3268) + %2356 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3269) + %2357 = "ttir.unsqueeze"(%2355, %2356) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3269) + %2358 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3270) + %2359 = "ttir.transpose"(%2357, %2358) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3270) + %2360 = tensor.empty() : tensor<12x3200xf32> loc(#loc3271) + %2361 = "ttir.reshape"(%2359, %2360) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3271) + %2362 = tensor.empty() : tensor<12x3200xf32> loc(#loc3272) + %2363 = "ttir.matmul"(%2361, %arg420, %2362) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3272) + %2364 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3273) + %2365 = "ttir.unsqueeze"(%2363, %2364) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3273) + %2366 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3274) + %2367 = "ttir.add"(%2241, %2365, %2366) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3274) + %2368 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3275) + %2369 = "ttir.multiply"(%2367, %2367, %2368) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3275) + %2370 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3276) + %2371 = "ttir.mean"(%2369, %2370) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3276) + %2372 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3277) + %2373 = "ttir.add"(%2371, %arg165, %2372) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3277) + %2374 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3278) + %2375 = "ttir.sqrt"(%2373, %2374) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3278) + %2376 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3279) + %2377 = "ttir.reciprocal"(%2375, %2376) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3279) + %2378 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3280) + %2379 = "ttir.multiply"(%2367, %2377, %2378) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3280) + %2380 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3281) + %2381 = "ttir.multiply"(%arg421, %2379, %2380) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3281) + %2382 = tensor.empty() : tensor<12x3200xf32> loc(#loc3282) + %2383 = "ttir.squeeze"(%2381, %2382) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3282) + %2384 = tensor.empty() : tensor<12x8640xf32> loc(#loc3283) + %2385 = "ttir.matmul"(%2383, %arg422, %2384) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc3283) + %2386 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3284) + %2387 = "ttir.unsqueeze"(%2385, %2386) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3284) + %2388 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3285) + %2389 = "ttir.sigmoid"(%2387, %2388) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3285) + %2390 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3286) + %2391 = "ttir.multiply"(%2387, %2389, %2390) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3286) + %2392 = tensor.empty() : tensor<12x8640xf32> loc(#loc3287) + %2393 = "ttir.matmul"(%2383, %arg423, %2392) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc3287) + %2394 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3288) + %2395 = "ttir.unsqueeze"(%2393, %2394) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3288) + %2396 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3289) + %2397 = "ttir.multiply"(%2391, %2395, %2396) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3289) + %2398 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3290) + %2399 = "ttir.matmul"(%2397, %arg424, %2398) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3290) + %2400 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3291) + %2401 = "ttir.add"(%2367, %2399, %2400) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3291) + %2402 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3292) + %2403 = "ttir.multiply"(%2401, %2401, %2402) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3292) + %2404 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3293) + %2405 = "ttir.mean"(%2403, %2404) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3293) + %2406 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3294) + %2407 = "ttir.add"(%2405, %arg166, %2406) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3294) + %2408 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3295) + %2409 = "ttir.sqrt"(%2407, %2408) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3295) + %2410 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3296) + %2411 = "ttir.reciprocal"(%2409, %2410) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3296) + %2412 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3297) + %2413 = "ttir.multiply"(%2401, %2411, %2412) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3297) + %2414 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3298) + %2415 = "ttir.multiply"(%arg425, %2413, %2414) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3298) + %2416 = tensor.empty() : tensor<12x3200xf32> loc(#loc3299) + %2417 = "ttir.squeeze"(%2415, %2416) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3299) + %2418 = tensor.empty() : tensor<12x3200xf32> loc(#loc3300) + %2419 = "ttir.matmul"(%2417, %arg426, %2418) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3300) + %2420 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3301) + %2421 = "ttir.reshape"(%2419, %2420) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3301) + %2422 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3302) + %2423 = "ttir.transpose"(%2421, %2422) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3302) + %2424 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3303) + %2425 = "ttir.concat"(%arg167, %arg167, %2424) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3303) + %2426 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3304) + %2427 = "ttir.sin"(%2425, %2426) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3304) + %2428 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc3305) + %2429 = "ttir.unsqueeze"(%2427, %2428) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc3305) + %2430 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3306) + %2431 = "ttir.multiply"(%2423, %2429, %2430) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3306) + %2432 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3307) + %2433 = "ttir.transpose"(%2423, %2432) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3307) + %2434 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3308) + %2435 = "ttir.matmul"(%arg168, %2433, %2434) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3308) + %2436 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3309) + %2437 = "ttir.transpose"(%2435, %2436) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3309) + %2438 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3310) + %2439 = "ttir.multiply"(%2437, %arg169, %2438) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3310) + %2440 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3311) + %2441 = "ttir.transpose"(%2423, %2440) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3311) + %2442 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3312) + %2443 = "ttir.matmul"(%arg170, %2441, %2442) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3312) + %2444 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3313) + %2445 = "ttir.transpose"(%2443, %2444) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3313) + %2446 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3314) + %2447 = "ttir.concat"(%2439, %2445, %2446) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3314) + %2448 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3315) + %2449 = "ttir.cos"(%2425, %2448) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3315) + %2450 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc3316) + %2451 = "ttir.unsqueeze"(%2449, %2450) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc3316) + %2452 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3317) + %2453 = "ttir.multiply"(%2447, %2451, %2452) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3317) + %2454 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3318) + %2455 = "ttir.add"(%2431, %2453, %2454) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3318) + %2456 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3319) + %2457 = "ttir.squeeze"(%2455, %2456) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3319) + %2458 = tensor.empty() : tensor<12x3200xf32> loc(#loc3320) + %2459 = "ttir.matmul"(%2417, %arg427, %2458) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3320) + %2460 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3321) + %2461 = "ttir.reshape"(%2459, %2460) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3321) + %2462 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3322) + %2463 = "ttir.transpose"(%2461, %2462) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3322) + %2464 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3323) + %2465 = "ttir.multiply"(%2463, %2429, %2464) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3323) + %2466 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3324) + %2467 = "ttir.transpose"(%2463, %2466) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3324) + %2468 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3325) + %2469 = "ttir.matmul"(%arg171, %2467, %2468) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3325) + %2470 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3326) + %2471 = "ttir.transpose"(%2469, %2470) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3326) + %2472 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3327) + %2473 = "ttir.multiply"(%2471, %arg172, %2472) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3327) + %2474 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3328) + %2475 = "ttir.transpose"(%2463, %2474) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3328) + %2476 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3329) + %2477 = "ttir.matmul"(%arg173, %2475, %2476) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3329) + %2478 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3330) + %2479 = "ttir.transpose"(%2477, %2478) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3330) + %2480 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3331) + %2481 = "ttir.concat"(%2473, %2479, %2480) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3331) + %2482 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3332) + %2483 = "ttir.multiply"(%2481, %2451, %2482) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3332) + %2484 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3333) + %2485 = "ttir.add"(%2465, %2483, %2484) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3333) + %2486 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3334) + %2487 = "ttir.squeeze"(%2485, %2486) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3334) + %2488 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3335) + %2489 = "ttir.transpose"(%2487, %2488) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3335) + %2490 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3336) + %2491 = "ttir.matmul"(%2457, %2489, %2490) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3336) + %2492 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3337) + %2493 = "ttir.unsqueeze"(%2491, %2492) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3337) + %2494 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3338) + %2495 = "ttir.multiply"(%2493, %arg174, %2494) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3338) + %2496 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3339) + %2497 = "ttir.add"(%2495, %arg175, %2496) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3339) + %2498 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3340) + %2499 = "ttir.softmax"(%2497, %2498) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3340) + %2500 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3341) + %2501 = "ttir.squeeze"(%2499, %2500) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3341) + %2502 = tensor.empty() : tensor<12x3200xf32> loc(#loc3342) + %2503 = "ttir.matmul"(%2417, %arg428, %2502) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3342) + %2504 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3343) + %2505 = "ttir.reshape"(%2503, %2504) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3343) + %2506 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3344) + %2507 = "ttir.transpose"(%2505, %2506) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3344) + %2508 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3345) + %2509 = "ttir.transpose"(%2507, %2508) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3345) + %2510 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3346) + %2511 = "ttir.squeeze"(%2509, %2510) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3346) + %2512 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3347) + %2513 = "ttir.transpose"(%2511, %2512) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3347) + %2514 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3348) + %2515 = "ttir.matmul"(%2501, %2513, %2514) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3348) + %2516 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3349) + %2517 = "ttir.unsqueeze"(%2515, %2516) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3349) + %2518 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3350) + %2519 = "ttir.transpose"(%2517, %2518) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3350) + %2520 = tensor.empty() : tensor<12x3200xf32> loc(#loc3351) + %2521 = "ttir.reshape"(%2519, %2520) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3351) + %2522 = tensor.empty() : tensor<12x3200xf32> loc(#loc3352) + %2523 = "ttir.matmul"(%2521, %arg429, %2522) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3352) + %2524 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3353) + %2525 = "ttir.unsqueeze"(%2523, %2524) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3353) + %2526 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3354) + %2527 = "ttir.add"(%2401, %2525, %2526) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3354) + %2528 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3355) + %2529 = "ttir.multiply"(%2527, %2527, %2528) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3355) + %2530 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3356) + %2531 = "ttir.mean"(%2529, %2530) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3356) + %2532 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3357) + %2533 = "ttir.add"(%2531, %arg176, %2532) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3357) + %2534 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3358) + %2535 = "ttir.sqrt"(%2533, %2534) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3358) + %2536 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3359) + %2537 = "ttir.reciprocal"(%2535, %2536) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3359) + %2538 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3360) + %2539 = "ttir.multiply"(%2527, %2537, %2538) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3360) + %2540 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3361) + %2541 = "ttir.multiply"(%arg430, %2539, %2540) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3361) + %2542 = tensor.empty() : tensor<12x3200xf32> loc(#loc3362) + %2543 = "ttir.squeeze"(%2541, %2542) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3362) + %2544 = tensor.empty() : tensor<12x8640xf32> loc(#loc3363) + %2545 = "ttir.matmul"(%2543, %arg431, %2544) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc3363) + %2546 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3364) + %2547 = "ttir.unsqueeze"(%2545, %2546) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3364) + %2548 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3365) + %2549 = "ttir.sigmoid"(%2547, %2548) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3365) + %2550 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3366) + %2551 = "ttir.multiply"(%2547, %2549, %2550) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3366) + %2552 = tensor.empty() : tensor<12x8640xf32> loc(#loc3367) + %2553 = "ttir.matmul"(%2543, %arg432, %2552) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc3367) + %2554 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3368) + %2555 = "ttir.unsqueeze"(%2553, %2554) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3368) + %2556 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3369) + %2557 = "ttir.multiply"(%2551, %2555, %2556) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3369) + %2558 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3370) + %2559 = "ttir.matmul"(%2557, %arg433, %2558) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3370) + %2560 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3371) + %2561 = "ttir.add"(%2527, %2559, %2560) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3371) + %2562 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3372) + %2563 = "ttir.multiply"(%2561, %2561, %2562) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3372) + %2564 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3373) + %2565 = "ttir.mean"(%2563, %2564) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3373) + %2566 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3374) + %2567 = "ttir.add"(%2565, %arg177, %2566) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3374) + %2568 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3375) + %2569 = "ttir.sqrt"(%2567, %2568) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3375) + %2570 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3376) + %2571 = "ttir.reciprocal"(%2569, %2570) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3376) + %2572 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3377) + %2573 = "ttir.multiply"(%2561, %2571, %2572) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3377) + %2574 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3378) + %2575 = "ttir.multiply"(%arg434, %2573, %2574) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3378) + %2576 = tensor.empty() : tensor<12x3200xf32> loc(#loc3379) + %2577 = "ttir.squeeze"(%2575, %2576) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3379) + %2578 = tensor.empty() : tensor<12x3200xf32> loc(#loc3380) + %2579 = "ttir.matmul"(%2577, %arg435, %2578) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3380) + %2580 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3381) + %2581 = "ttir.reshape"(%2579, %2580) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3381) + %2582 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3382) + %2583 = "ttir.transpose"(%2581, %2582) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3382) + %2584 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3383) + %2585 = "ttir.concat"(%arg178, %arg178, %2584) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3383) + %2586 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3384) + %2587 = "ttir.sin"(%2585, %2586) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3384) + %2588 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc3385) + %2589 = "ttir.unsqueeze"(%2587, %2588) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc3385) + %2590 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3386) + %2591 = "ttir.multiply"(%2583, %2589, %2590) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3386) + %2592 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3387) + %2593 = "ttir.transpose"(%2583, %2592) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3387) + %2594 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3388) + %2595 = "ttir.matmul"(%arg179, %2593, %2594) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3388) + %2596 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3389) + %2597 = "ttir.transpose"(%2595, %2596) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3389) + %2598 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3390) + %2599 = "ttir.multiply"(%2597, %arg180, %2598) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3390) + %2600 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3391) + %2601 = "ttir.transpose"(%2583, %2600) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3391) + %2602 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3392) + %2603 = "ttir.matmul"(%arg181, %2601, %2602) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3392) + %2604 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3393) + %2605 = "ttir.transpose"(%2603, %2604) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3393) + %2606 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3394) + %2607 = "ttir.concat"(%2599, %2605, %2606) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3394) + %2608 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3395) + %2609 = "ttir.cos"(%2585, %2608) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3395) + %2610 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc3396) + %2611 = "ttir.unsqueeze"(%2609, %2610) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc3396) + %2612 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3397) + %2613 = "ttir.multiply"(%2607, %2611, %2612) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3397) + %2614 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3398) + %2615 = "ttir.add"(%2591, %2613, %2614) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3398) + %2616 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3399) + %2617 = "ttir.squeeze"(%2615, %2616) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3399) + %2618 = tensor.empty() : tensor<12x3200xf32> loc(#loc3400) + %2619 = "ttir.matmul"(%2577, %arg436, %2618) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3400) + %2620 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3401) + %2621 = "ttir.reshape"(%2619, %2620) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3401) + %2622 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3402) + %2623 = "ttir.transpose"(%2621, %2622) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3402) + %2624 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3403) + %2625 = "ttir.multiply"(%2623, %2589, %2624) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3403) + %2626 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3404) + %2627 = "ttir.transpose"(%2623, %2626) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3404) + %2628 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3405) + %2629 = "ttir.matmul"(%arg182, %2627, %2628) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3405) + %2630 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3406) + %2631 = "ttir.transpose"(%2629, %2630) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3406) + %2632 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3407) + %2633 = "ttir.multiply"(%2631, %arg183, %2632) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3407) + %2634 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3408) + %2635 = "ttir.transpose"(%2623, %2634) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3408) + %2636 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3409) + %2637 = "ttir.matmul"(%arg184, %2635, %2636) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3409) + %2638 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3410) + %2639 = "ttir.transpose"(%2637, %2638) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3410) + %2640 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3411) + %2641 = "ttir.concat"(%2633, %2639, %2640) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3411) + %2642 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3412) + %2643 = "ttir.multiply"(%2641, %2611, %2642) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3412) + %2644 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3413) + %2645 = "ttir.add"(%2625, %2643, %2644) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3413) + %2646 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3414) + %2647 = "ttir.squeeze"(%2645, %2646) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3414) + %2648 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3415) + %2649 = "ttir.transpose"(%2647, %2648) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3415) + %2650 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3416) + %2651 = "ttir.matmul"(%2617, %2649, %2650) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3416) + %2652 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3417) + %2653 = "ttir.unsqueeze"(%2651, %2652) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3417) + %2654 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3418) + %2655 = "ttir.multiply"(%2653, %arg185, %2654) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3418) + %2656 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3419) + %2657 = "ttir.add"(%2655, %arg186, %2656) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3419) + %2658 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3420) + %2659 = "ttir.softmax"(%2657, %2658) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3420) + %2660 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3421) + %2661 = "ttir.squeeze"(%2659, %2660) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3421) + %2662 = tensor.empty() : tensor<12x3200xf32> loc(#loc3422) + %2663 = "ttir.matmul"(%2577, %arg437, %2662) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3422) + %2664 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3423) + %2665 = "ttir.reshape"(%2663, %2664) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3423) + %2666 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3424) + %2667 = "ttir.transpose"(%2665, %2666) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3424) + %2668 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3425) + %2669 = "ttir.transpose"(%2667, %2668) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3425) + %2670 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3426) + %2671 = "ttir.squeeze"(%2669, %2670) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3426) + %2672 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3427) + %2673 = "ttir.transpose"(%2671, %2672) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3427) + %2674 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3428) + %2675 = "ttir.matmul"(%2661, %2673, %2674) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3428) + %2676 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3429) + %2677 = "ttir.unsqueeze"(%2675, %2676) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3429) + %2678 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3430) + %2679 = "ttir.transpose"(%2677, %2678) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3430) + %2680 = tensor.empty() : tensor<12x3200xf32> loc(#loc3431) + %2681 = "ttir.reshape"(%2679, %2680) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3431) + %2682 = tensor.empty() : tensor<12x3200xf32> loc(#loc3432) + %2683 = "ttir.matmul"(%2681, %arg438, %2682) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3432) + %2684 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3433) + %2685 = "ttir.unsqueeze"(%2683, %2684) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3433) + %2686 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3434) + %2687 = "ttir.add"(%2561, %2685, %2686) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3434) + %2688 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3435) + %2689 = "ttir.multiply"(%2687, %2687, %2688) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3435) + %2690 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3436) + %2691 = "ttir.mean"(%2689, %2690) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3436) + %2692 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3437) + %2693 = "ttir.add"(%2691, %arg187, %2692) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3437) + %2694 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3438) + %2695 = "ttir.sqrt"(%2693, %2694) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3438) + %2696 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3439) + %2697 = "ttir.reciprocal"(%2695, %2696) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3439) + %2698 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3440) + %2699 = "ttir.multiply"(%2687, %2697, %2698) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3440) + %2700 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3441) + %2701 = "ttir.multiply"(%arg439, %2699, %2700) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3441) + %2702 = tensor.empty() : tensor<12x3200xf32> loc(#loc3442) + %2703 = "ttir.squeeze"(%2701, %2702) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3442) + %2704 = tensor.empty() : tensor<12x8640xf32> loc(#loc3443) + %2705 = "ttir.matmul"(%2703, %arg440, %2704) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc3443) + %2706 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3444) + %2707 = "ttir.unsqueeze"(%2705, %2706) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3444) + %2708 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3445) + %2709 = "ttir.sigmoid"(%2707, %2708) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3445) + %2710 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3446) + %2711 = "ttir.multiply"(%2707, %2709, %2710) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3446) + %2712 = tensor.empty() : tensor<12x8640xf32> loc(#loc3447) + %2713 = "ttir.matmul"(%2703, %arg441, %2712) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc3447) + %2714 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3448) + %2715 = "ttir.unsqueeze"(%2713, %2714) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3448) + %2716 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3449) + %2717 = "ttir.multiply"(%2711, %2715, %2716) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3449) + %2718 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3450) + %2719 = "ttir.matmul"(%2717, %arg442, %2718) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3450) + %2720 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3451) + %2721 = "ttir.add"(%2687, %2719, %2720) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3451) + %2722 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3452) + %2723 = "ttir.multiply"(%2721, %2721, %2722) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3452) + %2724 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3453) + %2725 = "ttir.mean"(%2723, %2724) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3453) + %2726 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3454) + %2727 = "ttir.add"(%2725, %arg188, %2726) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3454) + %2728 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3455) + %2729 = "ttir.sqrt"(%2727, %2728) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3455) + %2730 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3456) + %2731 = "ttir.reciprocal"(%2729, %2730) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3456) + %2732 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3457) + %2733 = "ttir.multiply"(%2721, %2731, %2732) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3457) + %2734 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3458) + %2735 = "ttir.multiply"(%arg443, %2733, %2734) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3458) + %2736 = tensor.empty() : tensor<12x3200xf32> loc(#loc3459) + %2737 = "ttir.squeeze"(%2735, %2736) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3459) + %2738 = tensor.empty() : tensor<12x3200xf32> loc(#loc3460) + %2739 = "ttir.matmul"(%2737, %arg444, %2738) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3460) + %2740 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3461) + %2741 = "ttir.reshape"(%2739, %2740) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3461) + %2742 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3462) + %2743 = "ttir.transpose"(%2741, %2742) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3462) + %2744 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3463) + %2745 = "ttir.concat"(%arg189, %arg189, %2744) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3463) + %2746 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3464) + %2747 = "ttir.sin"(%2745, %2746) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3464) + %2748 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc3465) + %2749 = "ttir.unsqueeze"(%2747, %2748) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc3465) + %2750 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3466) + %2751 = "ttir.multiply"(%2743, %2749, %2750) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3466) + %2752 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3467) + %2753 = "ttir.transpose"(%2743, %2752) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3467) + %2754 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3468) + %2755 = "ttir.matmul"(%arg190, %2753, %2754) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3468) + %2756 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3469) + %2757 = "ttir.transpose"(%2755, %2756) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3469) + %2758 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3470) + %2759 = "ttir.multiply"(%2757, %arg191, %2758) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3470) + %2760 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3471) + %2761 = "ttir.transpose"(%2743, %2760) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3471) + %2762 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3472) + %2763 = "ttir.matmul"(%arg192, %2761, %2762) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3472) + %2764 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3473) + %2765 = "ttir.transpose"(%2763, %2764) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3473) + %2766 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3474) + %2767 = "ttir.concat"(%2759, %2765, %2766) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3474) + %2768 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3475) + %2769 = "ttir.cos"(%2745, %2768) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3475) + %2770 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc3476) + %2771 = "ttir.unsqueeze"(%2769, %2770) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc3476) + %2772 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3477) + %2773 = "ttir.multiply"(%2767, %2771, %2772) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3477) + %2774 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3478) + %2775 = "ttir.add"(%2751, %2773, %2774) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3478) + %2776 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3479) + %2777 = "ttir.squeeze"(%2775, %2776) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3479) + %2778 = tensor.empty() : tensor<12x3200xf32> loc(#loc3480) + %2779 = "ttir.matmul"(%2737, %arg445, %2778) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3480) + %2780 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3481) + %2781 = "ttir.reshape"(%2779, %2780) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3481) + %2782 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3482) + %2783 = "ttir.transpose"(%2781, %2782) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3482) + %2784 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3483) + %2785 = "ttir.multiply"(%2783, %2749, %2784) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3483) + %2786 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3484) + %2787 = "ttir.transpose"(%2783, %2786) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3484) + %2788 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3485) + %2789 = "ttir.matmul"(%arg193, %2787, %2788) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3485) + %2790 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3486) + %2791 = "ttir.transpose"(%2789, %2790) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3486) + %2792 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3487) + %2793 = "ttir.multiply"(%2791, %arg194, %2792) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3487) + %2794 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3488) + %2795 = "ttir.transpose"(%2783, %2794) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3488) + %2796 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3489) + %2797 = "ttir.matmul"(%arg195, %2795, %2796) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3489) + %2798 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3490) + %2799 = "ttir.transpose"(%2797, %2798) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3490) + %2800 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3491) + %2801 = "ttir.concat"(%2793, %2799, %2800) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3491) + %2802 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3492) + %2803 = "ttir.multiply"(%2801, %2771, %2802) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3492) + %2804 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3493) + %2805 = "ttir.add"(%2785, %2803, %2804) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3493) + %2806 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3494) + %2807 = "ttir.squeeze"(%2805, %2806) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3494) + %2808 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3495) + %2809 = "ttir.transpose"(%2807, %2808) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3495) + %2810 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3496) + %2811 = "ttir.matmul"(%2777, %2809, %2810) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3496) + %2812 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3497) + %2813 = "ttir.unsqueeze"(%2811, %2812) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3497) + %2814 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3498) + %2815 = "ttir.multiply"(%2813, %arg196, %2814) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3498) + %2816 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3499) + %2817 = "ttir.add"(%2815, %arg197, %2816) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3499) + %2818 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3500) + %2819 = "ttir.softmax"(%2817, %2818) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3500) + %2820 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3501) + %2821 = "ttir.squeeze"(%2819, %2820) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3501) + %2822 = tensor.empty() : tensor<12x3200xf32> loc(#loc3502) + %2823 = "ttir.matmul"(%2737, %arg446, %2822) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3502) + %2824 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3503) + %2825 = "ttir.reshape"(%2823, %2824) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3503) + %2826 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3504) + %2827 = "ttir.transpose"(%2825, %2826) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3504) + %2828 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3505) + %2829 = "ttir.transpose"(%2827, %2828) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3505) + %2830 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3506) + %2831 = "ttir.squeeze"(%2829, %2830) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3506) + %2832 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3507) + %2833 = "ttir.transpose"(%2831, %2832) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3507) + %2834 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3508) + %2835 = "ttir.matmul"(%2821, %2833, %2834) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3508) + %2836 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3509) + %2837 = "ttir.unsqueeze"(%2835, %2836) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3509) + %2838 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3510) + %2839 = "ttir.transpose"(%2837, %2838) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3510) + %2840 = tensor.empty() : tensor<12x3200xf32> loc(#loc3511) + %2841 = "ttir.reshape"(%2839, %2840) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3511) + %2842 = tensor.empty() : tensor<12x3200xf32> loc(#loc3512) + %2843 = "ttir.matmul"(%2841, %arg447, %2842) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3512) + %2844 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3513) + %2845 = "ttir.unsqueeze"(%2843, %2844) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3513) + %2846 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3514) + %2847 = "ttir.add"(%2721, %2845, %2846) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3514) + %2848 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3515) + %2849 = "ttir.multiply"(%2847, %2847, %2848) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3515) + %2850 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3516) + %2851 = "ttir.mean"(%2849, %2850) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3516) + %2852 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3517) + %2853 = "ttir.add"(%2851, %arg198, %2852) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3517) + %2854 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3518) + %2855 = "ttir.sqrt"(%2853, %2854) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3518) + %2856 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3519) + %2857 = "ttir.reciprocal"(%2855, %2856) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3519) + %2858 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3520) + %2859 = "ttir.multiply"(%2847, %2857, %2858) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3520) + %2860 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3521) + %2861 = "ttir.multiply"(%arg448, %2859, %2860) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3521) + %2862 = tensor.empty() : tensor<12x3200xf32> loc(#loc3522) + %2863 = "ttir.squeeze"(%2861, %2862) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3522) + %2864 = tensor.empty() : tensor<12x8640xf32> loc(#loc3523) + %2865 = "ttir.matmul"(%2863, %arg449, %2864) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc3523) + %2866 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3524) + %2867 = "ttir.unsqueeze"(%2865, %2866) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3524) + %2868 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3525) + %2869 = "ttir.sigmoid"(%2867, %2868) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3525) + %2870 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3526) + %2871 = "ttir.multiply"(%2867, %2869, %2870) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3526) + %2872 = tensor.empty() : tensor<12x8640xf32> loc(#loc3527) + %2873 = "ttir.matmul"(%2863, %arg450, %2872) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc3527) + %2874 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3528) + %2875 = "ttir.unsqueeze"(%2873, %2874) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3528) + %2876 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3529) + %2877 = "ttir.multiply"(%2871, %2875, %2876) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3529) + %2878 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3530) + %2879 = "ttir.matmul"(%2877, %arg451, %2878) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3530) + %2880 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3531) + %2881 = "ttir.add"(%2847, %2879, %2880) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3531) + %2882 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3532) + %2883 = "ttir.multiply"(%2881, %2881, %2882) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3532) + %2884 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3533) + %2885 = "ttir.mean"(%2883, %2884) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3533) + %2886 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3534) + %2887 = "ttir.add"(%2885, %arg199, %2886) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3534) + %2888 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3535) + %2889 = "ttir.sqrt"(%2887, %2888) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3535) + %2890 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3536) + %2891 = "ttir.reciprocal"(%2889, %2890) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3536) + %2892 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3537) + %2893 = "ttir.multiply"(%2881, %2891, %2892) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3537) + %2894 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3538) + %2895 = "ttir.multiply"(%arg452, %2893, %2894) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3538) + %2896 = tensor.empty() : tensor<12x3200xf32> loc(#loc3539) + %2897 = "ttir.squeeze"(%2895, %2896) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3539) + %2898 = tensor.empty() : tensor<12x3200xf32> loc(#loc3540) + %2899 = "ttir.matmul"(%2897, %arg453, %2898) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3540) + %2900 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3541) + %2901 = "ttir.reshape"(%2899, %2900) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3541) + %2902 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3542) + %2903 = "ttir.transpose"(%2901, %2902) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3542) + %2904 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3543) + %2905 = "ttir.concat"(%arg200, %arg200, %2904) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3543) + %2906 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3544) + %2907 = "ttir.sin"(%2905, %2906) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3544) + %2908 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc3545) + %2909 = "ttir.unsqueeze"(%2907, %2908) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc3545) + %2910 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3546) + %2911 = "ttir.multiply"(%2903, %2909, %2910) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3546) + %2912 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3547) + %2913 = "ttir.transpose"(%2903, %2912) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3547) + %2914 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3548) + %2915 = "ttir.matmul"(%arg201, %2913, %2914) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3548) + %2916 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3549) + %2917 = "ttir.transpose"(%2915, %2916) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3549) + %2918 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3550) + %2919 = "ttir.multiply"(%2917, %arg202, %2918) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3550) + %2920 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3551) + %2921 = "ttir.transpose"(%2903, %2920) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3551) + %2922 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3552) + %2923 = "ttir.matmul"(%arg203, %2921, %2922) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3552) + %2924 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3553) + %2925 = "ttir.transpose"(%2923, %2924) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3553) + %2926 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3554) + %2927 = "ttir.concat"(%2919, %2925, %2926) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3554) + %2928 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3555) + %2929 = "ttir.cos"(%2905, %2928) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3555) + %2930 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc3556) + %2931 = "ttir.unsqueeze"(%2929, %2930) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc3556) + %2932 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3557) + %2933 = "ttir.multiply"(%2927, %2931, %2932) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3557) + %2934 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3558) + %2935 = "ttir.add"(%2911, %2933, %2934) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3558) + %2936 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3559) + %2937 = "ttir.squeeze"(%2935, %2936) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3559) + %2938 = tensor.empty() : tensor<12x3200xf32> loc(#loc3560) + %2939 = "ttir.matmul"(%2897, %arg454, %2938) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3560) + %2940 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3561) + %2941 = "ttir.reshape"(%2939, %2940) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3561) + %2942 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3562) + %2943 = "ttir.transpose"(%2941, %2942) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3562) + %2944 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3563) + %2945 = "ttir.multiply"(%2943, %2909, %2944) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3563) + %2946 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3564) + %2947 = "ttir.transpose"(%2943, %2946) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3564) + %2948 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3565) + %2949 = "ttir.matmul"(%arg204, %2947, %2948) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3565) + %2950 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3566) + %2951 = "ttir.transpose"(%2949, %2950) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3566) + %2952 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3567) + %2953 = "ttir.multiply"(%2951, %arg205, %2952) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3567) + %2954 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3568) + %2955 = "ttir.transpose"(%2943, %2954) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3568) + %2956 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3569) + %2957 = "ttir.matmul"(%arg206, %2955, %2956) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3569) + %2958 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3570) + %2959 = "ttir.transpose"(%2957, %2958) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3570) + %2960 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3571) + %2961 = "ttir.concat"(%2953, %2959, %2960) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3571) + %2962 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3572) + %2963 = "ttir.multiply"(%2961, %2931, %2962) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3572) + %2964 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3573) + %2965 = "ttir.add"(%2945, %2963, %2964) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3573) + %2966 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3574) + %2967 = "ttir.squeeze"(%2965, %2966) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3574) + %2968 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3575) + %2969 = "ttir.transpose"(%2967, %2968) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3575) + %2970 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3576) + %2971 = "ttir.matmul"(%2937, %2969, %2970) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3576) + %2972 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3577) + %2973 = "ttir.unsqueeze"(%2971, %2972) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3577) + %2974 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3578) + %2975 = "ttir.multiply"(%2973, %arg207, %2974) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3578) + %2976 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3579) + %2977 = "ttir.add"(%2975, %arg208, %2976) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3579) + %2978 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3580) + %2979 = "ttir.softmax"(%2977, %2978) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3580) + %2980 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3581) + %2981 = "ttir.squeeze"(%2979, %2980) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3581) + %2982 = tensor.empty() : tensor<12x3200xf32> loc(#loc3582) + %2983 = "ttir.matmul"(%2897, %arg455, %2982) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3582) + %2984 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3583) + %2985 = "ttir.reshape"(%2983, %2984) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3583) + %2986 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3584) + %2987 = "ttir.transpose"(%2985, %2986) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3584) + %2988 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3585) + %2989 = "ttir.transpose"(%2987, %2988) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3585) + %2990 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3586) + %2991 = "ttir.squeeze"(%2989, %2990) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3586) + %2992 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3587) + %2993 = "ttir.transpose"(%2991, %2992) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3587) + %2994 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3588) + %2995 = "ttir.matmul"(%2981, %2993, %2994) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3588) + %2996 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3589) + %2997 = "ttir.unsqueeze"(%2995, %2996) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3589) + %2998 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3590) + %2999 = "ttir.transpose"(%2997, %2998) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3590) + %3000 = tensor.empty() : tensor<12x3200xf32> loc(#loc3591) + %3001 = "ttir.reshape"(%2999, %3000) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3591) + %3002 = tensor.empty() : tensor<12x3200xf32> loc(#loc3592) + %3003 = "ttir.matmul"(%3001, %arg456, %3002) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3592) + %3004 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3593) + %3005 = "ttir.unsqueeze"(%3003, %3004) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3593) + %3006 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3594) + %3007 = "ttir.add"(%2881, %3005, %3006) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3594) + %3008 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3595) + %3009 = "ttir.multiply"(%3007, %3007, %3008) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3595) + %3010 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3596) + %3011 = "ttir.mean"(%3009, %3010) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3596) + %3012 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3597) + %3013 = "ttir.add"(%3011, %arg209, %3012) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3597) + %3014 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3598) + %3015 = "ttir.sqrt"(%3013, %3014) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3598) + %3016 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3599) + %3017 = "ttir.reciprocal"(%3015, %3016) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3599) + %3018 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3600) + %3019 = "ttir.multiply"(%3007, %3017, %3018) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3600) + %3020 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3601) + %3021 = "ttir.multiply"(%arg457, %3019, %3020) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3601) + %3022 = tensor.empty() : tensor<12x3200xf32> loc(#loc3602) + %3023 = "ttir.squeeze"(%3021, %3022) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3602) + %3024 = tensor.empty() : tensor<12x8640xf32> loc(#loc3603) + %3025 = "ttir.matmul"(%3023, %arg458, %3024) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc3603) + %3026 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3604) + %3027 = "ttir.unsqueeze"(%3025, %3026) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3604) + %3028 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3605) + %3029 = "ttir.sigmoid"(%3027, %3028) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3605) + %3030 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3606) + %3031 = "ttir.multiply"(%3027, %3029, %3030) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3606) + %3032 = tensor.empty() : tensor<12x8640xf32> loc(#loc3607) + %3033 = "ttir.matmul"(%3023, %arg459, %3032) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc3607) + %3034 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3608) + %3035 = "ttir.unsqueeze"(%3033, %3034) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3608) + %3036 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3609) + %3037 = "ttir.multiply"(%3031, %3035, %3036) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3609) + %3038 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3610) + %3039 = "ttir.matmul"(%3037, %arg460, %3038) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3610) + %3040 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3611) + %3041 = "ttir.add"(%3007, %3039, %3040) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3611) + %3042 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3612) + %3043 = "ttir.multiply"(%3041, %3041, %3042) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3612) + %3044 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3613) + %3045 = "ttir.mean"(%3043, %3044) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3613) + %3046 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3614) + %3047 = "ttir.add"(%3045, %arg210, %3046) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3614) + %3048 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3615) + %3049 = "ttir.sqrt"(%3047, %3048) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3615) + %3050 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3616) + %3051 = "ttir.reciprocal"(%3049, %3050) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3616) + %3052 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3617) + %3053 = "ttir.multiply"(%3041, %3051, %3052) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3617) + %3054 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3618) + %3055 = "ttir.multiply"(%arg461, %3053, %3054) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3618) + %3056 = tensor.empty() : tensor<12x3200xf32> loc(#loc3619) + %3057 = "ttir.squeeze"(%3055, %3056) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3619) + %3058 = tensor.empty() : tensor<12x3200xf32> loc(#loc3620) + %3059 = "ttir.matmul"(%3057, %arg462, %3058) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3620) + %3060 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3621) + %3061 = "ttir.reshape"(%3059, %3060) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3621) + %3062 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3622) + %3063 = "ttir.transpose"(%3061, %3062) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3622) + %3064 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3623) + %3065 = "ttir.concat"(%arg211, %arg211, %3064) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3623) + %3066 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3624) + %3067 = "ttir.sin"(%3065, %3066) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3624) + %3068 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc3625) + %3069 = "ttir.unsqueeze"(%3067, %3068) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc3625) + %3070 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3626) + %3071 = "ttir.multiply"(%3063, %3069, %3070) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3626) + %3072 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3627) + %3073 = "ttir.transpose"(%3063, %3072) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3627) + %3074 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3628) + %3075 = "ttir.matmul"(%arg212, %3073, %3074) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3628) + %3076 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3629) + %3077 = "ttir.transpose"(%3075, %3076) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3629) + %3078 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3630) + %3079 = "ttir.multiply"(%3077, %arg213, %3078) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3630) + %3080 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3631) + %3081 = "ttir.transpose"(%3063, %3080) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3631) + %3082 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3632) + %3083 = "ttir.matmul"(%arg214, %3081, %3082) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3632) + %3084 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3633) + %3085 = "ttir.transpose"(%3083, %3084) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3633) + %3086 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3634) + %3087 = "ttir.concat"(%3079, %3085, %3086) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3634) + %3088 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3635) + %3089 = "ttir.cos"(%3065, %3088) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3635) + %3090 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc3636) + %3091 = "ttir.unsqueeze"(%3089, %3090) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc3636) + %3092 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3637) + %3093 = "ttir.multiply"(%3087, %3091, %3092) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3637) + %3094 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3638) + %3095 = "ttir.add"(%3071, %3093, %3094) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3638) + %3096 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3639) + %3097 = "ttir.squeeze"(%3095, %3096) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3639) + %3098 = tensor.empty() : tensor<12x3200xf32> loc(#loc3640) + %3099 = "ttir.matmul"(%3057, %arg463, %3098) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3640) + %3100 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3641) + %3101 = "ttir.reshape"(%3099, %3100) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3641) + %3102 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3642) + %3103 = "ttir.transpose"(%3101, %3102) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3642) + %3104 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3643) + %3105 = "ttir.multiply"(%3103, %3069, %3104) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3643) + %3106 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3644) + %3107 = "ttir.transpose"(%3103, %3106) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3644) + %3108 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3645) + %3109 = "ttir.matmul"(%arg215, %3107, %3108) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3645) + %3110 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3646) + %3111 = "ttir.transpose"(%3109, %3110) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3646) + %3112 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3647) + %3113 = "ttir.multiply"(%3111, %arg216, %3112) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3647) + %3114 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3648) + %3115 = "ttir.transpose"(%3103, %3114) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3648) + %3116 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3649) + %3117 = "ttir.matmul"(%arg217, %3115, %3116) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3649) + %3118 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3650) + %3119 = "ttir.transpose"(%3117, %3118) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3650) + %3120 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3651) + %3121 = "ttir.concat"(%3113, %3119, %3120) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3651) + %3122 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3652) + %3123 = "ttir.multiply"(%3121, %3091, %3122) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3652) + %3124 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3653) + %3125 = "ttir.add"(%3105, %3123, %3124) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3653) + %3126 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3654) + %3127 = "ttir.squeeze"(%3125, %3126) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3654) + %3128 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3655) + %3129 = "ttir.transpose"(%3127, %3128) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3655) + %3130 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3656) + %3131 = "ttir.matmul"(%3097, %3129, %3130) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3656) + %3132 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3657) + %3133 = "ttir.unsqueeze"(%3131, %3132) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3657) + %3134 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3658) + %3135 = "ttir.multiply"(%3133, %arg218, %3134) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3658) + %3136 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3659) + %3137 = "ttir.add"(%3135, %arg219, %3136) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3659) + %3138 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3660) + %3139 = "ttir.softmax"(%3137, %3138) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3660) + %3140 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3661) + %3141 = "ttir.squeeze"(%3139, %3140) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3661) + %3142 = tensor.empty() : tensor<12x3200xf32> loc(#loc3662) + %3143 = "ttir.matmul"(%3057, %arg464, %3142) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3662) + %3144 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3663) + %3145 = "ttir.reshape"(%3143, %3144) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3663) + %3146 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3664) + %3147 = "ttir.transpose"(%3145, %3146) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3664) + %3148 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3665) + %3149 = "ttir.transpose"(%3147, %3148) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3665) + %3150 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3666) + %3151 = "ttir.squeeze"(%3149, %3150) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3666) + %3152 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3667) + %3153 = "ttir.transpose"(%3151, %3152) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3667) + %3154 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3668) + %3155 = "ttir.matmul"(%3141, %3153, %3154) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3668) + %3156 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3669) + %3157 = "ttir.unsqueeze"(%3155, %3156) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3669) + %3158 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3670) + %3159 = "ttir.transpose"(%3157, %3158) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3670) + %3160 = tensor.empty() : tensor<12x3200xf32> loc(#loc3671) + %3161 = "ttir.reshape"(%3159, %3160) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3671) + %3162 = tensor.empty() : tensor<12x3200xf32> loc(#loc3672) + %3163 = "ttir.matmul"(%3161, %arg465, %3162) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3672) + %3164 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3673) + %3165 = "ttir.unsqueeze"(%3163, %3164) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3673) + %3166 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3674) + %3167 = "ttir.add"(%3041, %3165, %3166) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3674) + %3168 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3675) + %3169 = "ttir.multiply"(%3167, %3167, %3168) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3675) + %3170 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3676) + %3171 = "ttir.mean"(%3169, %3170) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3676) + %3172 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3677) + %3173 = "ttir.add"(%3171, %arg220, %3172) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3677) + %3174 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3678) + %3175 = "ttir.sqrt"(%3173, %3174) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3678) + %3176 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3679) + %3177 = "ttir.reciprocal"(%3175, %3176) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3679) + %3178 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3680) + %3179 = "ttir.multiply"(%3167, %3177, %3178) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3680) + %3180 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3681) + %3181 = "ttir.multiply"(%arg466, %3179, %3180) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3681) + %3182 = tensor.empty() : tensor<12x3200xf32> loc(#loc3682) + %3183 = "ttir.squeeze"(%3181, %3182) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3682) + %3184 = tensor.empty() : tensor<12x8640xf32> loc(#loc3683) + %3185 = "ttir.matmul"(%3183, %arg467, %3184) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc3683) + %3186 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3684) + %3187 = "ttir.unsqueeze"(%3185, %3186) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3684) + %3188 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3685) + %3189 = "ttir.sigmoid"(%3187, %3188) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3685) + %3190 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3686) + %3191 = "ttir.multiply"(%3187, %3189, %3190) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3686) + %3192 = tensor.empty() : tensor<12x8640xf32> loc(#loc3687) + %3193 = "ttir.matmul"(%3183, %arg468, %3192) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc3687) + %3194 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3688) + %3195 = "ttir.unsqueeze"(%3193, %3194) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3688) + %3196 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3689) + %3197 = "ttir.multiply"(%3191, %3195, %3196) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3689) + %3198 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3690) + %3199 = "ttir.matmul"(%3197, %arg469, %3198) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3690) + %3200 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3691) + %3201 = "ttir.add"(%3167, %3199, %3200) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3691) + %3202 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3692) + %3203 = "ttir.multiply"(%3201, %3201, %3202) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3692) + %3204 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3693) + %3205 = "ttir.mean"(%3203, %3204) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3693) + %3206 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3694) + %3207 = "ttir.add"(%3205, %arg221, %3206) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3694) + %3208 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3695) + %3209 = "ttir.sqrt"(%3207, %3208) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3695) + %3210 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3696) + %3211 = "ttir.reciprocal"(%3209, %3210) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3696) + %3212 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3697) + %3213 = "ttir.multiply"(%3201, %3211, %3212) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3697) + %3214 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3698) + %3215 = "ttir.multiply"(%arg470, %3213, %3214) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3698) + %3216 = tensor.empty() : tensor<12x3200xf32> loc(#loc3699) + %3217 = "ttir.squeeze"(%3215, %3216) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3699) + %3218 = tensor.empty() : tensor<12x3200xf32> loc(#loc3700) + %3219 = "ttir.matmul"(%3217, %arg471, %3218) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3700) + %3220 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3701) + %3221 = "ttir.reshape"(%3219, %3220) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3701) + %3222 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3702) + %3223 = "ttir.transpose"(%3221, %3222) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3702) + %3224 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3703) + %3225 = "ttir.concat"(%arg222, %arg222, %3224) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3703) + %3226 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3704) + %3227 = "ttir.sin"(%3225, %3226) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3704) + %3228 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc3705) + %3229 = "ttir.unsqueeze"(%3227, %3228) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc3705) + %3230 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3706) + %3231 = "ttir.multiply"(%3223, %3229, %3230) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3706) + %3232 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3707) + %3233 = "ttir.transpose"(%3223, %3232) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3707) + %3234 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3708) + %3235 = "ttir.matmul"(%arg223, %3233, %3234) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3708) + %3236 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3709) + %3237 = "ttir.transpose"(%3235, %3236) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3709) + %3238 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3710) + %3239 = "ttir.multiply"(%3237, %arg224, %3238) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3710) + %3240 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3711) + %3241 = "ttir.transpose"(%3223, %3240) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3711) + %3242 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3712) + %3243 = "ttir.matmul"(%arg225, %3241, %3242) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3712) + %3244 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3713) + %3245 = "ttir.transpose"(%3243, %3244) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3713) + %3246 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3714) + %3247 = "ttir.concat"(%3239, %3245, %3246) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3714) + %3248 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3715) + %3249 = "ttir.cos"(%3225, %3248) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3715) + %3250 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc3716) + %3251 = "ttir.unsqueeze"(%3249, %3250) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc3716) + %3252 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3717) + %3253 = "ttir.multiply"(%3247, %3251, %3252) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3717) + %3254 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3718) + %3255 = "ttir.add"(%3231, %3253, %3254) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3718) + %3256 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3719) + %3257 = "ttir.squeeze"(%3255, %3256) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3719) + %3258 = tensor.empty() : tensor<12x3200xf32> loc(#loc3720) + %3259 = "ttir.matmul"(%3217, %arg472, %3258) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3720) + %3260 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3721) + %3261 = "ttir.reshape"(%3259, %3260) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3721) + %3262 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3722) + %3263 = "ttir.transpose"(%3261, %3262) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3722) + %3264 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3723) + %3265 = "ttir.multiply"(%3263, %3229, %3264) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3723) + %3266 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3724) + %3267 = "ttir.transpose"(%3263, %3266) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3724) + %3268 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3725) + %3269 = "ttir.matmul"(%arg226, %3267, %3268) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3725) + %3270 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3726) + %3271 = "ttir.transpose"(%3269, %3270) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3726) + %3272 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3727) + %3273 = "ttir.multiply"(%3271, %arg227, %3272) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3727) + %3274 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3728) + %3275 = "ttir.transpose"(%3263, %3274) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3728) + %3276 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3729) + %3277 = "ttir.matmul"(%arg228, %3275, %3276) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3729) + %3278 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3730) + %3279 = "ttir.transpose"(%3277, %3278) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3730) + %3280 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3731) + %3281 = "ttir.concat"(%3273, %3279, %3280) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3731) + %3282 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3732) + %3283 = "ttir.multiply"(%3281, %3251, %3282) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3732) + %3284 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3733) + %3285 = "ttir.add"(%3265, %3283, %3284) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3733) + %3286 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3734) + %3287 = "ttir.squeeze"(%3285, %3286) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3734) + %3288 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3735) + %3289 = "ttir.transpose"(%3287, %3288) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3735) + %3290 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3736) + %3291 = "ttir.matmul"(%3257, %3289, %3290) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3736) + %3292 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3737) + %3293 = "ttir.unsqueeze"(%3291, %3292) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3737) + %3294 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3738) + %3295 = "ttir.multiply"(%3293, %arg229, %3294) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3738) + %3296 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3739) + %3297 = "ttir.add"(%3295, %arg230, %3296) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3739) + %3298 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3740) + %3299 = "ttir.softmax"(%3297, %3298) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3740) + %3300 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3741) + %3301 = "ttir.squeeze"(%3299, %3300) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3741) + %3302 = tensor.empty() : tensor<12x3200xf32> loc(#loc3742) + %3303 = "ttir.matmul"(%3217, %arg473, %3302) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3742) + %3304 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3743) + %3305 = "ttir.reshape"(%3303, %3304) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3743) + %3306 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3744) + %3307 = "ttir.transpose"(%3305, %3306) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3744) + %3308 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3745) + %3309 = "ttir.transpose"(%3307, %3308) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3745) + %3310 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3746) + %3311 = "ttir.squeeze"(%3309, %3310) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3746) + %3312 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3747) + %3313 = "ttir.transpose"(%3311, %3312) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3747) + %3314 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3748) + %3315 = "ttir.matmul"(%3301, %3313, %3314) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3748) + %3316 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3749) + %3317 = "ttir.unsqueeze"(%3315, %3316) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3749) + %3318 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3750) + %3319 = "ttir.transpose"(%3317, %3318) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3750) + %3320 = tensor.empty() : tensor<12x3200xf32> loc(#loc3751) + %3321 = "ttir.reshape"(%3319, %3320) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3751) + %3322 = tensor.empty() : tensor<12x3200xf32> loc(#loc3752) + %3323 = "ttir.matmul"(%3321, %arg474, %3322) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3752) + %3324 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3753) + %3325 = "ttir.unsqueeze"(%3323, %3324) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3753) + %3326 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3754) + %3327 = "ttir.add"(%3201, %3325, %3326) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3754) + %3328 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3755) + %3329 = "ttir.multiply"(%3327, %3327, %3328) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3755) + %3330 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3756) + %3331 = "ttir.mean"(%3329, %3330) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3756) + %3332 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3757) + %3333 = "ttir.add"(%3331, %arg231, %3332) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3757) + %3334 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3758) + %3335 = "ttir.sqrt"(%3333, %3334) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3758) + %3336 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3759) + %3337 = "ttir.reciprocal"(%3335, %3336) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3759) + %3338 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3760) + %3339 = "ttir.multiply"(%3327, %3337, %3338) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3760) + %3340 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3761) + %3341 = "ttir.multiply"(%arg475, %3339, %3340) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3761) + %3342 = tensor.empty() : tensor<12x3200xf32> loc(#loc3762) + %3343 = "ttir.squeeze"(%3341, %3342) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3762) + %3344 = tensor.empty() : tensor<12x8640xf32> loc(#loc3763) + %3345 = "ttir.matmul"(%3343, %arg476, %3344) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc3763) + %3346 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3764) + %3347 = "ttir.unsqueeze"(%3345, %3346) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3764) + %3348 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3765) + %3349 = "ttir.sigmoid"(%3347, %3348) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3765) + %3350 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3766) + %3351 = "ttir.multiply"(%3347, %3349, %3350) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3766) + %3352 = tensor.empty() : tensor<12x8640xf32> loc(#loc3767) + %3353 = "ttir.matmul"(%3343, %arg477, %3352) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc3767) + %3354 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3768) + %3355 = "ttir.unsqueeze"(%3353, %3354) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3768) + %3356 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3769) + %3357 = "ttir.multiply"(%3351, %3355, %3356) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3769) + %3358 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3770) + %3359 = "ttir.matmul"(%3357, %arg478, %3358) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3770) + %3360 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3771) + %3361 = "ttir.add"(%3327, %3359, %3360) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3771) + %3362 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3772) + %3363 = "ttir.multiply"(%3361, %3361, %3362) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3772) + %3364 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3773) + %3365 = "ttir.mean"(%3363, %3364) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3773) + %3366 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3774) + %3367 = "ttir.add"(%3365, %arg232, %3366) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3774) + %3368 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3775) + %3369 = "ttir.sqrt"(%3367, %3368) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3775) + %3370 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3776) + %3371 = "ttir.reciprocal"(%3369, %3370) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3776) + %3372 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3777) + %3373 = "ttir.multiply"(%3361, %3371, %3372) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3777) + %3374 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3778) + %3375 = "ttir.multiply"(%arg479, %3373, %3374) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3778) + %3376 = tensor.empty() : tensor<12x3200xf32> loc(#loc3779) + %3377 = "ttir.squeeze"(%3375, %3376) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3779) + %3378 = tensor.empty() : tensor<12x3200xf32> loc(#loc3780) + %3379 = "ttir.matmul"(%3377, %arg480, %3378) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3780) + %3380 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3781) + %3381 = "ttir.reshape"(%3379, %3380) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3781) + %3382 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3782) + %3383 = "ttir.transpose"(%3381, %3382) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3782) + %3384 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3783) + %3385 = "ttir.concat"(%arg233, %arg233, %3384) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3783) + %3386 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3784) + %3387 = "ttir.sin"(%3385, %3386) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3784) + %3388 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc3785) + %3389 = "ttir.unsqueeze"(%3387, %3388) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc3785) + %3390 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3786) + %3391 = "ttir.multiply"(%3383, %3389, %3390) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3786) + %3392 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3787) + %3393 = "ttir.transpose"(%3383, %3392) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3787) + %3394 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3788) + %3395 = "ttir.matmul"(%arg234, %3393, %3394) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3788) + %3396 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3789) + %3397 = "ttir.transpose"(%3395, %3396) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3789) + %3398 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3790) + %3399 = "ttir.multiply"(%3397, %arg235, %3398) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3790) + %3400 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3791) + %3401 = "ttir.transpose"(%3383, %3400) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3791) + %3402 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3792) + %3403 = "ttir.matmul"(%arg236, %3401, %3402) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3792) + %3404 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3793) + %3405 = "ttir.transpose"(%3403, %3404) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3793) + %3406 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3794) + %3407 = "ttir.concat"(%3399, %3405, %3406) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3794) + %3408 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3795) + %3409 = "ttir.cos"(%3385, %3408) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3795) + %3410 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc3796) + %3411 = "ttir.unsqueeze"(%3409, %3410) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc3796) + %3412 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3797) + %3413 = "ttir.multiply"(%3407, %3411, %3412) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3797) + %3414 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3798) + %3415 = "ttir.add"(%3391, %3413, %3414) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3798) + %3416 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3799) + %3417 = "ttir.squeeze"(%3415, %3416) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3799) + %3418 = tensor.empty() : tensor<12x3200xf32> loc(#loc3800) + %3419 = "ttir.matmul"(%3377, %arg481, %3418) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3800) + %3420 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3801) + %3421 = "ttir.reshape"(%3419, %3420) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3801) + %3422 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3802) + %3423 = "ttir.transpose"(%3421, %3422) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3802) + %3424 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3803) + %3425 = "ttir.multiply"(%3423, %3389, %3424) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3803) + %3426 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3804) + %3427 = "ttir.transpose"(%3423, %3426) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3804) + %3428 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3805) + %3429 = "ttir.matmul"(%arg237, %3427, %3428) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3805) + %3430 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3806) + %3431 = "ttir.transpose"(%3429, %3430) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3806) + %3432 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3807) + %3433 = "ttir.multiply"(%3431, %arg238, %3432) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3807) + %3434 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3808) + %3435 = "ttir.transpose"(%3423, %3434) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3808) + %3436 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3809) + %3437 = "ttir.matmul"(%arg239, %3435, %3436) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3809) + %3438 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3810) + %3439 = "ttir.transpose"(%3437, %3438) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3810) + %3440 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3811) + %3441 = "ttir.concat"(%3433, %3439, %3440) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3811) + %3442 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3812) + %3443 = "ttir.multiply"(%3441, %3411, %3442) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3812) + %3444 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3813) + %3445 = "ttir.add"(%3425, %3443, %3444) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3813) + %3446 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3814) + %3447 = "ttir.squeeze"(%3445, %3446) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3814) + %3448 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3815) + %3449 = "ttir.transpose"(%3447, %3448) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3815) + %3450 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3816) + %3451 = "ttir.matmul"(%3417, %3449, %3450) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3816) + %3452 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3817) + %3453 = "ttir.unsqueeze"(%3451, %3452) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3817) + %3454 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3818) + %3455 = "ttir.multiply"(%3453, %arg240, %3454) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3818) + %3456 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3819) + %3457 = "ttir.add"(%3455, %arg241, %3456) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3819) + %3458 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3820) + %3459 = "ttir.softmax"(%3457, %3458) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3820) + %3460 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3821) + %3461 = "ttir.squeeze"(%3459, %3460) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3821) + %3462 = tensor.empty() : tensor<12x3200xf32> loc(#loc3822) + %3463 = "ttir.matmul"(%3377, %arg482, %3462) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3822) + %3464 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3823) + %3465 = "ttir.reshape"(%3463, %3464) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3823) + %3466 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3824) + %3467 = "ttir.transpose"(%3465, %3466) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3824) + %3468 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3825) + %3469 = "ttir.transpose"(%3467, %3468) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3825) + %3470 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3826) + %3471 = "ttir.squeeze"(%3469, %3470) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3826) + %3472 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3827) + %3473 = "ttir.transpose"(%3471, %3472) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3827) + %3474 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3828) + %3475 = "ttir.matmul"(%3461, %3473, %3474) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3828) + %3476 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3829) + %3477 = "ttir.unsqueeze"(%3475, %3476) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3829) + %3478 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3830) + %3479 = "ttir.transpose"(%3477, %3478) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3830) + %3480 = tensor.empty() : tensor<12x3200xf32> loc(#loc3831) + %3481 = "ttir.reshape"(%3479, %3480) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3831) + %3482 = tensor.empty() : tensor<12x3200xf32> loc(#loc3832) + %3483 = "ttir.matmul"(%3481, %arg483, %3482) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3832) + %3484 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3833) + %3485 = "ttir.unsqueeze"(%3483, %3484) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3833) + %3486 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3834) + %3487 = "ttir.add"(%3361, %3485, %3486) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3834) + %3488 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3835) + %3489 = "ttir.multiply"(%3487, %3487, %3488) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3835) + %3490 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3836) + %3491 = "ttir.mean"(%3489, %3490) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3836) + %3492 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3837) + %3493 = "ttir.add"(%3491, %arg242, %3492) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3837) + %3494 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3838) + %3495 = "ttir.sqrt"(%3493, %3494) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3838) + %3496 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3839) + %3497 = "ttir.reciprocal"(%3495, %3496) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3839) + %3498 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3840) + %3499 = "ttir.multiply"(%3487, %3497, %3498) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3840) + %3500 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3841) + %3501 = "ttir.multiply"(%arg484, %3499, %3500) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3841) + %3502 = tensor.empty() : tensor<12x3200xf32> loc(#loc3842) + %3503 = "ttir.squeeze"(%3501, %3502) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3842) + %3504 = tensor.empty() : tensor<12x8640xf32> loc(#loc3843) + %3505 = "ttir.matmul"(%3503, %arg485, %3504) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc3843) + %3506 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3844) + %3507 = "ttir.unsqueeze"(%3505, %3506) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3844) + %3508 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3845) + %3509 = "ttir.sigmoid"(%3507, %3508) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3845) + %3510 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3846) + %3511 = "ttir.multiply"(%3507, %3509, %3510) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3846) + %3512 = tensor.empty() : tensor<12x8640xf32> loc(#loc3847) + %3513 = "ttir.matmul"(%3503, %arg486, %3512) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc3847) + %3514 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3848) + %3515 = "ttir.unsqueeze"(%3513, %3514) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3848) + %3516 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3849) + %3517 = "ttir.multiply"(%3511, %3515, %3516) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3849) + %3518 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3850) + %3519 = "ttir.matmul"(%3517, %arg487, %3518) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3850) + %3520 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3851) + %3521 = "ttir.add"(%3487, %3519, %3520) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3851) + %3522 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3852) + %3523 = "ttir.multiply"(%3521, %3521, %3522) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3852) + %3524 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3853) + %3525 = "ttir.mean"(%3523, %3524) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3853) + %3526 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3854) + %3527 = "ttir.add"(%3525, %arg243, %3526) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3854) + %3528 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3855) + %3529 = "ttir.sqrt"(%3527, %3528) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3855) + %3530 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3856) + %3531 = "ttir.reciprocal"(%3529, %3530) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3856) + %3532 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3857) + %3533 = "ttir.multiply"(%3521, %3531, %3532) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3857) + %3534 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3858) + %3535 = "ttir.multiply"(%arg488, %3533, %3534) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3858) + %3536 = tensor.empty() : tensor<12x3200xf32> loc(#loc3859) + %3537 = "ttir.squeeze"(%3535, %3536) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3859) + %3538 = tensor.empty() : tensor<12x3200xf32> loc(#loc3860) + %3539 = "ttir.matmul"(%3537, %arg489, %3538) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3860) + %3540 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3861) + %3541 = "ttir.reshape"(%3539, %3540) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3861) + %3542 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3862) + %3543 = "ttir.transpose"(%3541, %3542) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3862) + %3544 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3863) + %3545 = "ttir.concat"(%arg244, %arg244, %3544) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3863) + %3546 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3864) + %3547 = "ttir.sin"(%3545, %3546) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3864) + %3548 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc3865) + %3549 = "ttir.unsqueeze"(%3547, %3548) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc3865) + %3550 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3866) + %3551 = "ttir.multiply"(%3543, %3549, %3550) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3866) + %3552 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3867) + %3553 = "ttir.transpose"(%3543, %3552) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3867) + %3554 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3868) + %3555 = "ttir.matmul"(%arg245, %3553, %3554) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3868) + %3556 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3869) + %3557 = "ttir.transpose"(%3555, %3556) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3869) + %3558 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3870) + %3559 = "ttir.multiply"(%3557, %arg246, %3558) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3870) + %3560 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3871) + %3561 = "ttir.transpose"(%3543, %3560) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3871) + %3562 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3872) + %3563 = "ttir.matmul"(%arg247, %3561, %3562) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3872) + %3564 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3873) + %3565 = "ttir.transpose"(%3563, %3564) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3873) + %3566 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3874) + %3567 = "ttir.concat"(%3559, %3565, %3566) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3874) + %3568 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3875) + %3569 = "ttir.cos"(%3545, %3568) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3875) + %3570 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc3876) + %3571 = "ttir.unsqueeze"(%3569, %3570) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc3876) + %3572 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3877) + %3573 = "ttir.multiply"(%3567, %3571, %3572) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3877) + %3574 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3878) + %3575 = "ttir.add"(%3551, %3573, %3574) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3878) + %3576 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3879) + %3577 = "ttir.squeeze"(%3575, %3576) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3879) + %3578 = tensor.empty() : tensor<12x3200xf32> loc(#loc3880) + %3579 = "ttir.matmul"(%3537, %arg490, %3578) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3880) + %3580 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3881) + %3581 = "ttir.reshape"(%3579, %3580) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3881) + %3582 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3882) + %3583 = "ttir.transpose"(%3581, %3582) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3882) + %3584 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3883) + %3585 = "ttir.multiply"(%3583, %3549, %3584) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3883) + %3586 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3884) + %3587 = "ttir.transpose"(%3583, %3586) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3884) + %3588 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3885) + %3589 = "ttir.matmul"(%arg248, %3587, %3588) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3885) + %3590 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3886) + %3591 = "ttir.transpose"(%3589, %3590) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3886) + %3592 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3887) + %3593 = "ttir.multiply"(%3591, %arg249, %3592) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3887) + %3594 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3888) + %3595 = "ttir.transpose"(%3583, %3594) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3888) + %3596 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3889) + %3597 = "ttir.matmul"(%arg250, %3595, %3596) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3889) + %3598 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3890) + %3599 = "ttir.transpose"(%3597, %3598) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3890) + %3600 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3891) + %3601 = "ttir.concat"(%3593, %3599, %3600) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3891) + %3602 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3892) + %3603 = "ttir.multiply"(%3601, %3571, %3602) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3892) + %3604 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3893) + %3605 = "ttir.add"(%3585, %3603, %3604) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3893) + %3606 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3894) + %3607 = "ttir.squeeze"(%3605, %3606) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3894) + %3608 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3895) + %3609 = "ttir.transpose"(%3607, %3608) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3895) + %3610 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3896) + %3611 = "ttir.matmul"(%3577, %3609, %3610) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3896) + %3612 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3897) + %3613 = "ttir.unsqueeze"(%3611, %3612) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3897) + %3614 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3898) + %3615 = "ttir.multiply"(%3613, %arg251, %3614) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3898) + %3616 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3899) + %3617 = "ttir.add"(%3615, %arg252, %3616) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3899) + %3618 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3900) + %3619 = "ttir.softmax"(%3617, %3618) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3900) + %3620 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3901) + %3621 = "ttir.squeeze"(%3619, %3620) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3901) + %3622 = tensor.empty() : tensor<12x3200xf32> loc(#loc3902) + %3623 = "ttir.matmul"(%3537, %arg491, %3622) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3902) + %3624 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3903) + %3625 = "ttir.reshape"(%3623, %3624) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3903) + %3626 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3904) + %3627 = "ttir.transpose"(%3625, %3626) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3904) + %3628 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3905) + %3629 = "ttir.transpose"(%3627, %3628) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3905) + %3630 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3906) + %3631 = "ttir.squeeze"(%3629, %3630) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3906) + %3632 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3907) + %3633 = "ttir.transpose"(%3631, %3632) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3907) + %3634 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3908) + %3635 = "ttir.matmul"(%3621, %3633, %3634) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3908) + %3636 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3909) + %3637 = "ttir.unsqueeze"(%3635, %3636) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3909) + %3638 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3910) + %3639 = "ttir.transpose"(%3637, %3638) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3910) + %3640 = tensor.empty() : tensor<12x3200xf32> loc(#loc3911) + %3641 = "ttir.reshape"(%3639, %3640) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3911) + %3642 = tensor.empty() : tensor<12x3200xf32> loc(#loc3912) + %3643 = "ttir.matmul"(%3641, %arg492, %3642) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3912) + %3644 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3913) + %3645 = "ttir.unsqueeze"(%3643, %3644) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3913) + %3646 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3914) + %3647 = "ttir.add"(%3521, %3645, %3646) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3914) + %3648 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3915) + %3649 = "ttir.multiply"(%3647, %3647, %3648) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3915) + %3650 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3916) + %3651 = "ttir.mean"(%3649, %3650) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3916) + %3652 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3917) + %3653 = "ttir.add"(%3651, %arg253, %3652) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3917) + %3654 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3918) + %3655 = "ttir.sqrt"(%3653, %3654) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3918) + %3656 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3919) + %3657 = "ttir.reciprocal"(%3655, %3656) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3919) + %3658 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3920) + %3659 = "ttir.multiply"(%3647, %3657, %3658) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3920) + %3660 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3921) + %3661 = "ttir.multiply"(%arg493, %3659, %3660) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3921) + %3662 = tensor.empty() : tensor<12x3200xf32> loc(#loc3922) + %3663 = "ttir.squeeze"(%3661, %3662) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3922) + %3664 = tensor.empty() : tensor<12x8640xf32> loc(#loc3923) + %3665 = "ttir.matmul"(%3663, %arg494, %3664) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc3923) + %3666 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3924) + %3667 = "ttir.unsqueeze"(%3665, %3666) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3924) + %3668 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3925) + %3669 = "ttir.sigmoid"(%3667, %3668) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3925) + %3670 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3926) + %3671 = "ttir.multiply"(%3667, %3669, %3670) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3926) + %3672 = tensor.empty() : tensor<12x8640xf32> loc(#loc3927) + %3673 = "ttir.matmul"(%3663, %arg495, %3672) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc3927) + %3674 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3928) + %3675 = "ttir.unsqueeze"(%3673, %3674) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3928) + %3676 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc3929) + %3677 = "ttir.multiply"(%3671, %3675, %3676) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc3929) + %3678 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3930) + %3679 = "ttir.matmul"(%3677, %arg496, %3678) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3930) + %3680 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3931) + %3681 = "ttir.add"(%3647, %3679, %3680) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3931) + %3682 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3932) + %3683 = "ttir.multiply"(%3681, %3681, %3682) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3932) + %3684 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3933) + %3685 = "ttir.mean"(%3683, %3684) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3933) + %3686 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3934) + %3687 = "ttir.add"(%3685, %arg254, %3686) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3934) + %3688 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3935) + %3689 = "ttir.sqrt"(%3687, %3688) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3935) + %3690 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3936) + %3691 = "ttir.reciprocal"(%3689, %3690) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3936) + %3692 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3937) + %3693 = "ttir.multiply"(%3681, %3691, %3692) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3937) + %3694 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3938) + %3695 = "ttir.multiply"(%arg497, %3693, %3694) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3938) + %3696 = tensor.empty() : tensor<12x3200xf32> loc(#loc3939) + %3697 = "ttir.squeeze"(%3695, %3696) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3939) + %3698 = tensor.empty() : tensor<12x3200xf32> loc(#loc3940) + %3699 = "ttir.matmul"(%3697, %arg498, %3698) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3940) + %3700 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3941) + %3701 = "ttir.reshape"(%3699, %3700) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3941) + %3702 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3942) + %3703 = "ttir.transpose"(%3701, %3702) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3942) + %3704 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3943) + %3705 = "ttir.concat"(%arg255, %arg255, %3704) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3943) + %3706 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3944) + %3707 = "ttir.sin"(%3705, %3706) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3944) + %3708 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc3945) + %3709 = "ttir.unsqueeze"(%3707, %3708) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc3945) + %3710 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3946) + %3711 = "ttir.multiply"(%3703, %3709, %3710) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3946) + %3712 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3947) + %3713 = "ttir.transpose"(%3703, %3712) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3947) + %3714 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3948) + %3715 = "ttir.matmul"(%arg256, %3713, %3714) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3948) + %3716 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3949) + %3717 = "ttir.transpose"(%3715, %3716) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3949) + %3718 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3950) + %3719 = "ttir.multiply"(%3717, %arg257, %3718) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3950) + %3720 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3951) + %3721 = "ttir.transpose"(%3703, %3720) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3951) + %3722 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3952) + %3723 = "ttir.matmul"(%arg258, %3721, %3722) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3952) + %3724 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3953) + %3725 = "ttir.transpose"(%3723, %3724) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3953) + %3726 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3954) + %3727 = "ttir.concat"(%3719, %3725, %3726) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3954) + %3728 = tensor.empty() : tensor<1x12x100xf32> loc(#loc3955) + %3729 = "ttir.cos"(%3705, %3728) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc3955) + %3730 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc3956) + %3731 = "ttir.unsqueeze"(%3729, %3730) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc3956) + %3732 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3957) + %3733 = "ttir.multiply"(%3727, %3731, %3732) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3957) + %3734 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3958) + %3735 = "ttir.add"(%3711, %3733, %3734) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3958) + %3736 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3959) + %3737 = "ttir.squeeze"(%3735, %3736) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3959) + %3738 = tensor.empty() : tensor<12x3200xf32> loc(#loc3960) + %3739 = "ttir.matmul"(%3697, %arg499, %3738) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3960) + %3740 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3961) + %3741 = "ttir.reshape"(%3739, %3740) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3961) + %3742 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3962) + %3743 = "ttir.transpose"(%3741, %3742) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3962) + %3744 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3963) + %3745 = "ttir.multiply"(%3743, %3709, %3744) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3963) + %3746 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3964) + %3747 = "ttir.transpose"(%3743, %3746) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3964) + %3748 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3965) + %3749 = "ttir.matmul"(%arg259, %3747, %3748) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3965) + %3750 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3966) + %3751 = "ttir.transpose"(%3749, %3750) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3966) + %3752 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3967) + %3753 = "ttir.multiply"(%3751, %arg260, %3752) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3967) + %3754 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3968) + %3755 = "ttir.transpose"(%3743, %3754) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3968) + %3756 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc3969) + %3757 = "ttir.matmul"(%arg261, %3755, %3756) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc3969) + %3758 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc3970) + %3759 = "ttir.transpose"(%3757, %3758) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc3970) + %3760 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3971) + %3761 = "ttir.concat"(%3753, %3759, %3760) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3971) + %3762 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3972) + %3763 = "ttir.multiply"(%3761, %3731, %3762) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3972) + %3764 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3973) + %3765 = "ttir.add"(%3745, %3763, %3764) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3973) + %3766 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3974) + %3767 = "ttir.squeeze"(%3765, %3766) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3974) + %3768 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3975) + %3769 = "ttir.transpose"(%3767, %3768) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3975) + %3770 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3976) + %3771 = "ttir.matmul"(%3737, %3769, %3770) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3976) + %3772 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3977) + %3773 = "ttir.unsqueeze"(%3771, %3772) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3977) + %3774 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3978) + %3775 = "ttir.multiply"(%3773, %arg262, %3774) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3978) + %3776 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3979) + %3777 = "ttir.add"(%3775, %arg263, %3776) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3979) + %3778 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc3980) + %3779 = "ttir.softmax"(%3777, %3778) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc3980) + %3780 = tensor.empty() : tensor<32x12x12xf32> loc(#loc3981) + %3781 = "ttir.squeeze"(%3779, %3780) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc3981) + %3782 = tensor.empty() : tensor<12x3200xf32> loc(#loc3982) + %3783 = "ttir.matmul"(%3697, %arg500, %3782) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3982) + %3784 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3983) + %3785 = "ttir.reshape"(%3783, %3784) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3983) + %3786 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3984) + %3787 = "ttir.transpose"(%3785, %3786) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3984) + %3788 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc3985) + %3789 = "ttir.transpose"(%3787, %3788) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc3985) + %3790 = tensor.empty() : tensor<32x100x12xf32> loc(#loc3986) + %3791 = "ttir.squeeze"(%3789, %3790) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc3986) + %3792 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3987) + %3793 = "ttir.transpose"(%3791, %3792) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3987) + %3794 = tensor.empty() : tensor<32x12x100xf32> loc(#loc3988) + %3795 = "ttir.matmul"(%3781, %3793, %3794) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc3988) + %3796 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc3989) + %3797 = "ttir.unsqueeze"(%3795, %3796) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc3989) + %3798 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc3990) + %3799 = "ttir.transpose"(%3797, %3798) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc3990) + %3800 = tensor.empty() : tensor<12x3200xf32> loc(#loc3991) + %3801 = "ttir.reshape"(%3799, %3800) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3991) + %3802 = tensor.empty() : tensor<12x3200xf32> loc(#loc3992) + %3803 = "ttir.matmul"(%3801, %arg501, %3802) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc3992) + %3804 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3993) + %3805 = "ttir.unsqueeze"(%3803, %3804) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3993) + %3806 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3994) + %3807 = "ttir.add"(%3681, %3805, %3806) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3994) + %3808 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc3995) + %3809 = "ttir.multiply"(%3807, %3807, %3808) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc3995) + %3810 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3996) + %3811 = "ttir.mean"(%3809, %3810) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3996) + %3812 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3997) + %3813 = "ttir.add"(%3811, %arg264, %3812) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3997) + %3814 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3998) + %3815 = "ttir.sqrt"(%3813, %3814) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3998) + %3816 = tensor.empty() : tensor<1x12x1xf32> loc(#loc3999) + %3817 = "ttir.reciprocal"(%3815, %3816) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc3999) + %3818 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4000) + %3819 = "ttir.multiply"(%3807, %3817, %3818) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4000) + %3820 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4001) + %3821 = "ttir.multiply"(%arg502, %3819, %3820) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4001) + %3822 = tensor.empty() : tensor<12x3200xf32> loc(#loc4002) + %3823 = "ttir.squeeze"(%3821, %3822) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc4002) + %3824 = tensor.empty() : tensor<12x8640xf32> loc(#loc4003) + %3825 = "ttir.matmul"(%3823, %arg503, %3824) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc4003) + %3826 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc4004) + %3827 = "ttir.unsqueeze"(%3825, %3826) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc4004) + %3828 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc4005) + %3829 = "ttir.sigmoid"(%3827, %3828) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc4005) + %3830 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc4006) + %3831 = "ttir.multiply"(%3827, %3829, %3830) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc4006) + %3832 = tensor.empty() : tensor<12x8640xf32> loc(#loc4007) + %3833 = "ttir.matmul"(%3823, %arg504, %3832) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc4007) + %3834 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc4008) + %3835 = "ttir.unsqueeze"(%3833, %3834) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc4008) + %3836 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc4009) + %3837 = "ttir.multiply"(%3831, %3835, %3836) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc4009) + %3838 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4010) + %3839 = "ttir.matmul"(%3837, %arg505, %3838) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4010) + %3840 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4011) + %3841 = "ttir.add"(%3807, %3839, %3840) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4011) + %3842 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4012) + %3843 = "ttir.multiply"(%3841, %3841, %3842) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4012) + %3844 = tensor.empty() : tensor<1x12x1xf32> loc(#loc4013) + %3845 = "ttir.mean"(%3843, %3844) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc4013) + %3846 = tensor.empty() : tensor<1x12x1xf32> loc(#loc4014) + %3847 = "ttir.add"(%3845, %arg265, %3846) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc4014) + %3848 = tensor.empty() : tensor<1x12x1xf32> loc(#loc4015) + %3849 = "ttir.sqrt"(%3847, %3848) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc4015) + %3850 = tensor.empty() : tensor<1x12x1xf32> loc(#loc4016) + %3851 = "ttir.reciprocal"(%3849, %3850) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc4016) + %3852 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4017) + %3853 = "ttir.multiply"(%3841, %3851, %3852) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4017) + %3854 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4018) + %3855 = "ttir.multiply"(%arg506, %3853, %3854) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4018) + %3856 = tensor.empty() : tensor<12x3200xf32> loc(#loc4019) + %3857 = "ttir.squeeze"(%3855, %3856) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc4019) + %3858 = tensor.empty() : tensor<12x3200xf32> loc(#loc4020) + %3859 = "ttir.matmul"(%3857, %arg507, %3858) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc4020) + %3860 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc4021) + %3861 = "ttir.reshape"(%3859, %3860) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc4021) + %3862 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc4022) + %3863 = "ttir.transpose"(%3861, %3862) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc4022) + %3864 = tensor.empty() : tensor<1x12x100xf32> loc(#loc4023) + %3865 = "ttir.concat"(%arg266, %arg266, %3864) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc4023) + %3866 = tensor.empty() : tensor<1x12x100xf32> loc(#loc4024) + %3867 = "ttir.sin"(%3865, %3866) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc4024) + %3868 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc4025) + %3869 = "ttir.unsqueeze"(%3867, %3868) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc4025) + %3870 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc4026) + %3871 = "ttir.multiply"(%3863, %3869, %3870) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc4026) + %3872 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc4027) + %3873 = "ttir.transpose"(%3863, %3872) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc4027) + %3874 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc4028) + %3875 = "ttir.matmul"(%arg267, %3873, %3874) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc4028) + %3876 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc4029) + %3877 = "ttir.transpose"(%3875, %3876) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc4029) + %3878 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc4030) + %3879 = "ttir.multiply"(%3877, %arg268, %3878) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc4030) + %3880 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc4031) + %3881 = "ttir.transpose"(%3863, %3880) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc4031) + %3882 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc4032) + %3883 = "ttir.matmul"(%arg269, %3881, %3882) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc4032) + %3884 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc4033) + %3885 = "ttir.transpose"(%3883, %3884) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc4033) + %3886 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc4034) + %3887 = "ttir.concat"(%3879, %3885, %3886) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc4034) + %3888 = tensor.empty() : tensor<1x12x100xf32> loc(#loc4035) + %3889 = "ttir.cos"(%3865, %3888) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc4035) + %3890 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc4036) + %3891 = "ttir.unsqueeze"(%3889, %3890) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc4036) + %3892 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc4037) + %3893 = "ttir.multiply"(%3887, %3891, %3892) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc4037) + %3894 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc4038) + %3895 = "ttir.add"(%3871, %3893, %3894) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc4038) + %3896 = tensor.empty() : tensor<32x12x100xf32> loc(#loc4039) + %3897 = "ttir.squeeze"(%3895, %3896) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc4039) + %3898 = tensor.empty() : tensor<12x3200xf32> loc(#loc4040) + %3899 = "ttir.matmul"(%3857, %arg508, %3898) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc4040) + %3900 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc4041) + %3901 = "ttir.reshape"(%3899, %3900) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc4041) + %3902 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc4042) + %3903 = "ttir.transpose"(%3901, %3902) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc4042) + %3904 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc4043) + %3905 = "ttir.multiply"(%3903, %3869, %3904) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc4043) + %3906 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc4044) + %3907 = "ttir.transpose"(%3903, %3906) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc4044) + %3908 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc4045) + %3909 = "ttir.matmul"(%arg270, %3907, %3908) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc4045) + %3910 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc4046) + %3911 = "ttir.transpose"(%3909, %3910) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc4046) + %3912 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc4047) + %3913 = "ttir.multiply"(%3911, %arg271, %3912) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc4047) + %3914 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc4048) + %3915 = "ttir.transpose"(%3903, %3914) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc4048) + %3916 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc4049) + %3917 = "ttir.matmul"(%arg272, %3915, %3916) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc4049) + %3918 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc4050) + %3919 = "ttir.transpose"(%3917, %3918) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc4050) + %3920 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc4051) + %3921 = "ttir.concat"(%3913, %3919, %3920) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc4051) + %3922 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc4052) + %3923 = "ttir.multiply"(%3921, %3891, %3922) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc4052) + %3924 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc4053) + %3925 = "ttir.add"(%3905, %3923, %3924) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc4053) + %3926 = tensor.empty() : tensor<32x12x100xf32> loc(#loc4054) + %3927 = "ttir.squeeze"(%3925, %3926) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc4054) + %3928 = tensor.empty() : tensor<32x100x12xf32> loc(#loc4055) + %3929 = "ttir.transpose"(%3927, %3928) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc4055) + %3930 = tensor.empty() : tensor<32x12x12xf32> loc(#loc4056) + %3931 = "ttir.matmul"(%3897, %3929, %3930) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc4056) + %3932 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc4057) + %3933 = "ttir.unsqueeze"(%3931, %3932) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc4057) + %3934 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc4058) + %3935 = "ttir.multiply"(%3933, %arg273, %3934) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc4058) + %3936 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc4059) + %3937 = "ttir.add"(%3935, %arg274, %3936) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc4059) + %3938 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc4060) + %3939 = "ttir.softmax"(%3937, %3938) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc4060) + %3940 = tensor.empty() : tensor<32x12x12xf32> loc(#loc4061) + %3941 = "ttir.squeeze"(%3939, %3940) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc4061) + %3942 = tensor.empty() : tensor<12x3200xf32> loc(#loc4062) + %3943 = "ttir.matmul"(%3857, %arg509, %3942) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc4062) + %3944 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc4063) + %3945 = "ttir.reshape"(%3943, %3944) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc4063) + %3946 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc4064) + %3947 = "ttir.transpose"(%3945, %3946) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc4064) + %3948 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc4065) + %3949 = "ttir.transpose"(%3947, %3948) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc4065) + %3950 = tensor.empty() : tensor<32x100x12xf32> loc(#loc4066) + %3951 = "ttir.squeeze"(%3949, %3950) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc4066) + %3952 = tensor.empty() : tensor<32x12x100xf32> loc(#loc4067) + %3953 = "ttir.transpose"(%3951, %3952) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc4067) + %3954 = tensor.empty() : tensor<32x12x100xf32> loc(#loc4068) + %3955 = "ttir.matmul"(%3941, %3953, %3954) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc4068) + %3956 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc4069) + %3957 = "ttir.unsqueeze"(%3955, %3956) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc4069) + %3958 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc4070) + %3959 = "ttir.transpose"(%3957, %3958) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc4070) + %3960 = tensor.empty() : tensor<12x3200xf32> loc(#loc4071) + %3961 = "ttir.reshape"(%3959, %3960) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc4071) + %3962 = tensor.empty() : tensor<12x3200xf32> loc(#loc4072) + %3963 = "ttir.matmul"(%3961, %arg510, %3962) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc4072) + %3964 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4073) + %3965 = "ttir.unsqueeze"(%3963, %3964) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4073) + %3966 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4074) + %3967 = "ttir.add"(%3841, %3965, %3966) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4074) + %3968 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4075) + %3969 = "ttir.multiply"(%3967, %3967, %3968) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4075) + %3970 = tensor.empty() : tensor<1x12x1xf32> loc(#loc4076) + %3971 = "ttir.mean"(%3969, %3970) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc4076) + %3972 = tensor.empty() : tensor<1x12x1xf32> loc(#loc4077) + %3973 = "ttir.add"(%3971, %arg275, %3972) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc4077) + %3974 = tensor.empty() : tensor<1x12x1xf32> loc(#loc4078) + %3975 = "ttir.sqrt"(%3973, %3974) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc4078) + %3976 = tensor.empty() : tensor<1x12x1xf32> loc(#loc4079) + %3977 = "ttir.reciprocal"(%3975, %3976) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc4079) + %3978 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4080) + %3979 = "ttir.multiply"(%3967, %3977, %3978) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4080) + %3980 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4081) + %3981 = "ttir.multiply"(%arg511, %3979, %3980) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4081) + %3982 = tensor.empty() : tensor<12x3200xf32> loc(#loc4082) + %3983 = "ttir.squeeze"(%3981, %3982) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc4082) + %3984 = tensor.empty() : tensor<12x8640xf32> loc(#loc4083) + %3985 = "ttir.matmul"(%3983, %arg512, %3984) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc4083) + %3986 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc4084) + %3987 = "ttir.unsqueeze"(%3985, %3986) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc4084) + %3988 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc4085) + %3989 = "ttir.sigmoid"(%3987, %3988) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc4085) + %3990 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc4086) + %3991 = "ttir.multiply"(%3987, %3989, %3990) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc4086) + %3992 = tensor.empty() : tensor<12x8640xf32> loc(#loc4087) + %3993 = "ttir.matmul"(%3983, %arg513, %3992) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc4087) + %3994 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc4088) + %3995 = "ttir.unsqueeze"(%3993, %3994) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc4088) + %3996 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc4089) + %3997 = "ttir.multiply"(%3991, %3995, %3996) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc4089) + %3998 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4090) + %3999 = "ttir.matmul"(%3997, %arg514, %3998) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4090) + %4000 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4091) + %4001 = "ttir.add"(%3967, %3999, %4000) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4091) + %4002 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4092) + %4003 = "ttir.multiply"(%4001, %4001, %4002) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4092) + %4004 = tensor.empty() : tensor<1x12x1xf32> loc(#loc4093) + %4005 = "ttir.mean"(%4003, %4004) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc4093) + %4006 = tensor.empty() : tensor<1x12x1xf32> loc(#loc4094) + %4007 = "ttir.add"(%4005, %arg276, %4006) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc4094) + %4008 = tensor.empty() : tensor<1x12x1xf32> loc(#loc4095) + %4009 = "ttir.sqrt"(%4007, %4008) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc4095) + %4010 = tensor.empty() : tensor<1x12x1xf32> loc(#loc4096) + %4011 = "ttir.reciprocal"(%4009, %4010) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc4096) + %4012 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4097) + %4013 = "ttir.multiply"(%4001, %4011, %4012) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4097) + %4014 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4098) + %4015 = "ttir.multiply"(%arg515, %4013, %4014) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4098) + %4016 = tensor.empty() : tensor<12x3200xf32> loc(#loc4099) + %4017 = "ttir.squeeze"(%4015, %4016) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc4099) + %4018 = tensor.empty() : tensor<12x3200xf32> loc(#loc4100) + %4019 = "ttir.matmul"(%4017, %arg516, %4018) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc4100) + %4020 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc4101) + %4021 = "ttir.reshape"(%4019, %4020) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc4101) + %4022 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc4102) + %4023 = "ttir.transpose"(%4021, %4022) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc4102) + %4024 = tensor.empty() : tensor<1x12x100xf32> loc(#loc4103) + %4025 = "ttir.concat"(%arg277, %arg277, %4024) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x50xf32>, tensor<1x12x50xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc4103) + %4026 = tensor.empty() : tensor<1x12x100xf32> loc(#loc4104) + %4027 = "ttir.sin"(%4025, %4026) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc4104) + %4028 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc4105) + %4029 = "ttir.unsqueeze"(%4027, %4028) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc4105) + %4030 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc4106) + %4031 = "ttir.multiply"(%4023, %4029, %4030) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc4106) + %4032 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc4107) + %4033 = "ttir.transpose"(%4023, %4032) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc4107) + %4034 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc4108) + %4035 = "ttir.matmul"(%arg278, %4033, %4034) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc4108) + %4036 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc4109) + %4037 = "ttir.transpose"(%4035, %4036) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc4109) + %4038 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc4110) + %4039 = "ttir.multiply"(%4037, %arg279, %4038) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc4110) + %4040 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc4111) + %4041 = "ttir.transpose"(%4023, %4040) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc4111) + %4042 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc4112) + %4043 = "ttir.matmul"(%arg280, %4041, %4042) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc4112) + %4044 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc4113) + %4045 = "ttir.transpose"(%4043, %4044) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc4113) + %4046 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc4114) + %4047 = "ttir.concat"(%4039, %4045, %4046) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc4114) + %4048 = tensor.empty() : tensor<1x12x100xf32> loc(#loc4115) + %4049 = "ttir.cos"(%4025, %4048) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x12x100xf32>) -> tensor<1x12x100xf32> loc(#loc4115) + %4050 = tensor.empty() : tensor<1x1x12x100xf32> loc(#loc4116) + %4051 = "ttir.unsqueeze"(%4049, %4050) <{dim = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x100xf32>, tensor<1x1x12x100xf32>) -> tensor<1x1x12x100xf32> loc(#loc4116) + %4052 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc4117) + %4053 = "ttir.multiply"(%4047, %4051, %4052) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc4117) + %4054 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc4118) + %4055 = "ttir.add"(%4031, %4053, %4054) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc4118) + %4056 = tensor.empty() : tensor<32x12x100xf32> loc(#loc4119) + %4057 = "ttir.squeeze"(%4055, %4056) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc4119) + %4058 = tensor.empty() : tensor<12x3200xf32> loc(#loc4120) + %4059 = "ttir.matmul"(%4017, %arg517, %4058) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc4120) + %4060 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc4121) + %4061 = "ttir.reshape"(%4059, %4060) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc4121) + %4062 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc4122) + %4063 = "ttir.transpose"(%4061, %4062) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc4122) + %4064 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc4123) + %4065 = "ttir.multiply"(%4063, %4029, %4064) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc4123) + %4066 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc4124) + %4067 = "ttir.transpose"(%4063, %4066) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc4124) + %4068 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc4125) + %4069 = "ttir.matmul"(%arg281, %4067, %4068) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc4125) + %4070 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc4126) + %4071 = "ttir.transpose"(%4069, %4070) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc4126) + %4072 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc4127) + %4073 = "ttir.multiply"(%4071, %arg282, %4072) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc4127) + %4074 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc4128) + %4075 = "ttir.transpose"(%4063, %4074) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc4128) + %4076 = tensor.empty() : tensor<1x32x50x12xf32> loc(#loc4129) + %4077 = "ttir.matmul"(%arg283, %4075, %4076) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x50x100xf32>, tensor<1x32x100x12xf32>, tensor<1x32x50x12xf32>) -> tensor<1x32x50x12xf32> loc(#loc4129) + %4078 = tensor.empty() : tensor<1x32x12x50xf32> loc(#loc4130) + %4079 = "ttir.transpose"(%4077, %4078) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x50x12xf32>, tensor<1x32x12x50xf32>) -> tensor<1x32x12x50xf32> loc(#loc4130) + %4080 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc4131) + %4081 = "ttir.concat"(%4073, %4079, %4080) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x50xf32>, tensor<1x32x12x50xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc4131) + %4082 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc4132) + %4083 = "ttir.multiply"(%4081, %4051, %4082) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x1x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc4132) + %4084 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc4133) + %4085 = "ttir.add"(%4065, %4083, %4084) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc4133) + %4086 = tensor.empty() : tensor<32x12x100xf32> loc(#loc4134) + %4087 = "ttir.squeeze"(%4085, %4086) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc4134) + %4088 = tensor.empty() : tensor<32x100x12xf32> loc(#loc4135) + %4089 = "ttir.transpose"(%4087, %4088) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc4135) + %4090 = tensor.empty() : tensor<32x12x12xf32> loc(#loc4136) + %4091 = "ttir.matmul"(%4057, %4089, %4090) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<32x100x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc4136) + %4092 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc4137) + %4093 = "ttir.unsqueeze"(%4091, %4092) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc4137) + %4094 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc4138) + %4095 = "ttir.multiply"(%4093, %arg284, %4094) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc4138) + %4096 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc4139) + %4097 = "ttir.add"(%4095, %arg285, %4096) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x1x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc4139) + %4098 = tensor.empty() : tensor<1x32x12x12xf32> loc(#loc4140) + %4099 = "ttir.softmax"(%4097, %4098) <{dimension = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<1x32x12x12xf32>) -> tensor<1x32x12x12xf32> loc(#loc4140) + %4100 = tensor.empty() : tensor<32x12x12xf32> loc(#loc4141) + %4101 = "ttir.squeeze"(%4099, %4100) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x12xf32>, tensor<32x12x12xf32>) -> tensor<32x12x12xf32> loc(#loc4141) + %4102 = tensor.empty() : tensor<12x3200xf32> loc(#loc4142) + %4103 = "ttir.matmul"(%4017, %arg518, %4102) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc4142) + %4104 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc4143) + %4105 = "ttir.reshape"(%4103, %4104) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 12 : i32, 32 : i32, 100 : i32]}> : (tensor<12x3200xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc4143) + %4106 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc4144) + %4107 = "ttir.transpose"(%4105, %4106) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x32x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc4144) + %4108 = tensor.empty() : tensor<1x32x100x12xf32> loc(#loc4145) + %4109 = "ttir.transpose"(%4107, %4108) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x32x100x12xf32>) -> tensor<1x32x100x12xf32> loc(#loc4145) + %4110 = tensor.empty() : tensor<32x100x12xf32> loc(#loc4146) + %4111 = "ttir.squeeze"(%4109, %4110) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x100x12xf32>, tensor<32x100x12xf32>) -> tensor<32x100x12xf32> loc(#loc4146) + %4112 = tensor.empty() : tensor<32x12x100xf32> loc(#loc4147) + %4113 = "ttir.transpose"(%4111, %4112) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x100x12xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc4147) + %4114 = tensor.empty() : tensor<32x12x100xf32> loc(#loc4148) + %4115 = "ttir.matmul"(%4101, %4113, %4114) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<32x12x12xf32>, tensor<32x12x100xf32>, tensor<32x12x100xf32>) -> tensor<32x12x100xf32> loc(#loc4148) + %4116 = tensor.empty() : tensor<1x32x12x100xf32> loc(#loc4149) + %4117 = "ttir.unsqueeze"(%4115, %4116) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<32x12x100xf32>, tensor<1x32x12x100xf32>) -> tensor<1x32x12x100xf32> loc(#loc4149) + %4118 = tensor.empty() : tensor<1x12x32x100xf32> loc(#loc4150) + %4119 = "ttir.transpose"(%4117, %4118) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x32x12x100xf32>, tensor<1x12x32x100xf32>) -> tensor<1x12x32x100xf32> loc(#loc4150) + %4120 = tensor.empty() : tensor<12x3200xf32> loc(#loc4151) + %4121 = "ttir.reshape"(%4119, %4120) <{operand_constraints = [#any_device, #any_device], shape = [12 : i32, 3200 : i32]}> : (tensor<1x12x32x100xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc4151) + %4122 = tensor.empty() : tensor<12x3200xf32> loc(#loc4152) + %4123 = "ttir.matmul"(%4121, %arg519, %4122) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc4152) + %4124 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4153) + %4125 = "ttir.unsqueeze"(%4123, %4124) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4153) + %4126 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4154) + %4127 = "ttir.add"(%4001, %4125, %4126) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4154) + %4128 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4155) + %4129 = "ttir.multiply"(%4127, %4127, %4128) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4155) + %4130 = tensor.empty() : tensor<1x12x1xf32> loc(#loc4156) + %4131 = "ttir.mean"(%4129, %4130) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc4156) + %4132 = tensor.empty() : tensor<1x12x1xf32> loc(#loc4157) + %4133 = "ttir.add"(%4131, %arg286, %4132) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc4157) + %4134 = tensor.empty() : tensor<1x12x1xf32> loc(#loc4158) + %4135 = "ttir.sqrt"(%4133, %4134) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc4158) + %4136 = tensor.empty() : tensor<1x12x1xf32> loc(#loc4159) + %4137 = "ttir.reciprocal"(%4135, %4136) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc4159) + %4138 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4160) + %4139 = "ttir.multiply"(%4127, %4137, %4138) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4160) + %4140 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4161) + %4141 = "ttir.multiply"(%arg520, %4139, %4140) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4161) + %4142 = tensor.empty() : tensor<12x3200xf32> loc(#loc4162) + %4143 = "ttir.squeeze"(%4141, %4142) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<12x3200xf32>) -> tensor<12x3200xf32> loc(#loc4162) + %4144 = tensor.empty() : tensor<12x8640xf32> loc(#loc4163) + %4145 = "ttir.matmul"(%4143, %arg521, %4144) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc4163) + %4146 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc4164) + %4147 = "ttir.unsqueeze"(%4145, %4146) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc4164) + %4148 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc4165) + %4149 = "ttir.sigmoid"(%4147, %4148) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc4165) + %4150 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc4166) + %4151 = "ttir.multiply"(%4147, %4149, %4150) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc4166) + %4152 = tensor.empty() : tensor<12x8640xf32> loc(#loc4167) + %4153 = "ttir.matmul"(%4143, %arg522, %4152) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<12x3200xf32>, tensor<3200x8640xf32>, tensor<12x8640xf32>) -> tensor<12x8640xf32> loc(#loc4167) + %4154 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc4168) + %4155 = "ttir.unsqueeze"(%4153, %4154) <{dim = 0 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc4168) + %4156 = tensor.empty() : tensor<1x12x8640xf32> loc(#loc4169) + %4157 = "ttir.multiply"(%4151, %4155, %4156) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<1x12x8640xf32>, tensor<1x12x8640xf32>) -> tensor<1x12x8640xf32> loc(#loc4169) + %4158 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4170) + %4159 = "ttir.matmul"(%4157, %arg523, %4158) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x8640xf32>, tensor<8640x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4170) + %4160 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4171) + %4161 = "ttir.add"(%4127, %4159, %4160) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4171) + %4162 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4172) + %4163 = "ttir.multiply"(%4161, %4161, %4162) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4172) + %4164 = tensor.empty() : tensor<1x12x1xf32> loc(#loc4173) + %4165 = "ttir.mean"(%4163, %4164) <{dim_arg = [-1 : i32], keep_dim = true, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc4173) + %4166 = tensor.empty() : tensor<1x12x1xf32> loc(#loc4174) + %4167 = "ttir.add"(%4165, %arg287, %4166) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc4174) + %4168 = tensor.empty() : tensor<1x12x1xf32> loc(#loc4175) + %4169 = "ttir.sqrt"(%4167, %4168) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc4175) + %4170 = tensor.empty() : tensor<1x12x1xf32> loc(#loc4176) + %4171 = "ttir.reciprocal"(%4169, %4170) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x12x1xf32>, tensor<1x12x1xf32>) -> tensor<1x12x1xf32> loc(#loc4176) + %4172 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4177) + %4173 = "ttir.multiply"(%4161, %4171, %4172) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<1x12x1xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4177) + %4174 = tensor.empty() : tensor<1x12x3200xf32> loc(#loc4178) + %4175 = "ttir.multiply"(%arg288, %4173, %4174) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<3200xf32>, tensor<1x12x3200xf32>, tensor<1x12x3200xf32>) -> tensor<1x12x3200xf32> loc(#loc4178) + %4176 = tensor.empty() : tensor<1x12x32000xf32> loc(#loc4179) + %4177 = "ttir.matmul"(%4175, %arg524, %4176) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x12x3200xf32>, tensor<3200x32000xf32>, tensor<1x12x32000xf32>) -> tensor<1x12x32000xf32> loc(#loc4179) + return %4177 : tensor<1x12x32000xf32> loc(#loc2090) + } loc(#loc) +} loc(#loc) +#loc1 = loc("forward":4294967295:4184) +#loc2 = loc("forward":4294967295:4186) +#loc3 = loc("forward":4294967295:4187) +#loc4 = loc("forward":4294967295:4189) +#loc5 = loc("forward":4294967295:4190) +#loc6 = loc("forward":4294967295:4191) +#loc7 = loc("forward":4294967295:4192) +#loc8 = loc("forward":4294967295:4193) +#loc9 = loc("forward":4294967295:4194) +#loc10 = loc("forward":4294967295:4196) +#loc11 = loc("forward":4294967295:4197) +#loc12 = loc("forward":4294967295:4198) +#loc13 = loc("forward":4294967295:4200) +#loc14 = loc("forward":4294967295:4201) +#loc15 = loc("forward":4294967295:4202) +#loc16 = loc("forward":4294967295:4203) +#loc17 = loc("forward":4294967295:4205) +#loc18 = loc("forward":4294967295:4206) +#loc19 = loc("forward":4294967295:4207) +#loc20 = loc("forward":4294967295:4209) +#loc21 = loc("forward":4294967295:4211) +#loc22 = loc("forward":4294967295:4212) +#loc23 = loc("forward":4294967295:4213) +#loc24 = loc("forward":4294967295:4214) +#loc25 = loc("forward":4294967295:4215) +#loc26 = loc("forward":4294967295:4216) +#loc27 = loc("forward":4294967295:4217) +#loc28 = loc("forward":4294967295:4218) +#loc29 = loc("forward":4294967295:4219) +#loc30 = loc("forward":4294967295:4221) +#loc31 = loc("forward":4294967295:4222) +#loc32 = loc("forward":4294967295:4223) +#loc33 = loc("forward":4294967295:4224) +#loc34 = loc("forward":4294967295:4226) +#loc35 = loc("forward":4294967295:4227) +#loc36 = loc("forward":4294967295:4228) +#loc37 = loc("forward":4294967295:4230) +#loc38 = loc("forward":4294967295:4232) +#loc39 = loc("forward":4294967295:4233) +#loc40 = loc("forward":4294967295:4234) +#loc41 = loc("forward":4294967295:4235) +#loc42 = loc("forward":4294967295:4236) +#loc43 = loc("forward":4294967295:4237) +#loc44 = loc("forward":4294967295:4238) +#loc45 = loc("forward":4294967295:4239) +#loc46 = loc("forward":4294967295:4240) +#loc47 = loc("forward":4294967295:4241) +#loc48 = loc("forward":4294967295:4243) +#loc49 = loc("forward":4294967295:4245) +#loc50 = loc("forward":4294967295:4246) +#loc51 = loc("forward":4294967295:4247) +#loc52 = loc("forward":4294967295:4249) +#loc53 = loc("forward":4294967295:4250) +#loc54 = loc("forward":4294967295:4251) +#loc55 = loc("forward":4294967295:4252) +#loc56 = loc("forward":4294967295:4253) +#loc57 = loc("forward":4294967295:4254) +#loc58 = loc("forward":4294967295:4255) +#loc59 = loc("forward":4294967295:4256) +#loc60 = loc("forward":4294967295:4257) +#loc61 = loc("forward":4294967295:4258) +#loc62 = loc("forward":4294967295:4260) +#loc63 = loc("forward":4294967295:4261) +#loc64 = loc("forward":4294967295:4262) +#loc65 = loc("forward":4294967295:4264) +#loc66 = loc("forward":4294967295:4265) +#loc67 = loc("forward":4294967295:4267) +#loc68 = loc("forward":4294967295:4268) +#loc69 = loc("forward":4294967295:4269) +#loc70 = loc("forward":4294967295:4270) +#loc71 = loc("forward":4294967295:4271) +#loc72 = loc("forward":4294967295:4272) +#loc73 = loc("forward":4294967295:4274) +#loc74 = loc("forward":4294967295:4275) +#loc75 = loc("forward":4294967295:4276) +#loc76 = loc("forward":4294967295:4277) +#loc77 = loc("forward":4294967295:4279) +#loc78 = loc("forward":4294967295:4280) +#loc79 = loc("forward":4294967295:4281) +#loc80 = loc("forward":4294967295:4283) +#loc81 = loc("forward":4294967295:4284) +#loc82 = loc("forward":4294967295:4286) +#loc83 = loc("forward":4294967295:4287) +#loc84 = loc("forward":4294967295:4289) +#loc85 = loc("forward":4294967295:4290) +#loc86 = loc("forward":4294967295:4291) +#loc87 = loc("forward":4294967295:4292) +#loc88 = loc("forward":4294967295:4293) +#loc89 = loc("forward":4294967295:4294) +#loc90 = loc("forward":4294967295:4296) +#loc91 = loc("forward":4294967295:4297) +#loc92 = loc("forward":4294967295:4298) +#loc93 = loc("forward":4294967295:4300) +#loc94 = loc("forward":4294967295:4301) +#loc95 = loc("forward":4294967295:4302) +#loc96 = loc("forward":4294967295:4303) +#loc97 = loc("forward":4294967295:4305) +#loc98 = loc("forward":4294967295:4306) +#loc99 = loc("forward":4294967295:4307) +#loc100 = loc("forward":4294967295:4309) +#loc101 = loc("forward":4294967295:4311) +#loc102 = loc("forward":4294967295:4312) +#loc103 = loc("forward":4294967295:4313) +#loc104 = loc("forward":4294967295:4314) +#loc105 = loc("forward":4294967295:4315) +#loc106 = loc("forward":4294967295:4316) +#loc107 = loc("forward":4294967295:4317) +#loc108 = loc("forward":4294967295:4318) +#loc109 = loc("forward":4294967295:4319) +#loc110 = loc("forward":4294967295:4321) +#loc111 = loc("forward":4294967295:4322) +#loc112 = loc("forward":4294967295:4323) +#loc113 = loc("forward":4294967295:4324) +#loc114 = loc("forward":4294967295:4326) +#loc115 = loc("forward":4294967295:4327) +#loc116 = loc("forward":4294967295:4328) +#loc117 = loc("forward":4294967295:4330) +#loc118 = loc("forward":4294967295:4332) +#loc119 = loc("forward":4294967295:4333) +#loc120 = loc("forward":4294967295:4334) +#loc121 = loc("forward":4294967295:4335) +#loc122 = loc("forward":4294967295:4336) +#loc123 = loc("forward":4294967295:4337) +#loc124 = loc("forward":4294967295:4338) +#loc125 = loc("forward":4294967295:4339) +#loc126 = loc("forward":4294967295:4340) +#loc127 = loc("forward":4294967295:4341) +#loc128 = loc("forward":4294967295:4343) +#loc129 = loc("forward":4294967295:4345) +#loc130 = loc("forward":4294967295:4346) +#loc131 = loc("forward":4294967295:4347) +#loc132 = loc("forward":4294967295:4349) +#loc133 = loc("forward":4294967295:4350) +#loc134 = loc("forward":4294967295:4351) +#loc135 = loc("forward":4294967295:4352) +#loc136 = loc("forward":4294967295:4353) +#loc137 = loc("forward":4294967295:4354) +#loc138 = loc("forward":4294967295:4355) +#loc139 = loc("forward":4294967295:4356) +#loc140 = loc("forward":4294967295:4357) +#loc141 = loc("forward":4294967295:4358) +#loc142 = loc("forward":4294967295:4360) +#loc143 = loc("forward":4294967295:4361) +#loc144 = loc("forward":4294967295:4362) +#loc145 = loc("forward":4294967295:4364) +#loc146 = loc("forward":4294967295:4365) +#loc147 = loc("forward":4294967295:4367) +#loc148 = loc("forward":4294967295:4368) +#loc149 = loc("forward":4294967295:4369) +#loc150 = loc("forward":4294967295:4370) +#loc151 = loc("forward":4294967295:4371) +#loc152 = loc("forward":4294967295:4372) +#loc153 = loc("forward":4294967295:4374) +#loc154 = loc("forward":4294967295:4375) +#loc155 = loc("forward":4294967295:4376) +#loc156 = loc("forward":4294967295:4377) +#loc157 = loc("forward":4294967295:4379) +#loc158 = loc("forward":4294967295:4380) +#loc159 = loc("forward":4294967295:4381) +#loc160 = loc("forward":4294967295:4383) +#loc161 = loc("forward":4294967295:4384) +#loc162 = loc("forward":4294967295:4386) +#loc163 = loc("forward":4294967295:4387) +#loc164 = loc("forward":4294967295:4389) +#loc165 = loc("forward":4294967295:4390) +#loc166 = loc("forward":4294967295:4391) +#loc167 = loc("forward":4294967295:4392) +#loc168 = loc("forward":4294967295:4393) +#loc169 = loc("forward":4294967295:4394) +#loc170 = loc("forward":4294967295:4396) +#loc171 = loc("forward":4294967295:4397) +#loc172 = loc("forward":4294967295:4398) +#loc173 = loc("forward":4294967295:4400) +#loc174 = loc("forward":4294967295:4401) +#loc175 = loc("forward":4294967295:4402) +#loc176 = loc("forward":4294967295:4403) +#loc177 = loc("forward":4294967295:4405) +#loc178 = loc("forward":4294967295:4406) +#loc179 = loc("forward":4294967295:4407) +#loc180 = loc("forward":4294967295:4409) +#loc181 = loc("forward":4294967295:4411) +#loc182 = loc("forward":4294967295:4412) +#loc183 = loc("forward":4294967295:4413) +#loc184 = loc("forward":4294967295:4414) +#loc185 = loc("forward":4294967295:4415) +#loc186 = loc("forward":4294967295:4416) +#loc187 = loc("forward":4294967295:4417) +#loc188 = loc("forward":4294967295:4418) +#loc189 = loc("forward":4294967295:4419) +#loc190 = loc("forward":4294967295:4421) +#loc191 = loc("forward":4294967295:4422) +#loc192 = loc("forward":4294967295:4423) +#loc193 = loc("forward":4294967295:4424) +#loc194 = loc("forward":4294967295:4426) +#loc195 = loc("forward":4294967295:4427) +#loc196 = loc("forward":4294967295:4428) +#loc197 = loc("forward":4294967295:4430) +#loc198 = loc("forward":4294967295:4432) +#loc199 = loc("forward":4294967295:4433) +#loc200 = loc("forward":4294967295:4434) +#loc201 = loc("forward":4294967295:4435) +#loc202 = loc("forward":4294967295:4436) +#loc203 = loc("forward":4294967295:4437) +#loc204 = loc("forward":4294967295:4438) +#loc205 = loc("forward":4294967295:4439) +#loc206 = loc("forward":4294967295:4440) +#loc207 = loc("forward":4294967295:4441) +#loc208 = loc("forward":4294967295:4443) +#loc209 = loc("forward":4294967295:4445) +#loc210 = loc("forward":4294967295:4446) +#loc211 = loc("forward":4294967295:4447) +#loc212 = loc("forward":4294967295:4449) +#loc213 = loc("forward":4294967295:4450) +#loc214 = loc("forward":4294967295:4451) +#loc215 = loc("forward":4294967295:4452) +#loc216 = loc("forward":4294967295:4453) +#loc217 = loc("forward":4294967295:4454) +#loc218 = loc("forward":4294967295:4455) +#loc219 = loc("forward":4294967295:4456) +#loc220 = loc("forward":4294967295:4457) +#loc221 = loc("forward":4294967295:4458) +#loc222 = loc("forward":4294967295:4460) +#loc223 = loc("forward":4294967295:4461) +#loc224 = loc("forward":4294967295:4462) +#loc225 = loc("forward":4294967295:4464) +#loc226 = loc("forward":4294967295:4465) +#loc227 = loc("forward":4294967295:4467) +#loc228 = loc("forward":4294967295:4468) +#loc229 = loc("forward":4294967295:4469) +#loc230 = loc("forward":4294967295:4470) +#loc231 = loc("forward":4294967295:4471) +#loc232 = loc("forward":4294967295:4472) +#loc233 = loc("forward":4294967295:4474) +#loc234 = loc("forward":4294967295:4475) +#loc235 = loc("forward":4294967295:4476) +#loc236 = loc("forward":4294967295:4477) +#loc237 = loc("forward":4294967295:4479) +#loc238 = loc("forward":4294967295:4480) +#loc239 = loc("forward":4294967295:4481) +#loc240 = loc("forward":4294967295:4483) +#loc241 = loc("forward":4294967295:4484) +#loc242 = loc("forward":4294967295:4486) +#loc243 = loc("forward":4294967295:4487) +#loc244 = loc("forward":4294967295:4489) +#loc245 = loc("forward":4294967295:4490) +#loc246 = loc("forward":4294967295:4491) +#loc247 = loc("forward":4294967295:4492) +#loc248 = loc("forward":4294967295:4493) +#loc249 = loc("forward":4294967295:4494) +#loc250 = loc("forward":4294967295:4496) +#loc251 = loc("forward":4294967295:4497) +#loc252 = loc("forward":4294967295:4498) +#loc253 = loc("forward":4294967295:4500) +#loc254 = loc("forward":4294967295:4501) +#loc255 = loc("forward":4294967295:4502) +#loc256 = loc("forward":4294967295:4503) +#loc257 = loc("forward":4294967295:4505) +#loc258 = loc("forward":4294967295:4506) +#loc259 = loc("forward":4294967295:4507) +#loc260 = loc("forward":4294967295:4509) +#loc261 = loc("forward":4294967295:4511) +#loc262 = loc("forward":4294967295:4512) +#loc263 = loc("forward":4294967295:4513) +#loc264 = loc("forward":4294967295:4514) +#loc265 = loc("forward":4294967295:4515) +#loc266 = loc("forward":4294967295:4516) +#loc267 = loc("forward":4294967295:4517) +#loc268 = loc("forward":4294967295:4518) +#loc269 = loc("forward":4294967295:4519) +#loc270 = loc("forward":4294967295:4521) +#loc271 = loc("forward":4294967295:4522) +#loc272 = loc("forward":4294967295:4523) +#loc273 = loc("forward":4294967295:4524) +#loc274 = loc("forward":4294967295:4526) +#loc275 = loc("forward":4294967295:4527) +#loc276 = loc("forward":4294967295:4528) +#loc277 = loc("forward":4294967295:4530) +#loc278 = loc("forward":4294967295:4532) +#loc279 = loc("forward":4294967295:4533) +#loc280 = loc("forward":4294967295:4534) +#loc281 = loc("forward":4294967295:4535) +#loc282 = loc("forward":4294967295:4536) +#loc283 = loc("forward":4294967295:4537) +#loc284 = loc("forward":4294967295:4538) +#loc285 = loc("forward":4294967295:4539) +#loc286 = loc("forward":4294967295:4540) +#loc287 = loc("forward":4294967295:4541) +#loc288 = loc("forward":4294967295:4543) +#loc289 = loc("forward":4294967295:4545) +#loc290 = loc("forward":4294967295:4546) +#loc291 = loc("forward":4294967295:4547) +#loc292 = loc("forward":4294967295:4549) +#loc293 = loc("forward":4294967295:4550) +#loc294 = loc("forward":4294967295:4551) +#loc295 = loc("forward":4294967295:4552) +#loc296 = loc("forward":4294967295:4553) +#loc297 = loc("forward":4294967295:4554) +#loc298 = loc("forward":4294967295:4555) +#loc299 = loc("forward":4294967295:4556) +#loc300 = loc("forward":4294967295:4557) +#loc301 = loc("forward":4294967295:4558) +#loc302 = loc("forward":4294967295:4560) +#loc303 = loc("forward":4294967295:4561) +#loc304 = loc("forward":4294967295:4562) +#loc305 = loc("forward":4294967295:4564) +#loc306 = loc("forward":4294967295:4565) +#loc307 = loc("forward":4294967295:4567) +#loc308 = loc("forward":4294967295:4568) +#loc309 = loc("forward":4294967295:4569) +#loc310 = loc("forward":4294967295:4570) +#loc311 = loc("forward":4294967295:4571) +#loc312 = loc("forward":4294967295:4572) +#loc313 = loc("forward":4294967295:4574) +#loc314 = loc("forward":4294967295:4575) +#loc315 = loc("forward":4294967295:4576) +#loc316 = loc("forward":4294967295:4577) +#loc317 = loc("forward":4294967295:4579) +#loc318 = loc("forward":4294967295:4580) +#loc319 = loc("forward":4294967295:4581) +#loc320 = loc("forward":4294967295:4583) +#loc321 = loc("forward":4294967295:4584) +#loc322 = loc("forward":4294967295:4586) +#loc323 = loc("forward":4294967295:4587) +#loc324 = loc("forward":4294967295:4589) +#loc325 = loc("forward":4294967295:4590) +#loc326 = loc("forward":4294967295:4591) +#loc327 = loc("forward":4294967295:4592) +#loc328 = loc("forward":4294967295:4593) +#loc329 = loc("forward":4294967295:4594) +#loc330 = loc("forward":4294967295:4596) +#loc331 = loc("forward":4294967295:4597) +#loc332 = loc("forward":4294967295:4598) +#loc333 = loc("forward":4294967295:4600) +#loc334 = loc("forward":4294967295:4601) +#loc335 = loc("forward":4294967295:4602) +#loc336 = loc("forward":4294967295:4603) +#loc337 = loc("forward":4294967295:4605) +#loc338 = loc("forward":4294967295:4606) +#loc339 = loc("forward":4294967295:4607) +#loc340 = loc("forward":4294967295:4609) +#loc341 = loc("forward":4294967295:4611) +#loc342 = loc("forward":4294967295:4612) +#loc343 = loc("forward":4294967295:4613) +#loc344 = loc("forward":4294967295:4614) +#loc345 = loc("forward":4294967295:4615) +#loc346 = loc("forward":4294967295:4616) +#loc347 = loc("forward":4294967295:4617) +#loc348 = loc("forward":4294967295:4618) +#loc349 = loc("forward":4294967295:4619) +#loc350 = loc("forward":4294967295:4621) +#loc351 = loc("forward":4294967295:4622) +#loc352 = loc("forward":4294967295:4623) +#loc353 = loc("forward":4294967295:4624) +#loc354 = loc("forward":4294967295:4626) +#loc355 = loc("forward":4294967295:4627) +#loc356 = loc("forward":4294967295:4628) +#loc357 = loc("forward":4294967295:4630) +#loc358 = loc("forward":4294967295:4632) +#loc359 = loc("forward":4294967295:4633) +#loc360 = loc("forward":4294967295:4634) +#loc361 = loc("forward":4294967295:4635) +#loc362 = loc("forward":4294967295:4636) +#loc363 = loc("forward":4294967295:4637) +#loc364 = loc("forward":4294967295:4638) +#loc365 = loc("forward":4294967295:4639) +#loc366 = loc("forward":4294967295:4640) +#loc367 = loc("forward":4294967295:4641) +#loc368 = loc("forward":4294967295:4643) +#loc369 = loc("forward":4294967295:4645) +#loc370 = loc("forward":4294967295:4646) +#loc371 = loc("forward":4294967295:4647) +#loc372 = loc("forward":4294967295:4649) +#loc373 = loc("forward":4294967295:4650) +#loc374 = loc("forward":4294967295:4651) +#loc375 = loc("forward":4294967295:4652) +#loc376 = loc("forward":4294967295:4653) +#loc377 = loc("forward":4294967295:4654) +#loc378 = loc("forward":4294967295:4655) +#loc379 = loc("forward":4294967295:4656) +#loc380 = loc("forward":4294967295:4657) +#loc381 = loc("forward":4294967295:4658) +#loc382 = loc("forward":4294967295:4660) +#loc383 = loc("forward":4294967295:4661) +#loc384 = loc("forward":4294967295:4662) +#loc385 = loc("forward":4294967295:4664) +#loc386 = loc("forward":4294967295:4665) +#loc387 = loc("forward":4294967295:4667) +#loc388 = loc("forward":4294967295:4668) +#loc389 = loc("forward":4294967295:4669) +#loc390 = loc("forward":4294967295:4670) +#loc391 = loc("forward":4294967295:4671) +#loc392 = loc("forward":4294967295:4672) +#loc393 = loc("forward":4294967295:4674) +#loc394 = loc("forward":4294967295:4675) +#loc395 = loc("forward":4294967295:4676) +#loc396 = loc("forward":4294967295:4677) +#loc397 = loc("forward":4294967295:4679) +#loc398 = loc("forward":4294967295:4680) +#loc399 = loc("forward":4294967295:4681) +#loc400 = loc("forward":4294967295:4683) +#loc401 = loc("forward":4294967295:4684) +#loc402 = loc("forward":4294967295:4686) +#loc403 = loc("forward":4294967295:4687) +#loc404 = loc("forward":4294967295:4689) +#loc405 = loc("forward":4294967295:4690) +#loc406 = loc("forward":4294967295:4691) +#loc407 = loc("forward":4294967295:4692) +#loc408 = loc("forward":4294967295:4693) +#loc409 = loc("forward":4294967295:4694) +#loc410 = loc("forward":4294967295:4696) +#loc411 = loc("forward":4294967295:4697) +#loc412 = loc("forward":4294967295:4698) +#loc413 = loc("forward":4294967295:4700) +#loc414 = loc("forward":4294967295:4701) +#loc415 = loc("forward":4294967295:4702) +#loc416 = loc("forward":4294967295:4703) +#loc417 = loc("forward":4294967295:4705) +#loc418 = loc("forward":4294967295:4706) +#loc419 = loc("forward":4294967295:4707) +#loc420 = loc("forward":4294967295:4709) +#loc421 = loc("forward":4294967295:4711) +#loc422 = loc("forward":4294967295:4712) +#loc423 = loc("forward":4294967295:4713) +#loc424 = loc("forward":4294967295:4714) +#loc425 = loc("forward":4294967295:4715) +#loc426 = loc("forward":4294967295:4716) +#loc427 = loc("forward":4294967295:4717) +#loc428 = loc("forward":4294967295:4718) +#loc429 = loc("forward":4294967295:4719) +#loc430 = loc("forward":4294967295:4721) +#loc431 = loc("forward":4294967295:4722) +#loc432 = loc("forward":4294967295:4723) +#loc433 = loc("forward":4294967295:4724) +#loc434 = loc("forward":4294967295:4726) +#loc435 = loc("forward":4294967295:4727) +#loc436 = loc("forward":4294967295:4728) +#loc437 = loc("forward":4294967295:4730) +#loc438 = loc("forward":4294967295:4732) +#loc439 = loc("forward":4294967295:4733) +#loc440 = loc("forward":4294967295:4734) +#loc441 = loc("forward":4294967295:4735) +#loc442 = loc("forward":4294967295:4736) +#loc443 = loc("forward":4294967295:4737) +#loc444 = loc("forward":4294967295:4738) +#loc445 = loc("forward":4294967295:4739) +#loc446 = loc("forward":4294967295:4740) +#loc447 = loc("forward":4294967295:4741) +#loc448 = loc("forward":4294967295:4743) +#loc449 = loc("forward":4294967295:4745) +#loc450 = loc("forward":4294967295:4746) +#loc451 = loc("forward":4294967295:4747) +#loc452 = loc("forward":4294967295:4749) +#loc453 = loc("forward":4294967295:4750) +#loc454 = loc("forward":4294967295:4751) +#loc455 = loc("forward":4294967295:4752) +#loc456 = loc("forward":4294967295:4753) +#loc457 = loc("forward":4294967295:4754) +#loc458 = loc("forward":4294967295:4755) +#loc459 = loc("forward":4294967295:4756) +#loc460 = loc("forward":4294967295:4757) +#loc461 = loc("forward":4294967295:4758) +#loc462 = loc("forward":4294967295:4760) +#loc463 = loc("forward":4294967295:4761) +#loc464 = loc("forward":4294967295:4762) +#loc465 = loc("forward":4294967295:4764) +#loc466 = loc("forward":4294967295:4765) +#loc467 = loc("forward":4294967295:4767) +#loc468 = loc("forward":4294967295:4768) +#loc469 = loc("forward":4294967295:4769) +#loc470 = loc("forward":4294967295:4770) +#loc471 = loc("forward":4294967295:4771) +#loc472 = loc("forward":4294967295:4772) +#loc473 = loc("forward":4294967295:4774) +#loc474 = loc("forward":4294967295:4775) +#loc475 = loc("forward":4294967295:4776) +#loc476 = loc("forward":4294967295:4777) +#loc477 = loc("forward":4294967295:4779) +#loc478 = loc("forward":4294967295:4780) +#loc479 = loc("forward":4294967295:4781) +#loc480 = loc("forward":4294967295:4783) +#loc481 = loc("forward":4294967295:4784) +#loc482 = loc("forward":4294967295:4786) +#loc483 = loc("forward":4294967295:4787) +#loc484 = loc("forward":4294967295:4789) +#loc485 = loc("forward":4294967295:4790) +#loc486 = loc("forward":4294967295:4791) +#loc487 = loc("forward":4294967295:4792) +#loc488 = loc("forward":4294967295:4793) +#loc489 = loc("forward":4294967295:4794) +#loc490 = loc("forward":4294967295:4796) +#loc491 = loc("forward":4294967295:4797) +#loc492 = loc("forward":4294967295:4798) +#loc493 = loc("forward":4294967295:4800) +#loc494 = loc("forward":4294967295:4801) +#loc495 = loc("forward":4294967295:4802) +#loc496 = loc("forward":4294967295:4803) +#loc497 = loc("forward":4294967295:4805) +#loc498 = loc("forward":4294967295:4806) +#loc499 = loc("forward":4294967295:4807) +#loc500 = loc("forward":4294967295:4809) +#loc501 = loc("forward":4294967295:4811) +#loc502 = loc("forward":4294967295:4812) +#loc503 = loc("forward":4294967295:4813) +#loc504 = loc("forward":4294967295:4814) +#loc505 = loc("forward":4294967295:4815) +#loc506 = loc("forward":4294967295:4816) +#loc507 = loc("forward":4294967295:4817) +#loc508 = loc("forward":4294967295:4818) +#loc509 = loc("forward":4294967295:4819) +#loc510 = loc("forward":4294967295:4821) +#loc511 = loc("forward":4294967295:4822) +#loc512 = loc("forward":4294967295:4823) +#loc513 = loc("forward":4294967295:4824) +#loc514 = loc("forward":4294967295:4826) +#loc515 = loc("forward":4294967295:4827) +#loc516 = loc("forward":4294967295:4828) +#loc517 = loc("forward":4294967295:4830) +#loc518 = loc("forward":4294967295:4832) +#loc519 = loc("forward":4294967295:4833) +#loc520 = loc("forward":4294967295:4834) +#loc521 = loc("forward":4294967295:4835) +#loc522 = loc("forward":4294967295:4836) +#loc523 = loc("forward":4294967295:4837) +#loc524 = loc("forward":4294967295:4838) +#loc525 = loc("forward":4294967295:4839) +#loc526 = loc("forward":4294967295:4840) +#loc527 = loc("forward":4294967295:4841) +#loc528 = loc("forward":4294967295:4843) +#loc529 = loc("forward":4294967295:4845) +#loc530 = loc("forward":4294967295:4846) +#loc531 = loc("forward":4294967295:4847) +#loc532 = loc("forward":4294967295:4849) +#loc533 = loc("forward":4294967295:4850) +#loc534 = loc("forward":4294967295:4851) +#loc535 = loc("forward":4294967295:4852) +#loc536 = loc("forward":4294967295:4853) +#loc537 = loc("forward":4294967295:4854) +#loc538 = loc("forward":4294967295:4855) +#loc539 = loc("forward":4294967295:4856) +#loc540 = loc("forward":4294967295:4857) +#loc541 = loc("forward":4294967295:4858) +#loc542 = loc("forward":4294967295:4860) +#loc543 = loc("forward":4294967295:4861) +#loc544 = loc("forward":4294967295:4862) +#loc545 = loc("forward":4294967295:4864) +#loc546 = loc("forward":4294967295:4865) +#loc547 = loc("forward":4294967295:4867) +#loc548 = loc("forward":4294967295:4868) +#loc549 = loc("forward":4294967295:4869) +#loc550 = loc("forward":4294967295:4870) +#loc551 = loc("forward":4294967295:4871) +#loc552 = loc("forward":4294967295:4872) +#loc553 = loc("forward":4294967295:4874) +#loc554 = loc("forward":4294967295:4875) +#loc555 = loc("forward":4294967295:4876) +#loc556 = loc("forward":4294967295:4877) +#loc557 = loc("forward":4294967295:4879) +#loc558 = loc("forward":4294967295:4880) +#loc559 = loc("forward":4294967295:4881) +#loc560 = loc("forward":4294967295:4883) +#loc561 = loc("forward":4294967295:4884) +#loc562 = loc("forward":4294967295:4886) +#loc563 = loc("forward":4294967295:4887) +#loc564 = loc("forward":4294967295:4889) +#loc565 = loc("forward":4294967295:4890) +#loc566 = loc("forward":4294967295:4891) +#loc567 = loc("forward":4294967295:4892) +#loc568 = loc("forward":4294967295:4893) +#loc569 = loc("forward":4294967295:4894) +#loc570 = loc("forward":4294967295:4896) +#loc571 = loc("forward":4294967295:4897) +#loc572 = loc("forward":4294967295:4898) +#loc573 = loc("forward":4294967295:4900) +#loc574 = loc("forward":4294967295:4901) +#loc575 = loc("forward":4294967295:4902) +#loc576 = loc("forward":4294967295:4903) +#loc577 = loc("forward":4294967295:4905) +#loc578 = loc("forward":4294967295:4906) +#loc579 = loc("forward":4294967295:4907) +#loc580 = loc("forward":4294967295:4909) +#loc581 = loc("forward":4294967295:4911) +#loc582 = loc("forward":4294967295:4912) +#loc583 = loc("forward":4294967295:4913) +#loc584 = loc("forward":4294967295:4914) +#loc585 = loc("forward":4294967295:4915) +#loc586 = loc("forward":4294967295:4916) +#loc587 = loc("forward":4294967295:4917) +#loc588 = loc("forward":4294967295:4918) +#loc589 = loc("forward":4294967295:4919) +#loc590 = loc("forward":4294967295:4921) +#loc591 = loc("forward":4294967295:4922) +#loc592 = loc("forward":4294967295:4923) +#loc593 = loc("forward":4294967295:4924) +#loc594 = loc("forward":4294967295:4926) +#loc595 = loc("forward":4294967295:4927) +#loc596 = loc("forward":4294967295:4928) +#loc597 = loc("forward":4294967295:4930) +#loc598 = loc("forward":4294967295:4932) +#loc599 = loc("forward":4294967295:4933) +#loc600 = loc("forward":4294967295:4934) +#loc601 = loc("forward":4294967295:4935) +#loc602 = loc("forward":4294967295:4936) +#loc603 = loc("forward":4294967295:4937) +#loc604 = loc("forward":4294967295:4938) +#loc605 = loc("forward":4294967295:4939) +#loc606 = loc("forward":4294967295:4940) +#loc607 = loc("forward":4294967295:4941) +#loc608 = loc("forward":4294967295:4943) +#loc609 = loc("forward":4294967295:4945) +#loc610 = loc("forward":4294967295:4946) +#loc611 = loc("forward":4294967295:4947) +#loc612 = loc("forward":4294967295:4949) +#loc613 = loc("forward":4294967295:4950) +#loc614 = loc("forward":4294967295:4951) +#loc615 = loc("forward":4294967295:4952) +#loc616 = loc("forward":4294967295:4953) +#loc617 = loc("forward":4294967295:4954) +#loc618 = loc("forward":4294967295:4955) +#loc619 = loc("forward":4294967295:4956) +#loc620 = loc("forward":4294967295:4957) +#loc621 = loc("forward":4294967295:4958) +#loc622 = loc("forward":4294967295:4960) +#loc623 = loc("forward":4294967295:4961) +#loc624 = loc("forward":4294967295:4962) +#loc625 = loc("forward":4294967295:4964) +#loc626 = loc("forward":4294967295:4965) +#loc627 = loc("forward":4294967295:4967) +#loc628 = loc("forward":4294967295:4968) +#loc629 = loc("forward":4294967295:4969) +#loc630 = loc("forward":4294967295:4970) +#loc631 = loc("forward":4294967295:4971) +#loc632 = loc("forward":4294967295:4972) +#loc633 = loc("forward":4294967295:4974) +#loc634 = loc("forward":4294967295:4975) +#loc635 = loc("forward":4294967295:4976) +#loc636 = loc("forward":4294967295:4977) +#loc637 = loc("forward":4294967295:4979) +#loc638 = loc("forward":4294967295:4980) +#loc639 = loc("forward":4294967295:4981) +#loc640 = loc("forward":4294967295:4983) +#loc641 = loc("forward":4294967295:4984) +#loc642 = loc("forward":4294967295:4986) +#loc643 = loc("forward":4294967295:4987) +#loc644 = loc("forward":4294967295:4989) +#loc645 = loc("forward":4294967295:4990) +#loc646 = loc("forward":4294967295:4991) +#loc647 = loc("forward":4294967295:4992) +#loc648 = loc("forward":4294967295:4993) +#loc649 = loc("forward":4294967295:4994) +#loc650 = loc("forward":4294967295:4996) +#loc651 = loc("forward":4294967295:4997) +#loc652 = loc("forward":4294967295:4998) +#loc653 = loc("forward":4294967295:5000) +#loc654 = loc("forward":4294967295:5001) +#loc655 = loc("forward":4294967295:5002) +#loc656 = loc("forward":4294967295:5003) +#loc657 = loc("forward":4294967295:5005) +#loc658 = loc("forward":4294967295:5006) +#loc659 = loc("forward":4294967295:5007) +#loc660 = loc("forward":4294967295:5009) +#loc661 = loc("forward":4294967295:5011) +#loc662 = loc("forward":4294967295:5012) +#loc663 = loc("forward":4294967295:5013) +#loc664 = loc("forward":4294967295:5014) +#loc665 = loc("forward":4294967295:5015) +#loc666 = loc("forward":4294967295:5016) +#loc667 = loc("forward":4294967295:5017) +#loc668 = loc("forward":4294967295:5018) +#loc669 = loc("forward":4294967295:5019) +#loc670 = loc("forward":4294967295:5021) +#loc671 = loc("forward":4294967295:5022) +#loc672 = loc("forward":4294967295:5023) +#loc673 = loc("forward":4294967295:5024) +#loc674 = loc("forward":4294967295:5026) +#loc675 = loc("forward":4294967295:5027) +#loc676 = loc("forward":4294967295:5028) +#loc677 = loc("forward":4294967295:5030) +#loc678 = loc("forward":4294967295:5032) +#loc679 = loc("forward":4294967295:5033) +#loc680 = loc("forward":4294967295:5034) +#loc681 = loc("forward":4294967295:5035) +#loc682 = loc("forward":4294967295:5036) +#loc683 = loc("forward":4294967295:5037) +#loc684 = loc("forward":4294967295:5038) +#loc685 = loc("forward":4294967295:5039) +#loc686 = loc("forward":4294967295:5040) +#loc687 = loc("forward":4294967295:5041) +#loc688 = loc("forward":4294967295:5043) +#loc689 = loc("forward":4294967295:5045) +#loc690 = loc("forward":4294967295:5046) +#loc691 = loc("forward":4294967295:5047) +#loc692 = loc("forward":4294967295:5049) +#loc693 = loc("forward":4294967295:5050) +#loc694 = loc("forward":4294967295:5051) +#loc695 = loc("forward":4294967295:5052) +#loc696 = loc("forward":4294967295:5053) +#loc697 = loc("forward":4294967295:5054) +#loc698 = loc("forward":4294967295:5055) +#loc699 = loc("forward":4294967295:5056) +#loc700 = loc("forward":4294967295:5057) +#loc701 = loc("forward":4294967295:5058) +#loc702 = loc("forward":4294967295:5060) +#loc703 = loc("forward":4294967295:5061) +#loc704 = loc("forward":4294967295:5062) +#loc705 = loc("forward":4294967295:5064) +#loc706 = loc("forward":4294967295:5065) +#loc707 = loc("forward":4294967295:5067) +#loc708 = loc("forward":4294967295:5068) +#loc709 = loc("forward":4294967295:5069) +#loc710 = loc("forward":4294967295:5070) +#loc711 = loc("forward":4294967295:5071) +#loc712 = loc("forward":4294967295:5072) +#loc713 = loc("forward":4294967295:5074) +#loc714 = loc("forward":4294967295:5075) +#loc715 = loc("forward":4294967295:5076) +#loc716 = loc("forward":4294967295:5077) +#loc717 = loc("forward":4294967295:5079) +#loc718 = loc("forward":4294967295:5080) +#loc719 = loc("forward":4294967295:5081) +#loc720 = loc("forward":4294967295:5083) +#loc721 = loc("forward":4294967295:5084) +#loc722 = loc("forward":4294967295:5086) +#loc723 = loc("forward":4294967295:5087) +#loc724 = loc("forward":4294967295:5089) +#loc725 = loc("forward":4294967295:5090) +#loc726 = loc("forward":4294967295:5091) +#loc727 = loc("forward":4294967295:5092) +#loc728 = loc("forward":4294967295:5093) +#loc729 = loc("forward":4294967295:5094) +#loc730 = loc("forward":4294967295:5096) +#loc731 = loc("forward":4294967295:5097) +#loc732 = loc("forward":4294967295:5098) +#loc733 = loc("forward":4294967295:5100) +#loc734 = loc("forward":4294967295:5101) +#loc735 = loc("forward":4294967295:5102) +#loc736 = loc("forward":4294967295:5103) +#loc737 = loc("forward":4294967295:5105) +#loc738 = loc("forward":4294967295:5106) +#loc739 = loc("forward":4294967295:5107) +#loc740 = loc("forward":4294967295:5109) +#loc741 = loc("forward":4294967295:5111) +#loc742 = loc("forward":4294967295:5112) +#loc743 = loc("forward":4294967295:5113) +#loc744 = loc("forward":4294967295:5114) +#loc745 = loc("forward":4294967295:5115) +#loc746 = loc("forward":4294967295:5116) +#loc747 = loc("forward":4294967295:5117) +#loc748 = loc("forward":4294967295:5118) +#loc749 = loc("forward":4294967295:5119) +#loc750 = loc("forward":4294967295:5121) +#loc751 = loc("forward":4294967295:5122) +#loc752 = loc("forward":4294967295:5123) +#loc753 = loc("forward":4294967295:5124) +#loc754 = loc("forward":4294967295:5126) +#loc755 = loc("forward":4294967295:5127) +#loc756 = loc("forward":4294967295:5128) +#loc757 = loc("forward":4294967295:5130) +#loc758 = loc("forward":4294967295:5132) +#loc759 = loc("forward":4294967295:5133) +#loc760 = loc("forward":4294967295:5134) +#loc761 = loc("forward":4294967295:5135) +#loc762 = loc("forward":4294967295:5136) +#loc763 = loc("forward":4294967295:5137) +#loc764 = loc("forward":4294967295:5138) +#loc765 = loc("forward":4294967295:5139) +#loc766 = loc("forward":4294967295:5140) +#loc767 = loc("forward":4294967295:5141) +#loc768 = loc("forward":4294967295:5143) +#loc769 = loc("forward":4294967295:5145) +#loc770 = loc("forward":4294967295:5146) +#loc771 = loc("forward":4294967295:5147) +#loc772 = loc("forward":4294967295:5149) +#loc773 = loc("forward":4294967295:5150) +#loc774 = loc("forward":4294967295:5151) +#loc775 = loc("forward":4294967295:5152) +#loc776 = loc("forward":4294967295:5153) +#loc777 = loc("forward":4294967295:5154) +#loc778 = loc("forward":4294967295:5155) +#loc779 = loc("forward":4294967295:5156) +#loc780 = loc("forward":4294967295:5157) +#loc781 = loc("forward":4294967295:5158) +#loc782 = loc("forward":4294967295:5160) +#loc783 = loc("forward":4294967295:5161) +#loc784 = loc("forward":4294967295:5162) +#loc785 = loc("forward":4294967295:5164) +#loc786 = loc("forward":4294967295:5165) +#loc787 = loc("forward":4294967295:5167) +#loc788 = loc("forward":4294967295:5168) +#loc789 = loc("forward":4294967295:5169) +#loc790 = loc("forward":4294967295:5170) +#loc791 = loc("forward":4294967295:5171) +#loc792 = loc("forward":4294967295:5172) +#loc793 = loc("forward":4294967295:5174) +#loc794 = loc("forward":4294967295:5175) +#loc795 = loc("forward":4294967295:5176) +#loc796 = loc("forward":4294967295:5177) +#loc797 = loc("forward":4294967295:5179) +#loc798 = loc("forward":4294967295:5180) +#loc799 = loc("forward":4294967295:5181) +#loc800 = loc("forward":4294967295:5183) +#loc801 = loc("forward":4294967295:5184) +#loc802 = loc("forward":4294967295:5186) +#loc803 = loc("forward":4294967295:5187) +#loc804 = loc("forward":4294967295:5189) +#loc805 = loc("forward":4294967295:5190) +#loc806 = loc("forward":4294967295:5191) +#loc807 = loc("forward":4294967295:5192) +#loc808 = loc("forward":4294967295:5193) +#loc809 = loc("forward":4294967295:5194) +#loc810 = loc("forward":4294967295:5196) +#loc811 = loc("forward":4294967295:5197) +#loc812 = loc("forward":4294967295:5198) +#loc813 = loc("forward":4294967295:5200) +#loc814 = loc("forward":4294967295:5201) +#loc815 = loc("forward":4294967295:5202) +#loc816 = loc("forward":4294967295:5203) +#loc817 = loc("forward":4294967295:5205) +#loc818 = loc("forward":4294967295:5206) +#loc819 = loc("forward":4294967295:5207) +#loc820 = loc("forward":4294967295:5209) +#loc821 = loc("forward":4294967295:5211) +#loc822 = loc("forward":4294967295:5212) +#loc823 = loc("forward":4294967295:5213) +#loc824 = loc("forward":4294967295:5214) +#loc825 = loc("forward":4294967295:5215) +#loc826 = loc("forward":4294967295:5216) +#loc827 = loc("forward":4294967295:5217) +#loc828 = loc("forward":4294967295:5218) +#loc829 = loc("forward":4294967295:5219) +#loc830 = loc("forward":4294967295:5221) +#loc831 = loc("forward":4294967295:5222) +#loc832 = loc("forward":4294967295:5223) +#loc833 = loc("forward":4294967295:5224) +#loc834 = loc("forward":4294967295:5226) +#loc835 = loc("forward":4294967295:5227) +#loc836 = loc("forward":4294967295:5228) +#loc837 = loc("forward":4294967295:5230) +#loc838 = loc("forward":4294967295:5232) +#loc839 = loc("forward":4294967295:5233) +#loc840 = loc("forward":4294967295:5234) +#loc841 = loc("forward":4294967295:5235) +#loc842 = loc("forward":4294967295:5236) +#loc843 = loc("forward":4294967295:5237) +#loc844 = loc("forward":4294967295:5238) +#loc845 = loc("forward":4294967295:5239) +#loc846 = loc("forward":4294967295:5240) +#loc847 = loc("forward":4294967295:5241) +#loc848 = loc("forward":4294967295:5243) +#loc849 = loc("forward":4294967295:5245) +#loc850 = loc("forward":4294967295:5246) +#loc851 = loc("forward":4294967295:5247) +#loc852 = loc("forward":4294967295:5249) +#loc853 = loc("forward":4294967295:5250) +#loc854 = loc("forward":4294967295:5251) +#loc855 = loc("forward":4294967295:5252) +#loc856 = loc("forward":4294967295:5253) +#loc857 = loc("forward":4294967295:5254) +#loc858 = loc("forward":4294967295:5255) +#loc859 = loc("forward":4294967295:5256) +#loc860 = loc("forward":4294967295:5257) +#loc861 = loc("forward":4294967295:5258) +#loc862 = loc("forward":4294967295:5260) +#loc863 = loc("forward":4294967295:5261) +#loc864 = loc("forward":4294967295:5262) +#loc865 = loc("forward":4294967295:5264) +#loc866 = loc("forward":4294967295:5265) +#loc867 = loc("forward":4294967295:5267) +#loc868 = loc("forward":4294967295:5268) +#loc869 = loc("forward":4294967295:5269) +#loc870 = loc("forward":4294967295:5270) +#loc871 = loc("forward":4294967295:5271) +#loc872 = loc("forward":4294967295:5272) +#loc873 = loc("forward":4294967295:5274) +#loc874 = loc("forward":4294967295:5275) +#loc875 = loc("forward":4294967295:5276) +#loc876 = loc("forward":4294967295:5277) +#loc877 = loc("forward":4294967295:5279) +#loc878 = loc("forward":4294967295:5280) +#loc879 = loc("forward":4294967295:5281) +#loc880 = loc("forward":4294967295:5283) +#loc881 = loc("forward":4294967295:5284) +#loc882 = loc("forward":4294967295:5286) +#loc883 = loc("forward":4294967295:5287) +#loc884 = loc("forward":4294967295:5289) +#loc885 = loc("forward":4294967295:5290) +#loc886 = loc("forward":4294967295:5291) +#loc887 = loc("forward":4294967295:5292) +#loc888 = loc("forward":4294967295:5293) +#loc889 = loc("forward":4294967295:5294) +#loc890 = loc("forward":4294967295:5296) +#loc891 = loc("forward":4294967295:5297) +#loc892 = loc("forward":4294967295:5298) +#loc893 = loc("forward":4294967295:5300) +#loc894 = loc("forward":4294967295:5301) +#loc895 = loc("forward":4294967295:5302) +#loc896 = loc("forward":4294967295:5303) +#loc897 = loc("forward":4294967295:5305) +#loc898 = loc("forward":4294967295:5306) +#loc899 = loc("forward":4294967295:5307) +#loc900 = loc("forward":4294967295:5309) +#loc901 = loc("forward":4294967295:5311) +#loc902 = loc("forward":4294967295:5312) +#loc903 = loc("forward":4294967295:5313) +#loc904 = loc("forward":4294967295:5314) +#loc905 = loc("forward":4294967295:5315) +#loc906 = loc("forward":4294967295:5316) +#loc907 = loc("forward":4294967295:5317) +#loc908 = loc("forward":4294967295:5318) +#loc909 = loc("forward":4294967295:5319) +#loc910 = loc("forward":4294967295:5321) +#loc911 = loc("forward":4294967295:5322) +#loc912 = loc("forward":4294967295:5323) +#loc913 = loc("forward":4294967295:5324) +#loc914 = loc("forward":4294967295:5326) +#loc915 = loc("forward":4294967295:5327) +#loc916 = loc("forward":4294967295:5328) +#loc917 = loc("forward":4294967295:5330) +#loc918 = loc("forward":4294967295:5332) +#loc919 = loc("forward":4294967295:5333) +#loc920 = loc("forward":4294967295:5334) +#loc921 = loc("forward":4294967295:5335) +#loc922 = loc("forward":4294967295:5336) +#loc923 = loc("forward":4294967295:5337) +#loc924 = loc("forward":4294967295:5338) +#loc925 = loc("forward":4294967295:5339) +#loc926 = loc("forward":4294967295:5340) +#loc927 = loc("forward":4294967295:5341) +#loc928 = loc("forward":4294967295:5343) +#loc929 = loc("forward":4294967295:5345) +#loc930 = loc("forward":4294967295:5346) +#loc931 = loc("forward":4294967295:5347) +#loc932 = loc("forward":4294967295:5349) +#loc933 = loc("forward":4294967295:5350) +#loc934 = loc("forward":4294967295:5351) +#loc935 = loc("forward":4294967295:5352) +#loc936 = loc("forward":4294967295:5353) +#loc937 = loc("forward":4294967295:5354) +#loc938 = loc("forward":4294967295:5355) +#loc939 = loc("forward":4294967295:5356) +#loc940 = loc("forward":4294967295:5357) +#loc941 = loc("forward":4294967295:5358) +#loc942 = loc("forward":4294967295:5360) +#loc943 = loc("forward":4294967295:5361) +#loc944 = loc("forward":4294967295:5362) +#loc945 = loc("forward":4294967295:5364) +#loc946 = loc("forward":4294967295:5365) +#loc947 = loc("forward":4294967295:5367) +#loc948 = loc("forward":4294967295:5368) +#loc949 = loc("forward":4294967295:5369) +#loc950 = loc("forward":4294967295:5370) +#loc951 = loc("forward":4294967295:5371) +#loc952 = loc("forward":4294967295:5372) +#loc953 = loc("forward":4294967295:5374) +#loc954 = loc("forward":4294967295:5375) +#loc955 = loc("forward":4294967295:5376) +#loc956 = loc("forward":4294967295:5377) +#loc957 = loc("forward":4294967295:5379) +#loc958 = loc("forward":4294967295:5380) +#loc959 = loc("forward":4294967295:5381) +#loc960 = loc("forward":4294967295:5383) +#loc961 = loc("forward":4294967295:5384) +#loc962 = loc("forward":4294967295:5386) +#loc963 = loc("forward":4294967295:5387) +#loc964 = loc("forward":4294967295:5389) +#loc965 = loc("forward":4294967295:5390) +#loc966 = loc("forward":4294967295:5391) +#loc967 = loc("forward":4294967295:5392) +#loc968 = loc("forward":4294967295:5393) +#loc969 = loc("forward":4294967295:5394) +#loc970 = loc("forward":4294967295:5396) +#loc971 = loc("forward":4294967295:5397) +#loc972 = loc("forward":4294967295:5398) +#loc973 = loc("forward":4294967295:5400) +#loc974 = loc("forward":4294967295:5401) +#loc975 = loc("forward":4294967295:5402) +#loc976 = loc("forward":4294967295:5403) +#loc977 = loc("forward":4294967295:5405) +#loc978 = loc("forward":4294967295:5406) +#loc979 = loc("forward":4294967295:5407) +#loc980 = loc("forward":4294967295:5409) +#loc981 = loc("forward":4294967295:5411) +#loc982 = loc("forward":4294967295:5412) +#loc983 = loc("forward":4294967295:5413) +#loc984 = loc("forward":4294967295:5414) +#loc985 = loc("forward":4294967295:5415) +#loc986 = loc("forward":4294967295:5416) +#loc987 = loc("forward":4294967295:5417) +#loc988 = loc("forward":4294967295:5418) +#loc989 = loc("forward":4294967295:5419) +#loc990 = loc("forward":4294967295:5421) +#loc991 = loc("forward":4294967295:5422) +#loc992 = loc("forward":4294967295:5423) +#loc993 = loc("forward":4294967295:5424) +#loc994 = loc("forward":4294967295:5426) +#loc995 = loc("forward":4294967295:5427) +#loc996 = loc("forward":4294967295:5428) +#loc997 = loc("forward":4294967295:5430) +#loc998 = loc("forward":4294967295:5432) +#loc999 = loc("forward":4294967295:5433) +#loc1000 = loc("forward":4294967295:5434) +#loc1001 = loc("forward":4294967295:5435) +#loc1002 = loc("forward":4294967295:5436) +#loc1003 = loc("forward":4294967295:5437) +#loc1004 = loc("forward":4294967295:5438) +#loc1005 = loc("forward":4294967295:5439) +#loc1006 = loc("forward":4294967295:5440) +#loc1007 = loc("forward":4294967295:5441) +#loc1008 = loc("forward":4294967295:5443) +#loc1009 = loc("forward":4294967295:5445) +#loc1010 = loc("forward":4294967295:5446) +#loc1011 = loc("forward":4294967295:5447) +#loc1012 = loc("forward":4294967295:5449) +#loc1013 = loc("forward":4294967295:5450) +#loc1014 = loc("forward":4294967295:5451) +#loc1015 = loc("forward":4294967295:5452) +#loc1016 = loc("forward":4294967295:5453) +#loc1017 = loc("forward":4294967295:5454) +#loc1018 = loc("forward":4294967295:5455) +#loc1019 = loc("forward":4294967295:5456) +#loc1020 = loc("forward":4294967295:5457) +#loc1021 = loc("forward":4294967295:5458) +#loc1022 = loc("forward":4294967295:5460) +#loc1023 = loc("forward":4294967295:5461) +#loc1024 = loc("forward":4294967295:5462) +#loc1025 = loc("forward":4294967295:5464) +#loc1026 = loc("forward":4294967295:5465) +#loc1027 = loc("forward":4294967295:5467) +#loc1028 = loc("forward":4294967295:5468) +#loc1029 = loc("forward":4294967295:5469) +#loc1030 = loc("forward":4294967295:5470) +#loc1031 = loc("forward":4294967295:5471) +#loc1032 = loc("forward":4294967295:5472) +#loc1033 = loc("forward":4294967295:5474) +#loc1034 = loc("forward":4294967295:5475) +#loc1035 = loc("forward":4294967295:5476) +#loc1036 = loc("forward":4294967295:5477) +#loc1037 = loc("forward":4294967295:5479) +#loc1038 = loc("forward":4294967295:5480) +#loc1039 = loc("forward":4294967295:5481) +#loc1040 = loc("forward":4294967295:5483) +#loc1041 = loc("forward":4294967295:5484) +#loc1042 = loc("forward":4294967295:5486) +#loc1043 = loc("forward":4294967295:5487) +#loc1044 = loc("forward":4294967295:5489) +#loc1045 = loc("forward":4294967295:5490) +#loc1046 = loc("forward":4294967295:5491) +#loc1047 = loc("forward":4294967295:5492) +#loc1048 = loc("forward":4294967295:5493) +#loc1049 = loc("forward":4294967295:5494) +#loc1050 = loc("forward":4294967295:5496) +#loc1051 = loc("forward":4294967295:5497) +#loc1052 = loc("forward":4294967295:5498) +#loc1053 = loc("forward":4294967295:5500) +#loc1054 = loc("forward":4294967295:5501) +#loc1055 = loc("forward":4294967295:5502) +#loc1056 = loc("forward":4294967295:5503) +#loc1057 = loc("forward":4294967295:5505) +#loc1058 = loc("forward":4294967295:5506) +#loc1059 = loc("forward":4294967295:5507) +#loc1060 = loc("forward":4294967295:5509) +#loc1061 = loc("forward":4294967295:5511) +#loc1062 = loc("forward":4294967295:5512) +#loc1063 = loc("forward":4294967295:5513) +#loc1064 = loc("forward":4294967295:5514) +#loc1065 = loc("forward":4294967295:5515) +#loc1066 = loc("forward":4294967295:5516) +#loc1067 = loc("forward":4294967295:5517) +#loc1068 = loc("forward":4294967295:5518) +#loc1069 = loc("forward":4294967295:5519) +#loc1070 = loc("forward":4294967295:5521) +#loc1071 = loc("forward":4294967295:5522) +#loc1072 = loc("forward":4294967295:5523) +#loc1073 = loc("forward":4294967295:5524) +#loc1074 = loc("forward":4294967295:5526) +#loc1075 = loc("forward":4294967295:5527) +#loc1076 = loc("forward":4294967295:5528) +#loc1077 = loc("forward":4294967295:5530) +#loc1078 = loc("forward":4294967295:5532) +#loc1079 = loc("forward":4294967295:5533) +#loc1080 = loc("forward":4294967295:5534) +#loc1081 = loc("forward":4294967295:5535) +#loc1082 = loc("forward":4294967295:5536) +#loc1083 = loc("forward":4294967295:5537) +#loc1084 = loc("forward":4294967295:5538) +#loc1085 = loc("forward":4294967295:5539) +#loc1086 = loc("forward":4294967295:5540) +#loc1087 = loc("forward":4294967295:5541) +#loc1088 = loc("forward":4294967295:5543) +#loc1089 = loc("forward":4294967295:5545) +#loc1090 = loc("forward":4294967295:5546) +#loc1091 = loc("forward":4294967295:5547) +#loc1092 = loc("forward":4294967295:5549) +#loc1093 = loc("forward":4294967295:5550) +#loc1094 = loc("forward":4294967295:5551) +#loc1095 = loc("forward":4294967295:5552) +#loc1096 = loc("forward":4294967295:5553) +#loc1097 = loc("forward":4294967295:5554) +#loc1098 = loc("forward":4294967295:5555) +#loc1099 = loc("forward":4294967295:5556) +#loc1100 = loc("forward":4294967295:5557) +#loc1101 = loc("forward":4294967295:5558) +#loc1102 = loc("forward":4294967295:5560) +#loc1103 = loc("forward":4294967295:5561) +#loc1104 = loc("forward":4294967295:5562) +#loc1105 = loc("forward":4294967295:5564) +#loc1106 = loc("forward":4294967295:5565) +#loc1107 = loc("forward":4294967295:5567) +#loc1108 = loc("forward":4294967295:5568) +#loc1109 = loc("forward":4294967295:5569) +#loc1110 = loc("forward":4294967295:5570) +#loc1111 = loc("forward":4294967295:5571) +#loc1112 = loc("forward":4294967295:5572) +#loc1113 = loc("forward":4294967295:5574) +#loc1114 = loc("forward":4294967295:5575) +#loc1115 = loc("forward":4294967295:5576) +#loc1116 = loc("forward":4294967295:5577) +#loc1117 = loc("forward":4294967295:5579) +#loc1118 = loc("forward":4294967295:5580) +#loc1119 = loc("forward":4294967295:5581) +#loc1120 = loc("forward":4294967295:5583) +#loc1121 = loc("forward":4294967295:5584) +#loc1122 = loc("forward":4294967295:5586) +#loc1123 = loc("forward":4294967295:5587) +#loc1124 = loc("forward":4294967295:5589) +#loc1125 = loc("forward":4294967295:5590) +#loc1126 = loc("forward":4294967295:5591) +#loc1127 = loc("forward":4294967295:5592) +#loc1128 = loc("forward":4294967295:5593) +#loc1129 = loc("forward":4294967295:5594) +#loc1130 = loc("forward":4294967295:5596) +#loc1131 = loc("forward":4294967295:5597) +#loc1132 = loc("forward":4294967295:5598) +#loc1133 = loc("forward":4294967295:5600) +#loc1134 = loc("forward":4294967295:5601) +#loc1135 = loc("forward":4294967295:5602) +#loc1136 = loc("forward":4294967295:5603) +#loc1137 = loc("forward":4294967295:5605) +#loc1138 = loc("forward":4294967295:5606) +#loc1139 = loc("forward":4294967295:5607) +#loc1140 = loc("forward":4294967295:5609) +#loc1141 = loc("forward":4294967295:5611) +#loc1142 = loc("forward":4294967295:5612) +#loc1143 = loc("forward":4294967295:5613) +#loc1144 = loc("forward":4294967295:5614) +#loc1145 = loc("forward":4294967295:5615) +#loc1146 = loc("forward":4294967295:5616) +#loc1147 = loc("forward":4294967295:5617) +#loc1148 = loc("forward":4294967295:5618) +#loc1149 = loc("forward":4294967295:5619) +#loc1150 = loc("forward":4294967295:5621) +#loc1151 = loc("forward":4294967295:5622) +#loc1152 = loc("forward":4294967295:5623) +#loc1153 = loc("forward":4294967295:5624) +#loc1154 = loc("forward":4294967295:5626) +#loc1155 = loc("forward":4294967295:5627) +#loc1156 = loc("forward":4294967295:5628) +#loc1157 = loc("forward":4294967295:5630) +#loc1158 = loc("forward":4294967295:5632) +#loc1159 = loc("forward":4294967295:5633) +#loc1160 = loc("forward":4294967295:5634) +#loc1161 = loc("forward":4294967295:5635) +#loc1162 = loc("forward":4294967295:5636) +#loc1163 = loc("forward":4294967295:5637) +#loc1164 = loc("forward":4294967295:5638) +#loc1165 = loc("forward":4294967295:5639) +#loc1166 = loc("forward":4294967295:5640) +#loc1167 = loc("forward":4294967295:5641) +#loc1168 = loc("forward":4294967295:5643) +#loc1169 = loc("forward":4294967295:5645) +#loc1170 = loc("forward":4294967295:5646) +#loc1171 = loc("forward":4294967295:5647) +#loc1172 = loc("forward":4294967295:5649) +#loc1173 = loc("forward":4294967295:5650) +#loc1174 = loc("forward":4294967295:5651) +#loc1175 = loc("forward":4294967295:5652) +#loc1176 = loc("forward":4294967295:5653) +#loc1177 = loc("forward":4294967295:5654) +#loc1178 = loc("forward":4294967295:5655) +#loc1179 = loc("forward":4294967295:5656) +#loc1180 = loc("forward":4294967295:5657) +#loc1181 = loc("forward":4294967295:5658) +#loc1182 = loc("forward":4294967295:5660) +#loc1183 = loc("forward":4294967295:5661) +#loc1184 = loc("forward":4294967295:5662) +#loc1185 = loc("forward":4294967295:5664) +#loc1186 = loc("forward":4294967295:5665) +#loc1187 = loc("forward":4294967295:5667) +#loc1188 = loc("forward":4294967295:5668) +#loc1189 = loc("forward":4294967295:5669) +#loc1190 = loc("forward":4294967295:5670) +#loc1191 = loc("forward":4294967295:5671) +#loc1192 = loc("forward":4294967295:5672) +#loc1193 = loc("forward":4294967295:5674) +#loc1194 = loc("forward":4294967295:5675) +#loc1195 = loc("forward":4294967295:5676) +#loc1196 = loc("forward":4294967295:5677) +#loc1197 = loc("forward":4294967295:5679) +#loc1198 = loc("forward":4294967295:5680) +#loc1199 = loc("forward":4294967295:5681) +#loc1200 = loc("forward":4294967295:5683) +#loc1201 = loc("forward":4294967295:5684) +#loc1202 = loc("forward":4294967295:5686) +#loc1203 = loc("forward":4294967295:5687) +#loc1204 = loc("forward":4294967295:5689) +#loc1205 = loc("forward":4294967295:5690) +#loc1206 = loc("forward":4294967295:5691) +#loc1207 = loc("forward":4294967295:5692) +#loc1208 = loc("forward":4294967295:5693) +#loc1209 = loc("forward":4294967295:5694) +#loc1210 = loc("forward":4294967295:5696) +#loc1211 = loc("forward":4294967295:5697) +#loc1212 = loc("forward":4294967295:5698) +#loc1213 = loc("forward":4294967295:5700) +#loc1214 = loc("forward":4294967295:5701) +#loc1215 = loc("forward":4294967295:5702) +#loc1216 = loc("forward":4294967295:5703) +#loc1217 = loc("forward":4294967295:5705) +#loc1218 = loc("forward":4294967295:5706) +#loc1219 = loc("forward":4294967295:5707) +#loc1220 = loc("forward":4294967295:5709) +#loc1221 = loc("forward":4294967295:5711) +#loc1222 = loc("forward":4294967295:5712) +#loc1223 = loc("forward":4294967295:5713) +#loc1224 = loc("forward":4294967295:5714) +#loc1225 = loc("forward":4294967295:5715) +#loc1226 = loc("forward":4294967295:5716) +#loc1227 = loc("forward":4294967295:5717) +#loc1228 = loc("forward":4294967295:5718) +#loc1229 = loc("forward":4294967295:5719) +#loc1230 = loc("forward":4294967295:5721) +#loc1231 = loc("forward":4294967295:5722) +#loc1232 = loc("forward":4294967295:5723) +#loc1233 = loc("forward":4294967295:5724) +#loc1234 = loc("forward":4294967295:5726) +#loc1235 = loc("forward":4294967295:5727) +#loc1236 = loc("forward":4294967295:5728) +#loc1237 = loc("forward":4294967295:5730) +#loc1238 = loc("forward":4294967295:5732) +#loc1239 = loc("forward":4294967295:5733) +#loc1240 = loc("forward":4294967295:5734) +#loc1241 = loc("forward":4294967295:5735) +#loc1242 = loc("forward":4294967295:5736) +#loc1243 = loc("forward":4294967295:5737) +#loc1244 = loc("forward":4294967295:5738) +#loc1245 = loc("forward":4294967295:5739) +#loc1246 = loc("forward":4294967295:5740) +#loc1247 = loc("forward":4294967295:5741) +#loc1248 = loc("forward":4294967295:5743) +#loc1249 = loc("forward":4294967295:5745) +#loc1250 = loc("forward":4294967295:5746) +#loc1251 = loc("forward":4294967295:5747) +#loc1252 = loc("forward":4294967295:5749) +#loc1253 = loc("forward":4294967295:5750) +#loc1254 = loc("forward":4294967295:5751) +#loc1255 = loc("forward":4294967295:5752) +#loc1256 = loc("forward":4294967295:5753) +#loc1257 = loc("forward":4294967295:5754) +#loc1258 = loc("forward":4294967295:5755) +#loc1259 = loc("forward":4294967295:5756) +#loc1260 = loc("forward":4294967295:5757) +#loc1261 = loc("forward":4294967295:5758) +#loc1262 = loc("forward":4294967295:5760) +#loc1263 = loc("forward":4294967295:5761) +#loc1264 = loc("forward":4294967295:5762) +#loc1265 = loc("forward":4294967295:5764) +#loc1266 = loc("forward":4294967295:5765) +#loc1267 = loc("forward":4294967295:5767) +#loc1268 = loc("forward":4294967295:5768) +#loc1269 = loc("forward":4294967295:5769) +#loc1270 = loc("forward":4294967295:5770) +#loc1271 = loc("forward":4294967295:5771) +#loc1272 = loc("forward":4294967295:5772) +#loc1273 = loc("forward":4294967295:5774) +#loc1274 = loc("forward":4294967295:5775) +#loc1275 = loc("forward":4294967295:5776) +#loc1276 = loc("forward":4294967295:5777) +#loc1277 = loc("forward":4294967295:5779) +#loc1278 = loc("forward":4294967295:5780) +#loc1279 = loc("forward":4294967295:5781) +#loc1280 = loc("forward":4294967295:5783) +#loc1281 = loc("forward":4294967295:5784) +#loc1282 = loc("forward":4294967295:5786) +#loc1283 = loc("forward":4294967295:5787) +#loc1284 = loc("forward":4294967295:5789) +#loc1285 = loc("forward":4294967295:5790) +#loc1286 = loc("forward":4294967295:5791) +#loc1287 = loc("forward":4294967295:5792) +#loc1288 = loc("forward":4294967295:5793) +#loc1289 = loc("forward":4294967295:5794) +#loc1290 = loc("forward":4294967295:5796) +#loc1291 = loc("forward":4294967295:5797) +#loc1292 = loc("forward":4294967295:5798) +#loc1293 = loc("forward":4294967295:5800) +#loc1294 = loc("forward":4294967295:5801) +#loc1295 = loc("forward":4294967295:5802) +#loc1296 = loc("forward":4294967295:5803) +#loc1297 = loc("forward":4294967295:5805) +#loc1298 = loc("forward":4294967295:5806) +#loc1299 = loc("forward":4294967295:5807) +#loc1300 = loc("forward":4294967295:5809) +#loc1301 = loc("forward":4294967295:5811) +#loc1302 = loc("forward":4294967295:5812) +#loc1303 = loc("forward":4294967295:5813) +#loc1304 = loc("forward":4294967295:5814) +#loc1305 = loc("forward":4294967295:5815) +#loc1306 = loc("forward":4294967295:5816) +#loc1307 = loc("forward":4294967295:5817) +#loc1308 = loc("forward":4294967295:5818) +#loc1309 = loc("forward":4294967295:5819) +#loc1310 = loc("forward":4294967295:5821) +#loc1311 = loc("forward":4294967295:5822) +#loc1312 = loc("forward":4294967295:5823) +#loc1313 = loc("forward":4294967295:5824) +#loc1314 = loc("forward":4294967295:5826) +#loc1315 = loc("forward":4294967295:5827) +#loc1316 = loc("forward":4294967295:5828) +#loc1317 = loc("forward":4294967295:5830) +#loc1318 = loc("forward":4294967295:5832) +#loc1319 = loc("forward":4294967295:5833) +#loc1320 = loc("forward":4294967295:5834) +#loc1321 = loc("forward":4294967295:5835) +#loc1322 = loc("forward":4294967295:5836) +#loc1323 = loc("forward":4294967295:5837) +#loc1324 = loc("forward":4294967295:5838) +#loc1325 = loc("forward":4294967295:5839) +#loc1326 = loc("forward":4294967295:5840) +#loc1327 = loc("forward":4294967295:5841) +#loc1328 = loc("forward":4294967295:5843) +#loc1329 = loc("forward":4294967295:5845) +#loc1330 = loc("forward":4294967295:5846) +#loc1331 = loc("forward":4294967295:5847) +#loc1332 = loc("forward":4294967295:5849) +#loc1333 = loc("forward":4294967295:5850) +#loc1334 = loc("forward":4294967295:5851) +#loc1335 = loc("forward":4294967295:5852) +#loc1336 = loc("forward":4294967295:5853) +#loc1337 = loc("forward":4294967295:5854) +#loc1338 = loc("forward":4294967295:5855) +#loc1339 = loc("forward":4294967295:5856) +#loc1340 = loc("forward":4294967295:5857) +#loc1341 = loc("forward":4294967295:5858) +#loc1342 = loc("forward":4294967295:5860) +#loc1343 = loc("forward":4294967295:5861) +#loc1344 = loc("forward":4294967295:5862) +#loc1345 = loc("forward":4294967295:5864) +#loc1346 = loc("forward":4294967295:5865) +#loc1347 = loc("forward":4294967295:5867) +#loc1348 = loc("forward":4294967295:5868) +#loc1349 = loc("forward":4294967295:5869) +#loc1350 = loc("forward":4294967295:5870) +#loc1351 = loc("forward":4294967295:5871) +#loc1352 = loc("forward":4294967295:5872) +#loc1353 = loc("forward":4294967295:5874) +#loc1354 = loc("forward":4294967295:5875) +#loc1355 = loc("forward":4294967295:5876) +#loc1356 = loc("forward":4294967295:5877) +#loc1357 = loc("forward":4294967295:5879) +#loc1358 = loc("forward":4294967295:5880) +#loc1359 = loc("forward":4294967295:5881) +#loc1360 = loc("forward":4294967295:5883) +#loc1361 = loc("forward":4294967295:5884) +#loc1362 = loc("forward":4294967295:5886) +#loc1363 = loc("forward":4294967295:5887) +#loc1364 = loc("forward":4294967295:5889) +#loc1365 = loc("forward":4294967295:5890) +#loc1366 = loc("forward":4294967295:5891) +#loc1367 = loc("forward":4294967295:5892) +#loc1368 = loc("forward":4294967295:5893) +#loc1369 = loc("forward":4294967295:5894) +#loc1370 = loc("forward":4294967295:5896) +#loc1371 = loc("forward":4294967295:5897) +#loc1372 = loc("forward":4294967295:5898) +#loc1373 = loc("forward":4294967295:5900) +#loc1374 = loc("forward":4294967295:5901) +#loc1375 = loc("forward":4294967295:5902) +#loc1376 = loc("forward":4294967295:5903) +#loc1377 = loc("forward":4294967295:5905) +#loc1378 = loc("forward":4294967295:5906) +#loc1379 = loc("forward":4294967295:5907) +#loc1380 = loc("forward":4294967295:5909) +#loc1381 = loc("forward":4294967295:5911) +#loc1382 = loc("forward":4294967295:5912) +#loc1383 = loc("forward":4294967295:5913) +#loc1384 = loc("forward":4294967295:5914) +#loc1385 = loc("forward":4294967295:5915) +#loc1386 = loc("forward":4294967295:5916) +#loc1387 = loc("forward":4294967295:5917) +#loc1388 = loc("forward":4294967295:5918) +#loc1389 = loc("forward":4294967295:5919) +#loc1390 = loc("forward":4294967295:5921) +#loc1391 = loc("forward":4294967295:5922) +#loc1392 = loc("forward":4294967295:5923) +#loc1393 = loc("forward":4294967295:5924) +#loc1394 = loc("forward":4294967295:5926) +#loc1395 = loc("forward":4294967295:5927) +#loc1396 = loc("forward":4294967295:5928) +#loc1397 = loc("forward":4294967295:5930) +#loc1398 = loc("forward":4294967295:5932) +#loc1399 = loc("forward":4294967295:5933) +#loc1400 = loc("forward":4294967295:5934) +#loc1401 = loc("forward":4294967295:5935) +#loc1402 = loc("forward":4294967295:5936) +#loc1403 = loc("forward":4294967295:5937) +#loc1404 = loc("forward":4294967295:5938) +#loc1405 = loc("forward":4294967295:5939) +#loc1406 = loc("forward":4294967295:5940) +#loc1407 = loc("forward":4294967295:5941) +#loc1408 = loc("forward":4294967295:5943) +#loc1409 = loc("forward":4294967295:5945) +#loc1410 = loc("forward":4294967295:5946) +#loc1411 = loc("forward":4294967295:5947) +#loc1412 = loc("forward":4294967295:5949) +#loc1413 = loc("forward":4294967295:5950) +#loc1414 = loc("forward":4294967295:5951) +#loc1415 = loc("forward":4294967295:5952) +#loc1416 = loc("forward":4294967295:5953) +#loc1417 = loc("forward":4294967295:5954) +#loc1418 = loc("forward":4294967295:5955) +#loc1419 = loc("forward":4294967295:5956) +#loc1420 = loc("forward":4294967295:5957) +#loc1421 = loc("forward":4294967295:5958) +#loc1422 = loc("forward":4294967295:5960) +#loc1423 = loc("forward":4294967295:5961) +#loc1424 = loc("forward":4294967295:5962) +#loc1425 = loc("forward":4294967295:5964) +#loc1426 = loc("forward":4294967295:5965) +#loc1427 = loc("forward":4294967295:5967) +#loc1428 = loc("forward":4294967295:5968) +#loc1429 = loc("forward":4294967295:5969) +#loc1430 = loc("forward":4294967295:5970) +#loc1431 = loc("forward":4294967295:5971) +#loc1432 = loc("forward":4294967295:5972) +#loc1433 = loc("forward":4294967295:5974) +#loc1434 = loc("forward":4294967295:5975) +#loc1435 = loc("forward":4294967295:5976) +#loc1436 = loc("forward":4294967295:5977) +#loc1437 = loc("forward":4294967295:5979) +#loc1438 = loc("forward":4294967295:5980) +#loc1439 = loc("forward":4294967295:5981) +#loc1440 = loc("forward":4294967295:5983) +#loc1441 = loc("forward":4294967295:5984) +#loc1442 = loc("forward":4294967295:5986) +#loc1443 = loc("forward":4294967295:5987) +#loc1444 = loc("forward":4294967295:5989) +#loc1445 = loc("forward":4294967295:5990) +#loc1446 = loc("forward":4294967295:5991) +#loc1447 = loc("forward":4294967295:5992) +#loc1448 = loc("forward":4294967295:5993) +#loc1449 = loc("forward":4294967295:5994) +#loc1450 = loc("forward":4294967295:5996) +#loc1451 = loc("forward":4294967295:5997) +#loc1452 = loc("forward":4294967295:5998) +#loc1453 = loc("forward":4294967295:6000) +#loc1454 = loc("forward":4294967295:6001) +#loc1455 = loc("forward":4294967295:6002) +#loc1456 = loc("forward":4294967295:6003) +#loc1457 = loc("forward":4294967295:6005) +#loc1458 = loc("forward":4294967295:6006) +#loc1459 = loc("forward":4294967295:6007) +#loc1460 = loc("forward":4294967295:6009) +#loc1461 = loc("forward":4294967295:6011) +#loc1462 = loc("forward":4294967295:6012) +#loc1463 = loc("forward":4294967295:6013) +#loc1464 = loc("forward":4294967295:6014) +#loc1465 = loc("forward":4294967295:6015) +#loc1466 = loc("forward":4294967295:6016) +#loc1467 = loc("forward":4294967295:6017) +#loc1468 = loc("forward":4294967295:6018) +#loc1469 = loc("forward":4294967295:6019) +#loc1470 = loc("forward":4294967295:6021) +#loc1471 = loc("forward":4294967295:6022) +#loc1472 = loc("forward":4294967295:6023) +#loc1473 = loc("forward":4294967295:6024) +#loc1474 = loc("forward":4294967295:6026) +#loc1475 = loc("forward":4294967295:6027) +#loc1476 = loc("forward":4294967295:6028) +#loc1477 = loc("forward":4294967295:6030) +#loc1478 = loc("forward":4294967295:6032) +#loc1479 = loc("forward":4294967295:6033) +#loc1480 = loc("forward":4294967295:6034) +#loc1481 = loc("forward":4294967295:6035) +#loc1482 = loc("forward":4294967295:6036) +#loc1483 = loc("forward":4294967295:6037) +#loc1484 = loc("forward":4294967295:6038) +#loc1485 = loc("forward":4294967295:6039) +#loc1486 = loc("forward":4294967295:6040) +#loc1487 = loc("forward":4294967295:6041) +#loc1488 = loc("forward":4294967295:6043) +#loc1489 = loc("forward":4294967295:6045) +#loc1490 = loc("forward":4294967295:6046) +#loc1491 = loc("forward":4294967295:6047) +#loc1492 = loc("forward":4294967295:6049) +#loc1493 = loc("forward":4294967295:6050) +#loc1494 = loc("forward":4294967295:6051) +#loc1495 = loc("forward":4294967295:6052) +#loc1496 = loc("forward":4294967295:6053) +#loc1497 = loc("forward":4294967295:6054) +#loc1498 = loc("forward":4294967295:6055) +#loc1499 = loc("forward":4294967295:6056) +#loc1500 = loc("forward":4294967295:6057) +#loc1501 = loc("forward":4294967295:6058) +#loc1502 = loc("forward":4294967295:6060) +#loc1503 = loc("forward":4294967295:6061) +#loc1504 = loc("forward":4294967295:6062) +#loc1505 = loc("forward":4294967295:6064) +#loc1506 = loc("forward":4294967295:6065) +#loc1507 = loc("forward":4294967295:6067) +#loc1508 = loc("forward":4294967295:6068) +#loc1509 = loc("forward":4294967295:6069) +#loc1510 = loc("forward":4294967295:6070) +#loc1511 = loc("forward":4294967295:6071) +#loc1512 = loc("forward":4294967295:6072) +#loc1513 = loc("forward":4294967295:6074) +#loc1514 = loc("forward":4294967295:6075) +#loc1515 = loc("forward":4294967295:6076) +#loc1516 = loc("forward":4294967295:6077) +#loc1517 = loc("forward":4294967295:6079) +#loc1518 = loc("forward":4294967295:6080) +#loc1519 = loc("forward":4294967295:6081) +#loc1520 = loc("forward":4294967295:6083) +#loc1521 = loc("forward":4294967295:6084) +#loc1522 = loc("forward":4294967295:6086) +#loc1523 = loc("forward":4294967295:6087) +#loc1524 = loc("forward":4294967295:6089) +#loc1525 = loc("forward":4294967295:6090) +#loc1526 = loc("forward":4294967295:6091) +#loc1527 = loc("forward":4294967295:6092) +#loc1528 = loc("forward":4294967295:6093) +#loc1529 = loc("forward":4294967295:6094) +#loc1530 = loc("forward":4294967295:6096) +#loc1531 = loc("forward":4294967295:6097) +#loc1532 = loc("forward":4294967295:6098) +#loc1533 = loc("forward":4294967295:6100) +#loc1534 = loc("forward":4294967295:6101) +#loc1535 = loc("forward":4294967295:6102) +#loc1536 = loc("forward":4294967295:6103) +#loc1537 = loc("forward":4294967295:6105) +#loc1538 = loc("forward":4294967295:6106) +#loc1539 = loc("forward":4294967295:6107) +#loc1540 = loc("forward":4294967295:6109) +#loc1541 = loc("forward":4294967295:6111) +#loc1542 = loc("forward":4294967295:6112) +#loc1543 = loc("forward":4294967295:6113) +#loc1544 = loc("forward":4294967295:6114) +#loc1545 = loc("forward":4294967295:6115) +#loc1546 = loc("forward":4294967295:6116) +#loc1547 = loc("forward":4294967295:6117) +#loc1548 = loc("forward":4294967295:6118) +#loc1549 = loc("forward":4294967295:6119) +#loc1550 = loc("forward":4294967295:6121) +#loc1551 = loc("forward":4294967295:6122) +#loc1552 = loc("forward":4294967295:6123) +#loc1553 = loc("forward":4294967295:6124) +#loc1554 = loc("forward":4294967295:6126) +#loc1555 = loc("forward":4294967295:6127) +#loc1556 = loc("forward":4294967295:6128) +#loc1557 = loc("forward":4294967295:6130) +#loc1558 = loc("forward":4294967295:6132) +#loc1559 = loc("forward":4294967295:6133) +#loc1560 = loc("forward":4294967295:6134) +#loc1561 = loc("forward":4294967295:6135) +#loc1562 = loc("forward":4294967295:6136) +#loc1563 = loc("forward":4294967295:6137) +#loc1564 = loc("forward":4294967295:6138) +#loc1565 = loc("forward":4294967295:6139) +#loc1566 = loc("forward":4294967295:6140) +#loc1567 = loc("forward":4294967295:6141) +#loc1568 = loc("forward":4294967295:6143) +#loc1569 = loc("forward":4294967295:6145) +#loc1570 = loc("forward":4294967295:6146) +#loc1571 = loc("forward":4294967295:6147) +#loc1572 = loc("forward":4294967295:6149) +#loc1573 = loc("forward":4294967295:6150) +#loc1574 = loc("forward":4294967295:6151) +#loc1575 = loc("forward":4294967295:6152) +#loc1576 = loc("forward":4294967295:6153) +#loc1577 = loc("forward":4294967295:6154) +#loc1578 = loc("forward":4294967295:6155) +#loc1579 = loc("forward":4294967295:6156) +#loc1580 = loc("forward":4294967295:6157) +#loc1581 = loc("forward":4294967295:6158) +#loc1582 = loc("forward":4294967295:6160) +#loc1583 = loc("forward":4294967295:6161) +#loc1584 = loc("forward":4294967295:6162) +#loc1585 = loc("forward":4294967295:6164) +#loc1586 = loc("forward":4294967295:6165) +#loc1587 = loc("forward":4294967295:6167) +#loc1588 = loc("forward":4294967295:6168) +#loc1589 = loc("forward":4294967295:6169) +#loc1590 = loc("forward":4294967295:6170) +#loc1591 = loc("forward":4294967295:6171) +#loc1592 = loc("forward":4294967295:6172) +#loc1593 = loc("forward":4294967295:6174) +#loc1594 = loc("forward":4294967295:6175) +#loc1595 = loc("forward":4294967295:6176) +#loc1596 = loc("forward":4294967295:6177) +#loc1597 = loc("forward":4294967295:6179) +#loc1598 = loc("forward":4294967295:6180) +#loc1599 = loc("forward":4294967295:6181) +#loc1600 = loc("forward":4294967295:6183) +#loc1601 = loc("forward":4294967295:6184) +#loc1602 = loc("forward":4294967295:6186) +#loc1603 = loc("forward":4294967295:6187) +#loc1604 = loc("forward":4294967295:6189) +#loc1605 = loc("forward":4294967295:6190) +#loc1606 = loc("forward":4294967295:6191) +#loc1607 = loc("forward":4294967295:6192) +#loc1608 = loc("forward":4294967295:6193) +#loc1609 = loc("forward":4294967295:6194) +#loc1610 = loc("forward":4294967295:6196) +#loc1611 = loc("forward":4294967295:6197) +#loc1612 = loc("forward":4294967295:6198) +#loc1613 = loc("forward":4294967295:6200) +#loc1614 = loc("forward":4294967295:6201) +#loc1615 = loc("forward":4294967295:6202) +#loc1616 = loc("forward":4294967295:6203) +#loc1617 = loc("forward":4294967295:6205) +#loc1618 = loc("forward":4294967295:6206) +#loc1619 = loc("forward":4294967295:6207) +#loc1620 = loc("forward":4294967295:6209) +#loc1621 = loc("forward":4294967295:6211) +#loc1622 = loc("forward":4294967295:6212) +#loc1623 = loc("forward":4294967295:6213) +#loc1624 = loc("forward":4294967295:6214) +#loc1625 = loc("forward":4294967295:6215) +#loc1626 = loc("forward":4294967295:6216) +#loc1627 = loc("forward":4294967295:6217) +#loc1628 = loc("forward":4294967295:6218) +#loc1629 = loc("forward":4294967295:6219) +#loc1630 = loc("forward":4294967295:6221) +#loc1631 = loc("forward":4294967295:6222) +#loc1632 = loc("forward":4294967295:6223) +#loc1633 = loc("forward":4294967295:6224) +#loc1634 = loc("forward":4294967295:6226) +#loc1635 = loc("forward":4294967295:6227) +#loc1636 = loc("forward":4294967295:6228) +#loc1637 = loc("forward":4294967295:6230) +#loc1638 = loc("forward":4294967295:6232) +#loc1639 = loc("forward":4294967295:6233) +#loc1640 = loc("forward":4294967295:6234) +#loc1641 = loc("forward":4294967295:6235) +#loc1642 = loc("forward":4294967295:6236) +#loc1643 = loc("forward":4294967295:6237) +#loc1644 = loc("forward":4294967295:6238) +#loc1645 = loc("forward":4294967295:6239) +#loc1646 = loc("forward":4294967295:6240) +#loc1647 = loc("forward":4294967295:6241) +#loc1648 = loc("forward":4294967295:6243) +#loc1649 = loc("forward":4294967295:6245) +#loc1650 = loc("forward":4294967295:6246) +#loc1651 = loc("forward":4294967295:6247) +#loc1652 = loc("forward":4294967295:6249) +#loc1653 = loc("forward":4294967295:6250) +#loc1654 = loc("forward":4294967295:6251) +#loc1655 = loc("forward":4294967295:6252) +#loc1656 = loc("forward":4294967295:6253) +#loc1657 = loc("forward":4294967295:6254) +#loc1658 = loc("forward":4294967295:6255) +#loc1659 = loc("forward":4294967295:6256) +#loc1660 = loc("forward":4294967295:6257) +#loc1661 = loc("forward":4294967295:6258) +#loc1662 = loc("forward":4294967295:6260) +#loc1663 = loc("forward":4294967295:6261) +#loc1664 = loc("forward":4294967295:6262) +#loc1665 = loc("forward":4294967295:6264) +#loc1666 = loc("forward":4294967295:6265) +#loc1667 = loc("forward":4294967295:6267) +#loc1668 = loc("forward":4294967295:6268) +#loc1669 = loc("forward":4294967295:6269) +#loc1670 = loc("forward":4294967295:6270) +#loc1671 = loc("forward":4294967295:6271) +#loc1672 = loc("forward":4294967295:6272) +#loc1673 = loc("forward":4294967295:6274) +#loc1674 = loc("forward":4294967295:6275) +#loc1675 = loc("forward":4294967295:6276) +#loc1676 = loc("forward":4294967295:6277) +#loc1677 = loc("forward":4294967295:6279) +#loc1678 = loc("forward":4294967295:6280) +#loc1679 = loc("forward":4294967295:6281) +#loc1680 = loc("forward":4294967295:6283) +#loc1681 = loc("forward":4294967295:6284) +#loc1682 = loc("forward":4294967295:6286) +#loc1683 = loc("forward":4294967295:6287) +#loc1684 = loc("forward":4294967295:6289) +#loc1685 = loc("forward":4294967295:6290) +#loc1686 = loc("forward":4294967295:6291) +#loc1687 = loc("forward":4294967295:6292) +#loc1688 = loc("forward":4294967295:6293) +#loc1689 = loc("forward":4294967295:6294) +#loc1690 = loc("forward":4294967295:6296) +#loc1691 = loc("forward":4294967295:6297) +#loc1692 = loc("forward":4294967295:6298) +#loc1693 = loc("forward":4294967295:6300) +#loc1694 = loc("forward":4294967295:6301) +#loc1695 = loc("forward":4294967295:6302) +#loc1696 = loc("forward":4294967295:6303) +#loc1697 = loc("forward":4294967295:6305) +#loc1698 = loc("forward":4294967295:6306) +#loc1699 = loc("forward":4294967295:6307) +#loc1700 = loc("forward":4294967295:6309) +#loc1701 = loc("forward":4294967295:6311) +#loc1702 = loc("forward":4294967295:6312) +#loc1703 = loc("forward":4294967295:6313) +#loc1704 = loc("forward":4294967295:6314) +#loc1705 = loc("forward":4294967295:6315) +#loc1706 = loc("forward":4294967295:6316) +#loc1707 = loc("forward":4294967295:6317) +#loc1708 = loc("forward":4294967295:6318) +#loc1709 = loc("forward":4294967295:6319) +#loc1710 = loc("forward":4294967295:6321) +#loc1711 = loc("forward":4294967295:6322) +#loc1712 = loc("forward":4294967295:6323) +#loc1713 = loc("forward":4294967295:6324) +#loc1714 = loc("forward":4294967295:6326) +#loc1715 = loc("forward":4294967295:6327) +#loc1716 = loc("forward":4294967295:6328) +#loc1717 = loc("forward":4294967295:6330) +#loc1718 = loc("forward":4294967295:6332) +#loc1719 = loc("forward":4294967295:6333) +#loc1720 = loc("forward":4294967295:6334) +#loc1721 = loc("forward":4294967295:6335) +#loc1722 = loc("forward":4294967295:6336) +#loc1723 = loc("forward":4294967295:6337) +#loc1724 = loc("forward":4294967295:6338) +#loc1725 = loc("forward":4294967295:6339) +#loc1726 = loc("forward":4294967295:6340) +#loc1727 = loc("forward":4294967295:6341) +#loc1728 = loc("forward":4294967295:6343) +#loc1729 = loc("forward":4294967295:6345) +#loc1730 = loc("forward":4294967295:6346) +#loc1731 = loc("forward":4294967295:6347) +#loc1732 = loc("forward":4294967295:6349) +#loc1733 = loc("forward":4294967295:6350) +#loc1734 = loc("forward":4294967295:6351) +#loc1735 = loc("forward":4294967295:6352) +#loc1736 = loc("forward":4294967295:6353) +#loc1737 = loc("forward":4294967295:6354) +#loc1738 = loc("forward":4294967295:6355) +#loc1739 = loc("forward":4294967295:6356) +#loc1740 = loc("forward":4294967295:6357) +#loc1741 = loc("forward":4294967295:6358) +#loc1742 = loc("forward":4294967295:6360) +#loc1743 = loc("forward":4294967295:6361) +#loc1744 = loc("forward":4294967295:6362) +#loc1745 = loc("forward":4294967295:6364) +#loc1746 = loc("forward":4294967295:6365) +#loc1747 = loc("forward":4294967295:6367) +#loc1748 = loc("forward":4294967295:6368) +#loc1749 = loc("forward":4294967295:6369) +#loc1750 = loc("forward":4294967295:6370) +#loc1751 = loc("forward":4294967295:6371) +#loc1752 = loc("forward":4294967295:6372) +#loc1753 = loc("forward":4294967295:6374) +#loc1754 = loc("forward":4294967295:6375) +#loc1755 = loc("forward":4294967295:6376) +#loc1756 = loc("forward":4294967295:6377) +#loc1757 = loc("forward":4294967295:6379) +#loc1758 = loc("forward":4294967295:6380) +#loc1759 = loc("forward":4294967295:6381) +#loc1760 = loc("forward":4294967295:6383) +#loc1761 = loc("forward":4294967295:6384) +#loc1762 = loc("forward":4294967295:6386) +#loc1763 = loc("forward":4294967295:6387) +#loc1764 = loc("forward":4294967295:6389) +#loc1765 = loc("forward":4294967295:6390) +#loc1766 = loc("forward":4294967295:6391) +#loc1767 = loc("forward":4294967295:6392) +#loc1768 = loc("forward":4294967295:6393) +#loc1769 = loc("forward":4294967295:6394) +#loc1770 = loc("forward":4294967295:6396) +#loc1771 = loc("forward":4294967295:6397) +#loc1772 = loc("forward":4294967295:6398) +#loc1773 = loc("forward":4294967295:6400) +#loc1774 = loc("forward":4294967295:6401) +#loc1775 = loc("forward":4294967295:6402) +#loc1776 = loc("forward":4294967295:6403) +#loc1777 = loc("forward":4294967295:6405) +#loc1778 = loc("forward":4294967295:6406) +#loc1779 = loc("forward":4294967295:6407) +#loc1780 = loc("forward":4294967295:6409) +#loc1781 = loc("forward":4294967295:6411) +#loc1782 = loc("forward":4294967295:6412) +#loc1783 = loc("forward":4294967295:6413) +#loc1784 = loc("forward":4294967295:6414) +#loc1785 = loc("forward":4294967295:6415) +#loc1786 = loc("forward":4294967295:6416) +#loc1787 = loc("forward":4294967295:6417) +#loc1788 = loc("forward":4294967295:6418) +#loc1789 = loc("forward":4294967295:6419) +#loc1790 = loc("forward":4294967295:6421) +#loc1791 = loc("forward":4294967295:6422) +#loc1792 = loc("forward":4294967295:6423) +#loc1793 = loc("forward":4294967295:6424) +#loc1794 = loc("forward":4294967295:6426) +#loc1795 = loc("forward":4294967295:6427) +#loc1796 = loc("forward":4294967295:6428) +#loc1797 = loc("forward":4294967295:6430) +#loc1798 = loc("forward":4294967295:6432) +#loc1799 = loc("forward":4294967295:6433) +#loc1800 = loc("forward":4294967295:6434) +#loc1801 = loc("forward":4294967295:6435) +#loc1802 = loc("forward":4294967295:6436) +#loc1803 = loc("forward":4294967295:6437) +#loc1804 = loc("forward":4294967295:6438) +#loc1805 = loc("forward":4294967295:6439) +#loc1806 = loc("forward":4294967295:6440) +#loc1807 = loc("forward":4294967295:6441) +#loc1808 = loc("forward":4294967295:6443) +#loc1809 = loc("forward":4294967295:6445) +#loc1810 = loc("forward":4294967295:6446) +#loc1811 = loc("forward":4294967295:6447) +#loc1812 = loc("forward":4294967295:6449) +#loc1813 = loc("forward":4294967295:6450) +#loc1814 = loc("forward":4294967295:6451) +#loc1815 = loc("forward":4294967295:6452) +#loc1816 = loc("forward":4294967295:6453) +#loc1817 = loc("forward":4294967295:6454) +#loc1818 = loc("forward":4294967295:6455) +#loc1819 = loc("forward":4294967295:6456) +#loc1820 = loc("forward":4294967295:6457) +#loc1821 = loc("forward":4294967295:6458) +#loc1822 = loc("forward":4294967295:6460) +#loc1823 = loc("forward":4294967295:6461) +#loc1824 = loc("forward":4294967295:6462) +#loc1825 = loc("forward":4294967295:6464) +#loc1826 = loc("forward":4294967295:6465) +#loc1827 = loc("forward":4294967295:6467) +#loc1828 = loc("forward":4294967295:6468) +#loc1829 = loc("forward":4294967295:6469) +#loc1830 = loc("forward":4294967295:6470) +#loc1831 = loc("forward":4294967295:6471) +#loc1832 = loc("forward":4294967295:6472) +#loc1833 = loc("forward":4294967295:6474) +#loc1834 = loc("forward":4294967295:6475) +#loc1835 = loc("forward":4294967295:6476) +#loc1836 = loc("forward":4294967295:6477) +#loc1837 = loc("forward":4294967295:6479) +#loc1838 = loc("forward":4294967295:6480) +#loc1839 = loc("forward":4294967295:6481) +#loc1840 = loc("forward":4294967295:6483) +#loc1841 = loc("forward":4294967295:6484) +#loc1842 = loc("forward":4294967295:6486) +#loc1843 = loc("forward":4294967295:6487) +#loc1844 = loc("forward":4294967295:6489) +#loc1845 = loc("forward":4294967295:6490) +#loc1846 = loc("forward":4294967295:6491) +#loc1847 = loc("forward":4294967295:6492) +#loc1848 = loc("forward":4294967295:6493) +#loc1849 = loc("forward":4294967295:6494) +#loc1850 = loc("forward":4294967295:6496) +#loc1851 = loc("forward":4294967295:6497) +#loc1852 = loc("forward":4294967295:6498) +#loc1853 = loc("forward":4294967295:6500) +#loc1854 = loc("forward":4294967295:6501) +#loc1855 = loc("forward":4294967295:6502) +#loc1856 = loc("forward":4294967295:6503) +#loc1857 = loc("forward":4294967295:6505) +#loc1858 = loc("forward":4294967295:6506) +#loc1859 = loc("forward":4294967295:6507) +#loc1860 = loc("forward":4294967295:6509) +#loc1861 = loc("forward":4294967295:6511) +#loc1862 = loc("forward":4294967295:6512) +#loc1863 = loc("forward":4294967295:6513) +#loc1864 = loc("forward":4294967295:6514) +#loc1865 = loc("forward":4294967295:6515) +#loc1866 = loc("forward":4294967295:6516) +#loc1867 = loc("forward":4294967295:6517) +#loc1868 = loc("forward":4294967295:6518) +#loc1869 = loc("forward":4294967295:6519) +#loc1870 = loc("forward":4294967295:6521) +#loc1871 = loc("forward":4294967295:6522) +#loc1872 = loc("forward":4294967295:6523) +#loc1873 = loc("forward":4294967295:6524) +#loc1874 = loc("forward":4294967295:6526) +#loc1875 = loc("forward":4294967295:6527) +#loc1876 = loc("forward":4294967295:6528) +#loc1877 = loc("forward":4294967295:6530) +#loc1878 = loc("forward":4294967295:6532) +#loc1879 = loc("forward":4294967295:6533) +#loc1880 = loc("forward":4294967295:6534) +#loc1881 = loc("forward":4294967295:6535) +#loc1882 = loc("forward":4294967295:6536) +#loc1883 = loc("forward":4294967295:6537) +#loc1884 = loc("forward":4294967295:6538) +#loc1885 = loc("forward":4294967295:6539) +#loc1886 = loc("forward":4294967295:6540) +#loc1887 = loc("forward":4294967295:6541) +#loc1888 = loc("forward":4294967295:6543) +#loc1889 = loc("forward":4294967295:6545) +#loc1890 = loc("forward":4294967295:6546) +#loc1891 = loc("forward":4294967295:6547) +#loc1892 = loc("forward":4294967295:6549) +#loc1893 = loc("forward":4294967295:6550) +#loc1894 = loc("forward":4294967295:6551) +#loc1895 = loc("forward":4294967295:6552) +#loc1896 = loc("forward":4294967295:6553) +#loc1897 = loc("forward":4294967295:6554) +#loc1898 = loc("forward":4294967295:6555) +#loc1899 = loc("forward":4294967295:6556) +#loc1900 = loc("forward":4294967295:6557) +#loc1901 = loc("forward":4294967295:6558) +#loc1902 = loc("forward":4294967295:6560) +#loc1903 = loc("forward":4294967295:6561) +#loc1904 = loc("forward":4294967295:6562) +#loc1905 = loc("forward":4294967295:6564) +#loc1906 = loc("forward":4294967295:6565) +#loc1907 = loc("forward":4294967295:6567) +#loc1908 = loc("forward":4294967295:6568) +#loc1909 = loc("forward":4294967295:6569) +#loc1910 = loc("forward":4294967295:6570) +#loc1911 = loc("forward":4294967295:6571) +#loc1912 = loc("forward":4294967295:6572) +#loc1913 = loc("forward":4294967295:6574) +#loc1914 = loc("forward":4294967295:6575) +#loc1915 = loc("forward":4294967295:6576) +#loc1916 = loc("forward":4294967295:6577) +#loc1917 = loc("forward":4294967295:6579) +#loc1918 = loc("forward":4294967295:6580) +#loc1919 = loc("forward":4294967295:6581) +#loc1920 = loc("forward":4294967295:6583) +#loc1921 = loc("forward":4294967295:6584) +#loc1922 = loc("forward":4294967295:6586) +#loc1923 = loc("forward":4294967295:6587) +#loc1924 = loc("forward":4294967295:6589) +#loc1925 = loc("forward":4294967295:6590) +#loc1926 = loc("forward":4294967295:6591) +#loc1927 = loc("forward":4294967295:6592) +#loc1928 = loc("forward":4294967295:6593) +#loc1929 = loc("forward":4294967295:6594) +#loc1930 = loc("forward":4294967295:6596) +#loc1931 = loc("forward":4294967295:6597) +#loc1932 = loc("forward":4294967295:6598) +#loc1933 = loc("forward":4294967295:6600) +#loc1934 = loc("forward":4294967295:6601) +#loc1935 = loc("forward":4294967295:6602) +#loc1936 = loc("forward":4294967295:6603) +#loc1937 = loc("forward":4294967295:6605) +#loc1938 = loc("forward":4294967295:6606) +#loc1939 = loc("forward":4294967295:6607) +#loc1940 = loc("forward":4294967295:6609) +#loc1941 = loc("forward":4294967295:6611) +#loc1942 = loc("forward":4294967295:6612) +#loc1943 = loc("forward":4294967295:6613) +#loc1944 = loc("forward":4294967295:6614) +#loc1945 = loc("forward":4294967295:6615) +#loc1946 = loc("forward":4294967295:6616) +#loc1947 = loc("forward":4294967295:6617) +#loc1948 = loc("forward":4294967295:6618) +#loc1949 = loc("forward":4294967295:6619) +#loc1950 = loc("forward":4294967295:6621) +#loc1951 = loc("forward":4294967295:6622) +#loc1952 = loc("forward":4294967295:6623) +#loc1953 = loc("forward":4294967295:6624) +#loc1954 = loc("forward":4294967295:6626) +#loc1955 = loc("forward":4294967295:6627) +#loc1956 = loc("forward":4294967295:6628) +#loc1957 = loc("forward":4294967295:6630) +#loc1958 = loc("forward":4294967295:6632) +#loc1959 = loc("forward":4294967295:6633) +#loc1960 = loc("forward":4294967295:6634) +#loc1961 = loc("forward":4294967295:6635) +#loc1962 = loc("forward":4294967295:6636) +#loc1963 = loc("forward":4294967295:6637) +#loc1964 = loc("forward":4294967295:6638) +#loc1965 = loc("forward":4294967295:6639) +#loc1966 = loc("forward":4294967295:6640) +#loc1967 = loc("forward":4294967295:6641) +#loc1968 = loc("forward":4294967295:6643) +#loc1969 = loc("forward":4294967295:6645) +#loc1970 = loc("forward":4294967295:6646) +#loc1971 = loc("forward":4294967295:6647) +#loc1972 = loc("forward":4294967295:6649) +#loc1973 = loc("forward":4294967295:6650) +#loc1974 = loc("forward":4294967295:6651) +#loc1975 = loc("forward":4294967295:6652) +#loc1976 = loc("forward":4294967295:6653) +#loc1977 = loc("forward":4294967295:6654) +#loc1978 = loc("forward":4294967295:6655) +#loc1979 = loc("forward":4294967295:6656) +#loc1980 = loc("forward":4294967295:6657) +#loc1981 = loc("forward":4294967295:6658) +#loc1982 = loc("forward":4294967295:6660) +#loc1983 = loc("forward":4294967295:6661) +#loc1984 = loc("forward":4294967295:6662) +#loc1985 = loc("forward":4294967295:6664) +#loc1986 = loc("forward":4294967295:6665) +#loc1987 = loc("forward":4294967295:6667) +#loc1988 = loc("forward":4294967295:6668) +#loc1989 = loc("forward":4294967295:6669) +#loc1990 = loc("forward":4294967295:6670) +#loc1991 = loc("forward":4294967295:6671) +#loc1992 = loc("forward":4294967295:6672) +#loc1993 = loc("forward":4294967295:6674) +#loc1994 = loc("forward":4294967295:6675) +#loc1995 = loc("forward":4294967295:6676) +#loc1996 = loc("forward":4294967295:6677) +#loc1997 = loc("forward":4294967295:6679) +#loc1998 = loc("forward":4294967295:6680) +#loc1999 = loc("forward":4294967295:6681) +#loc2000 = loc("forward":4294967295:6683) +#loc2001 = loc("forward":4294967295:6684) +#loc2002 = loc("forward":4294967295:6686) +#loc2003 = loc("forward":4294967295:6687) +#loc2004 = loc("forward":4294967295:6689) +#loc2005 = loc("forward":4294967295:6690) +#loc2006 = loc("forward":4294967295:6691) +#loc2007 = loc("forward":4294967295:6692) +#loc2008 = loc("forward":4294967295:6693) +#loc2009 = loc("forward":4294967295:6694) +#loc2010 = loc("forward":4294967295:6696) +#loc2011 = loc("forward":4294967295:6697) +#loc2012 = loc("forward":4294967295:6698) +#loc2013 = loc("forward":4294967295:6700) +#loc2014 = loc("forward":4294967295:6701) +#loc2015 = loc("forward":4294967295:6702) +#loc2016 = loc("forward":4294967295:6703) +#loc2017 = loc("forward":4294967295:6705) +#loc2018 = loc("forward":4294967295:6706) +#loc2019 = loc("forward":4294967295:6707) +#loc2020 = loc("forward":4294967295:6709) +#loc2021 = loc("forward":4294967295:6711) +#loc2022 = loc("forward":4294967295:6712) +#loc2023 = loc("forward":4294967295:6713) +#loc2024 = loc("forward":4294967295:6714) +#loc2025 = loc("forward":4294967295:6715) +#loc2026 = loc("forward":4294967295:6716) +#loc2027 = loc("forward":4294967295:6717) +#loc2028 = loc("forward":4294967295:6718) +#loc2029 = loc("forward":4294967295:6719) +#loc2030 = loc("forward":4294967295:6721) +#loc2031 = loc("forward":4294967295:6722) +#loc2032 = loc("forward":4294967295:6723) +#loc2033 = loc("forward":4294967295:6724) +#loc2034 = loc("forward":4294967295:6726) +#loc2035 = loc("forward":4294967295:6727) +#loc2036 = loc("forward":4294967295:6728) +#loc2037 = loc("forward":4294967295:6730) +#loc2038 = loc("forward":4294967295:6732) +#loc2039 = loc("forward":4294967295:6733) +#loc2040 = loc("forward":4294967295:6734) +#loc2041 = loc("forward":4294967295:6735) +#loc2042 = loc("forward":4294967295:6736) +#loc2043 = loc("forward":4294967295:6737) +#loc2044 = loc("forward":4294967295:6738) +#loc2045 = loc("forward":4294967295:6739) +#loc2046 = loc("forward":4294967295:6740) +#loc2047 = loc("forward":4294967295:6741) +#loc2048 = loc("forward":4294967295:6743) +#loc2049 = loc("forward":4294967295:6745) +#loc2050 = loc("forward":4294967295:6746) +#loc2051 = loc("forward":4294967295:6747) +#loc2052 = loc("forward":4294967295:6749) +#loc2053 = loc("forward":4294967295:6750) +#loc2054 = loc("forward":4294967295:6751) +#loc2055 = loc("forward":4294967295:6752) +#loc2056 = loc("forward":4294967295:6753) +#loc2057 = loc("forward":4294967295:6754) +#loc2058 = loc("forward":4294967295:6755) +#loc2059 = loc("forward":4294967295:6756) +#loc2060 = loc("forward":4294967295:6757) +#loc2061 = loc("forward":4294967295:6758) +#loc2062 = loc("forward":4294967295:6760) +#loc2063 = loc("forward":4294967295:6761) +#loc2064 = loc("forward":4294967295:6762) +#loc2065 = loc("forward":4294967295:6764) +#loc2066 = loc("forward":4294967295:6765) +#loc2067 = loc("forward":4294967295:6767) +#loc2068 = loc("forward":4294967295:6768) +#loc2069 = loc("forward":4294967295:6769) +#loc2070 = loc("forward":4294967295:6770) +#loc2071 = loc("forward":4294967295:6771) +#loc2072 = loc("forward":4294967295:6772) +#loc2073 = loc("forward":4294967295:6774) +#loc2074 = loc("forward":4294967295:6775) +#loc2075 = loc("forward":4294967295:6776) +#loc2076 = loc("forward":4294967295:6777) +#loc2077 = loc("forward":4294967295:6779) +#loc2078 = loc("forward":4294967295:6780) +#loc2079 = loc("forward":4294967295:6781) +#loc2080 = loc("forward":4294967295:6783) +#loc2081 = loc("forward":4294967295:6784) +#loc2082 = loc("forward":4294967295:6785) +#loc2083 = loc("forward":4294967295:6786) +#loc2084 = loc("forward":4294967295:6788) +#loc2085 = loc("forward":4294967295:6789) +#loc2086 = loc("forward":4294967295:6790) +#loc2087 = loc("forward":4294967295:6791) +#loc2088 = loc("forward":4294967295:6792) +#loc2089 = loc("forward":4294967295:6794) +#loc2090 = loc(unknown) +#loc2091 = loc("embedding_1"(#loc1)) +#loc2092 = loc("multiply_2"(#loc2)) +#loc2093 = loc("reduce_avg_3"(#loc3)) +#loc2094 = loc("add_4"(#loc4)) +#loc2095 = loc("sqrt_5"(#loc5)) +#loc2096 = loc("reciprocal_6"(#loc6)) +#loc2097 = loc("multiply_7"(#loc7)) +#loc2098 = loc("multiply_8"(#loc8)) +#loc2099 = loc("reshape_9.dc.squeeze.0"(#loc9)) +#loc2100 = loc("matmul_11"(#loc10)) +#loc2101 = loc("reshape_12"(#loc11)) +#loc2102 = loc("transpose_13"(#loc12)) +#loc2103 = loc("concatenate_20"(#loc13)) +#loc2104 = loc("cosine_21"(#loc14)) +#loc2105 = loc("unsqueeze_22"(#loc15)) +#loc2106 = loc("multiply_23"(#loc16)) +#loc2107 = loc("index_24.dc.transpose.0"(#loc17)) +#loc2108 = loc("index_24.dc.matmul.2"(#loc18)) +#loc2109 = loc("index_24.dc.transpose.3"(#loc19)) +#loc2110 = loc("multiply_25"(#loc20)) +#loc2111 = loc("index_26.dc.transpose.0"(#loc21)) +#loc2112 = loc("index_26.dc.matmul.2"(#loc22)) +#loc2113 = loc("index_26.dc.transpose.3"(#loc23)) +#loc2114 = loc("concatenate_27"(#loc24)) +#loc2115 = loc("sine_28"(#loc25)) +#loc2116 = loc("unsqueeze_29"(#loc26)) +#loc2117 = loc("multiply_30"(#loc27)) +#loc2118 = loc("add_31"(#loc28)) +#loc2119 = loc("reshape_32.dc.squeeze.0"(#loc29)) +#loc2120 = loc("matmul_34"(#loc30)) +#loc2121 = loc("reshape_35"(#loc31)) +#loc2122 = loc("transpose_36"(#loc32)) +#loc2123 = loc("multiply_37"(#loc33)) +#loc2124 = loc("index_38.dc.transpose.0"(#loc34)) +#loc2125 = loc("index_38.dc.matmul.2"(#loc35)) +#loc2126 = loc("index_38.dc.transpose.3"(#loc36)) +#loc2127 = loc("multiply_39"(#loc37)) +#loc2128 = loc("index_40.dc.transpose.0"(#loc38)) +#loc2129 = loc("index_40.dc.matmul.2"(#loc39)) +#loc2130 = loc("index_40.dc.transpose.3"(#loc40)) +#loc2131 = loc("concatenate_41"(#loc41)) +#loc2132 = loc("multiply_42"(#loc42)) +#loc2133 = loc("add_43"(#loc43)) +#loc2134 = loc("reshape_44.dc.squeeze.0"(#loc44)) +#loc2135 = loc("transpose_45"(#loc45)) +#loc2136 = loc("matmul_46"(#loc46)) +#loc2137 = loc("reshape_47.dc.unsqueeze.0"(#loc47)) +#loc2138 = loc("multiply_48"(#loc48)) +#loc2139 = loc("add_49"(#loc49)) +#loc2140 = loc("softmax_50"(#loc50)) +#loc2141 = loc("reshape_52.dc.squeeze.0"(#loc51)) +#loc2142 = loc("matmul_54"(#loc52)) +#loc2143 = loc("reshape_55"(#loc53)) +#loc2144 = loc("transpose_56"(#loc54)) +#loc2145 = loc("transpose_57"(#loc55)) +#loc2146 = loc("reshape_58.dc.squeeze.0"(#loc56)) +#loc2147 = loc("transpose_59"(#loc57)) +#loc2148 = loc("matmul_60"(#loc58)) +#loc2149 = loc("reshape_61.dc.unsqueeze.0"(#loc59)) +#loc2150 = loc("transpose_62"(#loc60)) +#loc2151 = loc("reshape_63"(#loc61)) +#loc2152 = loc("matmul_65"(#loc62)) +#loc2153 = loc("reshape_66.dc.unsqueeze.0"(#loc63)) +#loc2154 = loc("add_67"(#loc64)) +#loc2155 = loc("multiply_68"(#loc65)) +#loc2156 = loc("reduce_avg_69"(#loc66)) +#loc2157 = loc("add_70"(#loc67)) +#loc2158 = loc("sqrt_71"(#loc68)) +#loc2159 = loc("reciprocal_72"(#loc69)) +#loc2160 = loc("multiply_73"(#loc70)) +#loc2161 = loc("multiply_74"(#loc71)) +#loc2162 = loc("reshape_75.dc.squeeze.0"(#loc72)) +#loc2163 = loc("matmul_77"(#loc73)) +#loc2164 = loc("reshape_78.dc.unsqueeze.0"(#loc74)) +#loc2165 = loc("sigmoid_79"(#loc75)) +#loc2166 = loc("multiply_80"(#loc76)) +#loc2167 = loc("matmul_82"(#loc77)) +#loc2168 = loc("reshape_83.dc.unsqueeze.0"(#loc78)) +#loc2169 = loc("multiply_84"(#loc79)) +#loc2170 = loc("matmul_86"(#loc80)) +#loc2171 = loc("add_87"(#loc81)) +#loc2172 = loc("multiply_88"(#loc82)) +#loc2173 = loc("reduce_avg_89"(#loc83)) +#loc2174 = loc("add_90"(#loc84)) +#loc2175 = loc("sqrt_91"(#loc85)) +#loc2176 = loc("reciprocal_92"(#loc86)) +#loc2177 = loc("multiply_93"(#loc87)) +#loc2178 = loc("multiply_94"(#loc88)) +#loc2179 = loc("reshape_95.dc.squeeze.0"(#loc89)) +#loc2180 = loc("matmul_97"(#loc90)) +#loc2181 = loc("reshape_98"(#loc91)) +#loc2182 = loc("transpose_99"(#loc92)) +#loc2183 = loc("concatenate_106"(#loc93)) +#loc2184 = loc("cosine_107"(#loc94)) +#loc2185 = loc("unsqueeze_108"(#loc95)) +#loc2186 = loc("multiply_109"(#loc96)) +#loc2187 = loc("index_110.dc.transpose.0"(#loc97)) +#loc2188 = loc("index_110.dc.matmul.2"(#loc98)) +#loc2189 = loc("index_110.dc.transpose.3"(#loc99)) +#loc2190 = loc("multiply_111"(#loc100)) +#loc2191 = loc("index_112.dc.transpose.0"(#loc101)) +#loc2192 = loc("index_112.dc.matmul.2"(#loc102)) +#loc2193 = loc("index_112.dc.transpose.3"(#loc103)) +#loc2194 = loc("concatenate_113"(#loc104)) +#loc2195 = loc("sine_114"(#loc105)) +#loc2196 = loc("unsqueeze_115"(#loc106)) +#loc2197 = loc("multiply_116"(#loc107)) +#loc2198 = loc("add_117"(#loc108)) +#loc2199 = loc("reshape_118.dc.squeeze.0"(#loc109)) +#loc2200 = loc("matmul_120"(#loc110)) +#loc2201 = loc("reshape_121"(#loc111)) +#loc2202 = loc("transpose_122"(#loc112)) +#loc2203 = loc("multiply_123"(#loc113)) +#loc2204 = loc("index_124.dc.transpose.0"(#loc114)) +#loc2205 = loc("index_124.dc.matmul.2"(#loc115)) +#loc2206 = loc("index_124.dc.transpose.3"(#loc116)) +#loc2207 = loc("multiply_125"(#loc117)) +#loc2208 = loc("index_126.dc.transpose.0"(#loc118)) +#loc2209 = loc("index_126.dc.matmul.2"(#loc119)) +#loc2210 = loc("index_126.dc.transpose.3"(#loc120)) +#loc2211 = loc("concatenate_127"(#loc121)) +#loc2212 = loc("multiply_128"(#loc122)) +#loc2213 = loc("add_129"(#loc123)) +#loc2214 = loc("reshape_130.dc.squeeze.0"(#loc124)) +#loc2215 = loc("transpose_131"(#loc125)) +#loc2216 = loc("matmul_132"(#loc126)) +#loc2217 = loc("reshape_133.dc.unsqueeze.0"(#loc127)) +#loc2218 = loc("multiply_134"(#loc128)) +#loc2219 = loc("add_135"(#loc129)) +#loc2220 = loc("softmax_136"(#loc130)) +#loc2221 = loc("reshape_138.dc.squeeze.0"(#loc131)) +#loc2222 = loc("matmul_140"(#loc132)) +#loc2223 = loc("reshape_141"(#loc133)) +#loc2224 = loc("transpose_142"(#loc134)) +#loc2225 = loc("transpose_143"(#loc135)) +#loc2226 = loc("reshape_144.dc.squeeze.0"(#loc136)) +#loc2227 = loc("transpose_145"(#loc137)) +#loc2228 = loc("matmul_146"(#loc138)) +#loc2229 = loc("reshape_147.dc.unsqueeze.0"(#loc139)) +#loc2230 = loc("transpose_148"(#loc140)) +#loc2231 = loc("reshape_149"(#loc141)) +#loc2232 = loc("matmul_151"(#loc142)) +#loc2233 = loc("reshape_152.dc.unsqueeze.0"(#loc143)) +#loc2234 = loc("add_153"(#loc144)) +#loc2235 = loc("multiply_154"(#loc145)) +#loc2236 = loc("reduce_avg_155"(#loc146)) +#loc2237 = loc("add_156"(#loc147)) +#loc2238 = loc("sqrt_157"(#loc148)) +#loc2239 = loc("reciprocal_158"(#loc149)) +#loc2240 = loc("multiply_159"(#loc150)) +#loc2241 = loc("multiply_160"(#loc151)) +#loc2242 = loc("reshape_161.dc.squeeze.0"(#loc152)) +#loc2243 = loc("matmul_163"(#loc153)) +#loc2244 = loc("reshape_164.dc.unsqueeze.0"(#loc154)) +#loc2245 = loc("sigmoid_165"(#loc155)) +#loc2246 = loc("multiply_166"(#loc156)) +#loc2247 = loc("matmul_168"(#loc157)) +#loc2248 = loc("reshape_169.dc.unsqueeze.0"(#loc158)) +#loc2249 = loc("multiply_170"(#loc159)) +#loc2250 = loc("matmul_172"(#loc160)) +#loc2251 = loc("add_173"(#loc161)) +#loc2252 = loc("multiply_174"(#loc162)) +#loc2253 = loc("reduce_avg_175"(#loc163)) +#loc2254 = loc("add_176"(#loc164)) +#loc2255 = loc("sqrt_177"(#loc165)) +#loc2256 = loc("reciprocal_178"(#loc166)) +#loc2257 = loc("multiply_179"(#loc167)) +#loc2258 = loc("multiply_180"(#loc168)) +#loc2259 = loc("reshape_181.dc.squeeze.0"(#loc169)) +#loc2260 = loc("matmul_183"(#loc170)) +#loc2261 = loc("reshape_184"(#loc171)) +#loc2262 = loc("transpose_185"(#loc172)) +#loc2263 = loc("concatenate_192"(#loc173)) +#loc2264 = loc("cosine_193"(#loc174)) +#loc2265 = loc("unsqueeze_194"(#loc175)) +#loc2266 = loc("multiply_195"(#loc176)) +#loc2267 = loc("index_196.dc.transpose.0"(#loc177)) +#loc2268 = loc("index_196.dc.matmul.2"(#loc178)) +#loc2269 = loc("index_196.dc.transpose.3"(#loc179)) +#loc2270 = loc("multiply_197"(#loc180)) +#loc2271 = loc("index_198.dc.transpose.0"(#loc181)) +#loc2272 = loc("index_198.dc.matmul.2"(#loc182)) +#loc2273 = loc("index_198.dc.transpose.3"(#loc183)) +#loc2274 = loc("concatenate_199"(#loc184)) +#loc2275 = loc("sine_200"(#loc185)) +#loc2276 = loc("unsqueeze_201"(#loc186)) +#loc2277 = loc("multiply_202"(#loc187)) +#loc2278 = loc("add_203"(#loc188)) +#loc2279 = loc("reshape_204.dc.squeeze.0"(#loc189)) +#loc2280 = loc("matmul_206"(#loc190)) +#loc2281 = loc("reshape_207"(#loc191)) +#loc2282 = loc("transpose_208"(#loc192)) +#loc2283 = loc("multiply_209"(#loc193)) +#loc2284 = loc("index_210.dc.transpose.0"(#loc194)) +#loc2285 = loc("index_210.dc.matmul.2"(#loc195)) +#loc2286 = loc("index_210.dc.transpose.3"(#loc196)) +#loc2287 = loc("multiply_211"(#loc197)) +#loc2288 = loc("index_212.dc.transpose.0"(#loc198)) +#loc2289 = loc("index_212.dc.matmul.2"(#loc199)) +#loc2290 = loc("index_212.dc.transpose.3"(#loc200)) +#loc2291 = loc("concatenate_213"(#loc201)) +#loc2292 = loc("multiply_214"(#loc202)) +#loc2293 = loc("add_215"(#loc203)) +#loc2294 = loc("reshape_216.dc.squeeze.0"(#loc204)) +#loc2295 = loc("transpose_217"(#loc205)) +#loc2296 = loc("matmul_218"(#loc206)) +#loc2297 = loc("reshape_219.dc.unsqueeze.0"(#loc207)) +#loc2298 = loc("multiply_220"(#loc208)) +#loc2299 = loc("add_221"(#loc209)) +#loc2300 = loc("softmax_222"(#loc210)) +#loc2301 = loc("reshape_224.dc.squeeze.0"(#loc211)) +#loc2302 = loc("matmul_226"(#loc212)) +#loc2303 = loc("reshape_227"(#loc213)) +#loc2304 = loc("transpose_228"(#loc214)) +#loc2305 = loc("transpose_229"(#loc215)) +#loc2306 = loc("reshape_230.dc.squeeze.0"(#loc216)) +#loc2307 = loc("transpose_231"(#loc217)) +#loc2308 = loc("matmul_232"(#loc218)) +#loc2309 = loc("reshape_233.dc.unsqueeze.0"(#loc219)) +#loc2310 = loc("transpose_234"(#loc220)) +#loc2311 = loc("reshape_235"(#loc221)) +#loc2312 = loc("matmul_237"(#loc222)) +#loc2313 = loc("reshape_238.dc.unsqueeze.0"(#loc223)) +#loc2314 = loc("add_239"(#loc224)) +#loc2315 = loc("multiply_240"(#loc225)) +#loc2316 = loc("reduce_avg_241"(#loc226)) +#loc2317 = loc("add_242"(#loc227)) +#loc2318 = loc("sqrt_243"(#loc228)) +#loc2319 = loc("reciprocal_244"(#loc229)) +#loc2320 = loc("multiply_245"(#loc230)) +#loc2321 = loc("multiply_246"(#loc231)) +#loc2322 = loc("reshape_247.dc.squeeze.0"(#loc232)) +#loc2323 = loc("matmul_249"(#loc233)) +#loc2324 = loc("reshape_250.dc.unsqueeze.0"(#loc234)) +#loc2325 = loc("sigmoid_251"(#loc235)) +#loc2326 = loc("multiply_252"(#loc236)) +#loc2327 = loc("matmul_254"(#loc237)) +#loc2328 = loc("reshape_255.dc.unsqueeze.0"(#loc238)) +#loc2329 = loc("multiply_256"(#loc239)) +#loc2330 = loc("matmul_258"(#loc240)) +#loc2331 = loc("add_259"(#loc241)) +#loc2332 = loc("multiply_260"(#loc242)) +#loc2333 = loc("reduce_avg_261"(#loc243)) +#loc2334 = loc("add_262"(#loc244)) +#loc2335 = loc("sqrt_263"(#loc245)) +#loc2336 = loc("reciprocal_264"(#loc246)) +#loc2337 = loc("multiply_265"(#loc247)) +#loc2338 = loc("multiply_266"(#loc248)) +#loc2339 = loc("reshape_267.dc.squeeze.0"(#loc249)) +#loc2340 = loc("matmul_269"(#loc250)) +#loc2341 = loc("reshape_270"(#loc251)) +#loc2342 = loc("transpose_271"(#loc252)) +#loc2343 = loc("concatenate_278"(#loc253)) +#loc2344 = loc("cosine_279"(#loc254)) +#loc2345 = loc("unsqueeze_280"(#loc255)) +#loc2346 = loc("multiply_281"(#loc256)) +#loc2347 = loc("index_282.dc.transpose.0"(#loc257)) +#loc2348 = loc("index_282.dc.matmul.2"(#loc258)) +#loc2349 = loc("index_282.dc.transpose.3"(#loc259)) +#loc2350 = loc("multiply_283"(#loc260)) +#loc2351 = loc("index_284.dc.transpose.0"(#loc261)) +#loc2352 = loc("index_284.dc.matmul.2"(#loc262)) +#loc2353 = loc("index_284.dc.transpose.3"(#loc263)) +#loc2354 = loc("concatenate_285"(#loc264)) +#loc2355 = loc("sine_286"(#loc265)) +#loc2356 = loc("unsqueeze_287"(#loc266)) +#loc2357 = loc("multiply_288"(#loc267)) +#loc2358 = loc("add_289"(#loc268)) +#loc2359 = loc("reshape_290.dc.squeeze.0"(#loc269)) +#loc2360 = loc("matmul_292"(#loc270)) +#loc2361 = loc("reshape_293"(#loc271)) +#loc2362 = loc("transpose_294"(#loc272)) +#loc2363 = loc("multiply_295"(#loc273)) +#loc2364 = loc("index_296.dc.transpose.0"(#loc274)) +#loc2365 = loc("index_296.dc.matmul.2"(#loc275)) +#loc2366 = loc("index_296.dc.transpose.3"(#loc276)) +#loc2367 = loc("multiply_297"(#loc277)) +#loc2368 = loc("index_298.dc.transpose.0"(#loc278)) +#loc2369 = loc("index_298.dc.matmul.2"(#loc279)) +#loc2370 = loc("index_298.dc.transpose.3"(#loc280)) +#loc2371 = loc("concatenate_299"(#loc281)) +#loc2372 = loc("multiply_300"(#loc282)) +#loc2373 = loc("add_301"(#loc283)) +#loc2374 = loc("reshape_302.dc.squeeze.0"(#loc284)) +#loc2375 = loc("transpose_303"(#loc285)) +#loc2376 = loc("matmul_304"(#loc286)) +#loc2377 = loc("reshape_305.dc.unsqueeze.0"(#loc287)) +#loc2378 = loc("multiply_306"(#loc288)) +#loc2379 = loc("add_307"(#loc289)) +#loc2380 = loc("softmax_308"(#loc290)) +#loc2381 = loc("reshape_310.dc.squeeze.0"(#loc291)) +#loc2382 = loc("matmul_312"(#loc292)) +#loc2383 = loc("reshape_313"(#loc293)) +#loc2384 = loc("transpose_314"(#loc294)) +#loc2385 = loc("transpose_315"(#loc295)) +#loc2386 = loc("reshape_316.dc.squeeze.0"(#loc296)) +#loc2387 = loc("transpose_317"(#loc297)) +#loc2388 = loc("matmul_318"(#loc298)) +#loc2389 = loc("reshape_319.dc.unsqueeze.0"(#loc299)) +#loc2390 = loc("transpose_320"(#loc300)) +#loc2391 = loc("reshape_321"(#loc301)) +#loc2392 = loc("matmul_323"(#loc302)) +#loc2393 = loc("reshape_324.dc.unsqueeze.0"(#loc303)) +#loc2394 = loc("add_325"(#loc304)) +#loc2395 = loc("multiply_326"(#loc305)) +#loc2396 = loc("reduce_avg_327"(#loc306)) +#loc2397 = loc("add_328"(#loc307)) +#loc2398 = loc("sqrt_329"(#loc308)) +#loc2399 = loc("reciprocal_330"(#loc309)) +#loc2400 = loc("multiply_331"(#loc310)) +#loc2401 = loc("multiply_332"(#loc311)) +#loc2402 = loc("reshape_333.dc.squeeze.0"(#loc312)) +#loc2403 = loc("matmul_335"(#loc313)) +#loc2404 = loc("reshape_336.dc.unsqueeze.0"(#loc314)) +#loc2405 = loc("sigmoid_337"(#loc315)) +#loc2406 = loc("multiply_338"(#loc316)) +#loc2407 = loc("matmul_340"(#loc317)) +#loc2408 = loc("reshape_341.dc.unsqueeze.0"(#loc318)) +#loc2409 = loc("multiply_342"(#loc319)) +#loc2410 = loc("matmul_344"(#loc320)) +#loc2411 = loc("add_345"(#loc321)) +#loc2412 = loc("multiply_346"(#loc322)) +#loc2413 = loc("reduce_avg_347"(#loc323)) +#loc2414 = loc("add_348"(#loc324)) +#loc2415 = loc("sqrt_349"(#loc325)) +#loc2416 = loc("reciprocal_350"(#loc326)) +#loc2417 = loc("multiply_351"(#loc327)) +#loc2418 = loc("multiply_352"(#loc328)) +#loc2419 = loc("reshape_353.dc.squeeze.0"(#loc329)) +#loc2420 = loc("matmul_355"(#loc330)) +#loc2421 = loc("reshape_356"(#loc331)) +#loc2422 = loc("transpose_357"(#loc332)) +#loc2423 = loc("concatenate_364"(#loc333)) +#loc2424 = loc("cosine_365"(#loc334)) +#loc2425 = loc("unsqueeze_366"(#loc335)) +#loc2426 = loc("multiply_367"(#loc336)) +#loc2427 = loc("index_368.dc.transpose.0"(#loc337)) +#loc2428 = loc("index_368.dc.matmul.2"(#loc338)) +#loc2429 = loc("index_368.dc.transpose.3"(#loc339)) +#loc2430 = loc("multiply_369"(#loc340)) +#loc2431 = loc("index_370.dc.transpose.0"(#loc341)) +#loc2432 = loc("index_370.dc.matmul.2"(#loc342)) +#loc2433 = loc("index_370.dc.transpose.3"(#loc343)) +#loc2434 = loc("concatenate_371"(#loc344)) +#loc2435 = loc("sine_372"(#loc345)) +#loc2436 = loc("unsqueeze_373"(#loc346)) +#loc2437 = loc("multiply_374"(#loc347)) +#loc2438 = loc("add_375"(#loc348)) +#loc2439 = loc("reshape_376.dc.squeeze.0"(#loc349)) +#loc2440 = loc("matmul_378"(#loc350)) +#loc2441 = loc("reshape_379"(#loc351)) +#loc2442 = loc("transpose_380"(#loc352)) +#loc2443 = loc("multiply_381"(#loc353)) +#loc2444 = loc("index_382.dc.transpose.0"(#loc354)) +#loc2445 = loc("index_382.dc.matmul.2"(#loc355)) +#loc2446 = loc("index_382.dc.transpose.3"(#loc356)) +#loc2447 = loc("multiply_383"(#loc357)) +#loc2448 = loc("index_384.dc.transpose.0"(#loc358)) +#loc2449 = loc("index_384.dc.matmul.2"(#loc359)) +#loc2450 = loc("index_384.dc.transpose.3"(#loc360)) +#loc2451 = loc("concatenate_385"(#loc361)) +#loc2452 = loc("multiply_386"(#loc362)) +#loc2453 = loc("add_387"(#loc363)) +#loc2454 = loc("reshape_388.dc.squeeze.0"(#loc364)) +#loc2455 = loc("transpose_389"(#loc365)) +#loc2456 = loc("matmul_390"(#loc366)) +#loc2457 = loc("reshape_391.dc.unsqueeze.0"(#loc367)) +#loc2458 = loc("multiply_392"(#loc368)) +#loc2459 = loc("add_393"(#loc369)) +#loc2460 = loc("softmax_394"(#loc370)) +#loc2461 = loc("reshape_396.dc.squeeze.0"(#loc371)) +#loc2462 = loc("matmul_398"(#loc372)) +#loc2463 = loc("reshape_399"(#loc373)) +#loc2464 = loc("transpose_400"(#loc374)) +#loc2465 = loc("transpose_401"(#loc375)) +#loc2466 = loc("reshape_402.dc.squeeze.0"(#loc376)) +#loc2467 = loc("transpose_403"(#loc377)) +#loc2468 = loc("matmul_404"(#loc378)) +#loc2469 = loc("reshape_405.dc.unsqueeze.0"(#loc379)) +#loc2470 = loc("transpose_406"(#loc380)) +#loc2471 = loc("reshape_407"(#loc381)) +#loc2472 = loc("matmul_409"(#loc382)) +#loc2473 = loc("reshape_410.dc.unsqueeze.0"(#loc383)) +#loc2474 = loc("add_411"(#loc384)) +#loc2475 = loc("multiply_412"(#loc385)) +#loc2476 = loc("reduce_avg_413"(#loc386)) +#loc2477 = loc("add_414"(#loc387)) +#loc2478 = loc("sqrt_415"(#loc388)) +#loc2479 = loc("reciprocal_416"(#loc389)) +#loc2480 = loc("multiply_417"(#loc390)) +#loc2481 = loc("multiply_418"(#loc391)) +#loc2482 = loc("reshape_419.dc.squeeze.0"(#loc392)) +#loc2483 = loc("matmul_421"(#loc393)) +#loc2484 = loc("reshape_422.dc.unsqueeze.0"(#loc394)) +#loc2485 = loc("sigmoid_423"(#loc395)) +#loc2486 = loc("multiply_424"(#loc396)) +#loc2487 = loc("matmul_426"(#loc397)) +#loc2488 = loc("reshape_427.dc.unsqueeze.0"(#loc398)) +#loc2489 = loc("multiply_428"(#loc399)) +#loc2490 = loc("matmul_430"(#loc400)) +#loc2491 = loc("add_431"(#loc401)) +#loc2492 = loc("multiply_432"(#loc402)) +#loc2493 = loc("reduce_avg_433"(#loc403)) +#loc2494 = loc("add_434"(#loc404)) +#loc2495 = loc("sqrt_435"(#loc405)) +#loc2496 = loc("reciprocal_436"(#loc406)) +#loc2497 = loc("multiply_437"(#loc407)) +#loc2498 = loc("multiply_438"(#loc408)) +#loc2499 = loc("reshape_439.dc.squeeze.0"(#loc409)) +#loc2500 = loc("matmul_441"(#loc410)) +#loc2501 = loc("reshape_442"(#loc411)) +#loc2502 = loc("transpose_443"(#loc412)) +#loc2503 = loc("concatenate_450"(#loc413)) +#loc2504 = loc("cosine_451"(#loc414)) +#loc2505 = loc("unsqueeze_452"(#loc415)) +#loc2506 = loc("multiply_453"(#loc416)) +#loc2507 = loc("index_454.dc.transpose.0"(#loc417)) +#loc2508 = loc("index_454.dc.matmul.2"(#loc418)) +#loc2509 = loc("index_454.dc.transpose.3"(#loc419)) +#loc2510 = loc("multiply_455"(#loc420)) +#loc2511 = loc("index_456.dc.transpose.0"(#loc421)) +#loc2512 = loc("index_456.dc.matmul.2"(#loc422)) +#loc2513 = loc("index_456.dc.transpose.3"(#loc423)) +#loc2514 = loc("concatenate_457"(#loc424)) +#loc2515 = loc("sine_458"(#loc425)) +#loc2516 = loc("unsqueeze_459"(#loc426)) +#loc2517 = loc("multiply_460"(#loc427)) +#loc2518 = loc("add_461"(#loc428)) +#loc2519 = loc("reshape_462.dc.squeeze.0"(#loc429)) +#loc2520 = loc("matmul_464"(#loc430)) +#loc2521 = loc("reshape_465"(#loc431)) +#loc2522 = loc("transpose_466"(#loc432)) +#loc2523 = loc("multiply_467"(#loc433)) +#loc2524 = loc("index_468.dc.transpose.0"(#loc434)) +#loc2525 = loc("index_468.dc.matmul.2"(#loc435)) +#loc2526 = loc("index_468.dc.transpose.3"(#loc436)) +#loc2527 = loc("multiply_469"(#loc437)) +#loc2528 = loc("index_470.dc.transpose.0"(#loc438)) +#loc2529 = loc("index_470.dc.matmul.2"(#loc439)) +#loc2530 = loc("index_470.dc.transpose.3"(#loc440)) +#loc2531 = loc("concatenate_471"(#loc441)) +#loc2532 = loc("multiply_472"(#loc442)) +#loc2533 = loc("add_473"(#loc443)) +#loc2534 = loc("reshape_474.dc.squeeze.0"(#loc444)) +#loc2535 = loc("transpose_475"(#loc445)) +#loc2536 = loc("matmul_476"(#loc446)) +#loc2537 = loc("reshape_477.dc.unsqueeze.0"(#loc447)) +#loc2538 = loc("multiply_478"(#loc448)) +#loc2539 = loc("add_479"(#loc449)) +#loc2540 = loc("softmax_480"(#loc450)) +#loc2541 = loc("reshape_482.dc.squeeze.0"(#loc451)) +#loc2542 = loc("matmul_484"(#loc452)) +#loc2543 = loc("reshape_485"(#loc453)) +#loc2544 = loc("transpose_486"(#loc454)) +#loc2545 = loc("transpose_487"(#loc455)) +#loc2546 = loc("reshape_488.dc.squeeze.0"(#loc456)) +#loc2547 = loc("transpose_489"(#loc457)) +#loc2548 = loc("matmul_490"(#loc458)) +#loc2549 = loc("reshape_491.dc.unsqueeze.0"(#loc459)) +#loc2550 = loc("transpose_492"(#loc460)) +#loc2551 = loc("reshape_493"(#loc461)) +#loc2552 = loc("matmul_495"(#loc462)) +#loc2553 = loc("reshape_496.dc.unsqueeze.0"(#loc463)) +#loc2554 = loc("add_497"(#loc464)) +#loc2555 = loc("multiply_498"(#loc465)) +#loc2556 = loc("reduce_avg_499"(#loc466)) +#loc2557 = loc("add_500"(#loc467)) +#loc2558 = loc("sqrt_501"(#loc468)) +#loc2559 = loc("reciprocal_502"(#loc469)) +#loc2560 = loc("multiply_503"(#loc470)) +#loc2561 = loc("multiply_504"(#loc471)) +#loc2562 = loc("reshape_505.dc.squeeze.0"(#loc472)) +#loc2563 = loc("matmul_507"(#loc473)) +#loc2564 = loc("reshape_508.dc.unsqueeze.0"(#loc474)) +#loc2565 = loc("sigmoid_509"(#loc475)) +#loc2566 = loc("multiply_510"(#loc476)) +#loc2567 = loc("matmul_512"(#loc477)) +#loc2568 = loc("reshape_513.dc.unsqueeze.0"(#loc478)) +#loc2569 = loc("multiply_514"(#loc479)) +#loc2570 = loc("matmul_516"(#loc480)) +#loc2571 = loc("add_517"(#loc481)) +#loc2572 = loc("multiply_518"(#loc482)) +#loc2573 = loc("reduce_avg_519"(#loc483)) +#loc2574 = loc("add_520"(#loc484)) +#loc2575 = loc("sqrt_521"(#loc485)) +#loc2576 = loc("reciprocal_522"(#loc486)) +#loc2577 = loc("multiply_523"(#loc487)) +#loc2578 = loc("multiply_524"(#loc488)) +#loc2579 = loc("reshape_525.dc.squeeze.0"(#loc489)) +#loc2580 = loc("matmul_527"(#loc490)) +#loc2581 = loc("reshape_528"(#loc491)) +#loc2582 = loc("transpose_529"(#loc492)) +#loc2583 = loc("concatenate_536"(#loc493)) +#loc2584 = loc("cosine_537"(#loc494)) +#loc2585 = loc("unsqueeze_538"(#loc495)) +#loc2586 = loc("multiply_539"(#loc496)) +#loc2587 = loc("index_540.dc.transpose.0"(#loc497)) +#loc2588 = loc("index_540.dc.matmul.2"(#loc498)) +#loc2589 = loc("index_540.dc.transpose.3"(#loc499)) +#loc2590 = loc("multiply_541"(#loc500)) +#loc2591 = loc("index_542.dc.transpose.0"(#loc501)) +#loc2592 = loc("index_542.dc.matmul.2"(#loc502)) +#loc2593 = loc("index_542.dc.transpose.3"(#loc503)) +#loc2594 = loc("concatenate_543"(#loc504)) +#loc2595 = loc("sine_544"(#loc505)) +#loc2596 = loc("unsqueeze_545"(#loc506)) +#loc2597 = loc("multiply_546"(#loc507)) +#loc2598 = loc("add_547"(#loc508)) +#loc2599 = loc("reshape_548.dc.squeeze.0"(#loc509)) +#loc2600 = loc("matmul_550"(#loc510)) +#loc2601 = loc("reshape_551"(#loc511)) +#loc2602 = loc("transpose_552"(#loc512)) +#loc2603 = loc("multiply_553"(#loc513)) +#loc2604 = loc("index_554.dc.transpose.0"(#loc514)) +#loc2605 = loc("index_554.dc.matmul.2"(#loc515)) +#loc2606 = loc("index_554.dc.transpose.3"(#loc516)) +#loc2607 = loc("multiply_555"(#loc517)) +#loc2608 = loc("index_556.dc.transpose.0"(#loc518)) +#loc2609 = loc("index_556.dc.matmul.2"(#loc519)) +#loc2610 = loc("index_556.dc.transpose.3"(#loc520)) +#loc2611 = loc("concatenate_557"(#loc521)) +#loc2612 = loc("multiply_558"(#loc522)) +#loc2613 = loc("add_559"(#loc523)) +#loc2614 = loc("reshape_560.dc.squeeze.0"(#loc524)) +#loc2615 = loc("transpose_561"(#loc525)) +#loc2616 = loc("matmul_562"(#loc526)) +#loc2617 = loc("reshape_563.dc.unsqueeze.0"(#loc527)) +#loc2618 = loc("multiply_564"(#loc528)) +#loc2619 = loc("add_565"(#loc529)) +#loc2620 = loc("softmax_566"(#loc530)) +#loc2621 = loc("reshape_568.dc.squeeze.0"(#loc531)) +#loc2622 = loc("matmul_570"(#loc532)) +#loc2623 = loc("reshape_571"(#loc533)) +#loc2624 = loc("transpose_572"(#loc534)) +#loc2625 = loc("transpose_573"(#loc535)) +#loc2626 = loc("reshape_574.dc.squeeze.0"(#loc536)) +#loc2627 = loc("transpose_575"(#loc537)) +#loc2628 = loc("matmul_576"(#loc538)) +#loc2629 = loc("reshape_577.dc.unsqueeze.0"(#loc539)) +#loc2630 = loc("transpose_578"(#loc540)) +#loc2631 = loc("reshape_579"(#loc541)) +#loc2632 = loc("matmul_581"(#loc542)) +#loc2633 = loc("reshape_582.dc.unsqueeze.0"(#loc543)) +#loc2634 = loc("add_583"(#loc544)) +#loc2635 = loc("multiply_584"(#loc545)) +#loc2636 = loc("reduce_avg_585"(#loc546)) +#loc2637 = loc("add_586"(#loc547)) +#loc2638 = loc("sqrt_587"(#loc548)) +#loc2639 = loc("reciprocal_588"(#loc549)) +#loc2640 = loc("multiply_589"(#loc550)) +#loc2641 = loc("multiply_590"(#loc551)) +#loc2642 = loc("reshape_591.dc.squeeze.0"(#loc552)) +#loc2643 = loc("matmul_593"(#loc553)) +#loc2644 = loc("reshape_594.dc.unsqueeze.0"(#loc554)) +#loc2645 = loc("sigmoid_595"(#loc555)) +#loc2646 = loc("multiply_596"(#loc556)) +#loc2647 = loc("matmul_598"(#loc557)) +#loc2648 = loc("reshape_599.dc.unsqueeze.0"(#loc558)) +#loc2649 = loc("multiply_600"(#loc559)) +#loc2650 = loc("matmul_602"(#loc560)) +#loc2651 = loc("add_603"(#loc561)) +#loc2652 = loc("multiply_604"(#loc562)) +#loc2653 = loc("reduce_avg_605"(#loc563)) +#loc2654 = loc("add_606"(#loc564)) +#loc2655 = loc("sqrt_607"(#loc565)) +#loc2656 = loc("reciprocal_608"(#loc566)) +#loc2657 = loc("multiply_609"(#loc567)) +#loc2658 = loc("multiply_610"(#loc568)) +#loc2659 = loc("reshape_611.dc.squeeze.0"(#loc569)) +#loc2660 = loc("matmul_613"(#loc570)) +#loc2661 = loc("reshape_614"(#loc571)) +#loc2662 = loc("transpose_615"(#loc572)) +#loc2663 = loc("concatenate_622"(#loc573)) +#loc2664 = loc("cosine_623"(#loc574)) +#loc2665 = loc("unsqueeze_624"(#loc575)) +#loc2666 = loc("multiply_625"(#loc576)) +#loc2667 = loc("index_626.dc.transpose.0"(#loc577)) +#loc2668 = loc("index_626.dc.matmul.2"(#loc578)) +#loc2669 = loc("index_626.dc.transpose.3"(#loc579)) +#loc2670 = loc("multiply_627"(#loc580)) +#loc2671 = loc("index_628.dc.transpose.0"(#loc581)) +#loc2672 = loc("index_628.dc.matmul.2"(#loc582)) +#loc2673 = loc("index_628.dc.transpose.3"(#loc583)) +#loc2674 = loc("concatenate_629"(#loc584)) +#loc2675 = loc("sine_630"(#loc585)) +#loc2676 = loc("unsqueeze_631"(#loc586)) +#loc2677 = loc("multiply_632"(#loc587)) +#loc2678 = loc("add_633"(#loc588)) +#loc2679 = loc("reshape_634.dc.squeeze.0"(#loc589)) +#loc2680 = loc("matmul_636"(#loc590)) +#loc2681 = loc("reshape_637"(#loc591)) +#loc2682 = loc("transpose_638"(#loc592)) +#loc2683 = loc("multiply_639"(#loc593)) +#loc2684 = loc("index_640.dc.transpose.0"(#loc594)) +#loc2685 = loc("index_640.dc.matmul.2"(#loc595)) +#loc2686 = loc("index_640.dc.transpose.3"(#loc596)) +#loc2687 = loc("multiply_641"(#loc597)) +#loc2688 = loc("index_642.dc.transpose.0"(#loc598)) +#loc2689 = loc("index_642.dc.matmul.2"(#loc599)) +#loc2690 = loc("index_642.dc.transpose.3"(#loc600)) +#loc2691 = loc("concatenate_643"(#loc601)) +#loc2692 = loc("multiply_644"(#loc602)) +#loc2693 = loc("add_645"(#loc603)) +#loc2694 = loc("reshape_646.dc.squeeze.0"(#loc604)) +#loc2695 = loc("transpose_647"(#loc605)) +#loc2696 = loc("matmul_648"(#loc606)) +#loc2697 = loc("reshape_649.dc.unsqueeze.0"(#loc607)) +#loc2698 = loc("multiply_650"(#loc608)) +#loc2699 = loc("add_651"(#loc609)) +#loc2700 = loc("softmax_652"(#loc610)) +#loc2701 = loc("reshape_654.dc.squeeze.0"(#loc611)) +#loc2702 = loc("matmul_656"(#loc612)) +#loc2703 = loc("reshape_657"(#loc613)) +#loc2704 = loc("transpose_658"(#loc614)) +#loc2705 = loc("transpose_659"(#loc615)) +#loc2706 = loc("reshape_660.dc.squeeze.0"(#loc616)) +#loc2707 = loc("transpose_661"(#loc617)) +#loc2708 = loc("matmul_662"(#loc618)) +#loc2709 = loc("reshape_663.dc.unsqueeze.0"(#loc619)) +#loc2710 = loc("transpose_664"(#loc620)) +#loc2711 = loc("reshape_665"(#loc621)) +#loc2712 = loc("matmul_667"(#loc622)) +#loc2713 = loc("reshape_668.dc.unsqueeze.0"(#loc623)) +#loc2714 = loc("add_669"(#loc624)) +#loc2715 = loc("multiply_670"(#loc625)) +#loc2716 = loc("reduce_avg_671"(#loc626)) +#loc2717 = loc("add_672"(#loc627)) +#loc2718 = loc("sqrt_673"(#loc628)) +#loc2719 = loc("reciprocal_674"(#loc629)) +#loc2720 = loc("multiply_675"(#loc630)) +#loc2721 = loc("multiply_676"(#loc631)) +#loc2722 = loc("reshape_677.dc.squeeze.0"(#loc632)) +#loc2723 = loc("matmul_679"(#loc633)) +#loc2724 = loc("reshape_680.dc.unsqueeze.0"(#loc634)) +#loc2725 = loc("sigmoid_681"(#loc635)) +#loc2726 = loc("multiply_682"(#loc636)) +#loc2727 = loc("matmul_684"(#loc637)) +#loc2728 = loc("reshape_685.dc.unsqueeze.0"(#loc638)) +#loc2729 = loc("multiply_686"(#loc639)) +#loc2730 = loc("matmul_688"(#loc640)) +#loc2731 = loc("add_689"(#loc641)) +#loc2732 = loc("multiply_690"(#loc642)) +#loc2733 = loc("reduce_avg_691"(#loc643)) +#loc2734 = loc("add_692"(#loc644)) +#loc2735 = loc("sqrt_693"(#loc645)) +#loc2736 = loc("reciprocal_694"(#loc646)) +#loc2737 = loc("multiply_695"(#loc647)) +#loc2738 = loc("multiply_696"(#loc648)) +#loc2739 = loc("reshape_697.dc.squeeze.0"(#loc649)) +#loc2740 = loc("matmul_699"(#loc650)) +#loc2741 = loc("reshape_700"(#loc651)) +#loc2742 = loc("transpose_701"(#loc652)) +#loc2743 = loc("concatenate_708"(#loc653)) +#loc2744 = loc("cosine_709"(#loc654)) +#loc2745 = loc("unsqueeze_710"(#loc655)) +#loc2746 = loc("multiply_711"(#loc656)) +#loc2747 = loc("index_712.dc.transpose.0"(#loc657)) +#loc2748 = loc("index_712.dc.matmul.2"(#loc658)) +#loc2749 = loc("index_712.dc.transpose.3"(#loc659)) +#loc2750 = loc("multiply_713"(#loc660)) +#loc2751 = loc("index_714.dc.transpose.0"(#loc661)) +#loc2752 = loc("index_714.dc.matmul.2"(#loc662)) +#loc2753 = loc("index_714.dc.transpose.3"(#loc663)) +#loc2754 = loc("concatenate_715"(#loc664)) +#loc2755 = loc("sine_716"(#loc665)) +#loc2756 = loc("unsqueeze_717"(#loc666)) +#loc2757 = loc("multiply_718"(#loc667)) +#loc2758 = loc("add_719"(#loc668)) +#loc2759 = loc("reshape_720.dc.squeeze.0"(#loc669)) +#loc2760 = loc("matmul_722"(#loc670)) +#loc2761 = loc("reshape_723"(#loc671)) +#loc2762 = loc("transpose_724"(#loc672)) +#loc2763 = loc("multiply_725"(#loc673)) +#loc2764 = loc("index_726.dc.transpose.0"(#loc674)) +#loc2765 = loc("index_726.dc.matmul.2"(#loc675)) +#loc2766 = loc("index_726.dc.transpose.3"(#loc676)) +#loc2767 = loc("multiply_727"(#loc677)) +#loc2768 = loc("index_728.dc.transpose.0"(#loc678)) +#loc2769 = loc("index_728.dc.matmul.2"(#loc679)) +#loc2770 = loc("index_728.dc.transpose.3"(#loc680)) +#loc2771 = loc("concatenate_729"(#loc681)) +#loc2772 = loc("multiply_730"(#loc682)) +#loc2773 = loc("add_731"(#loc683)) +#loc2774 = loc("reshape_732.dc.squeeze.0"(#loc684)) +#loc2775 = loc("transpose_733"(#loc685)) +#loc2776 = loc("matmul_734"(#loc686)) +#loc2777 = loc("reshape_735.dc.unsqueeze.0"(#loc687)) +#loc2778 = loc("multiply_736"(#loc688)) +#loc2779 = loc("add_737"(#loc689)) +#loc2780 = loc("softmax_738"(#loc690)) +#loc2781 = loc("reshape_740.dc.squeeze.0"(#loc691)) +#loc2782 = loc("matmul_742"(#loc692)) +#loc2783 = loc("reshape_743"(#loc693)) +#loc2784 = loc("transpose_744"(#loc694)) +#loc2785 = loc("transpose_745"(#loc695)) +#loc2786 = loc("reshape_746.dc.squeeze.0"(#loc696)) +#loc2787 = loc("transpose_747"(#loc697)) +#loc2788 = loc("matmul_748"(#loc698)) +#loc2789 = loc("reshape_749.dc.unsqueeze.0"(#loc699)) +#loc2790 = loc("transpose_750"(#loc700)) +#loc2791 = loc("reshape_751"(#loc701)) +#loc2792 = loc("matmul_753"(#loc702)) +#loc2793 = loc("reshape_754.dc.unsqueeze.0"(#loc703)) +#loc2794 = loc("add_755"(#loc704)) +#loc2795 = loc("multiply_756"(#loc705)) +#loc2796 = loc("reduce_avg_757"(#loc706)) +#loc2797 = loc("add_758"(#loc707)) +#loc2798 = loc("sqrt_759"(#loc708)) +#loc2799 = loc("reciprocal_760"(#loc709)) +#loc2800 = loc("multiply_761"(#loc710)) +#loc2801 = loc("multiply_762"(#loc711)) +#loc2802 = loc("reshape_763.dc.squeeze.0"(#loc712)) +#loc2803 = loc("matmul_765"(#loc713)) +#loc2804 = loc("reshape_766.dc.unsqueeze.0"(#loc714)) +#loc2805 = loc("sigmoid_767"(#loc715)) +#loc2806 = loc("multiply_768"(#loc716)) +#loc2807 = loc("matmul_770"(#loc717)) +#loc2808 = loc("reshape_771.dc.unsqueeze.0"(#loc718)) +#loc2809 = loc("multiply_772"(#loc719)) +#loc2810 = loc("matmul_774"(#loc720)) +#loc2811 = loc("add_775"(#loc721)) +#loc2812 = loc("multiply_776"(#loc722)) +#loc2813 = loc("reduce_avg_777"(#loc723)) +#loc2814 = loc("add_778"(#loc724)) +#loc2815 = loc("sqrt_779"(#loc725)) +#loc2816 = loc("reciprocal_780"(#loc726)) +#loc2817 = loc("multiply_781"(#loc727)) +#loc2818 = loc("multiply_782"(#loc728)) +#loc2819 = loc("reshape_783.dc.squeeze.0"(#loc729)) +#loc2820 = loc("matmul_785"(#loc730)) +#loc2821 = loc("reshape_786"(#loc731)) +#loc2822 = loc("transpose_787"(#loc732)) +#loc2823 = loc("concatenate_794"(#loc733)) +#loc2824 = loc("cosine_795"(#loc734)) +#loc2825 = loc("unsqueeze_796"(#loc735)) +#loc2826 = loc("multiply_797"(#loc736)) +#loc2827 = loc("index_798.dc.transpose.0"(#loc737)) +#loc2828 = loc("index_798.dc.matmul.2"(#loc738)) +#loc2829 = loc("index_798.dc.transpose.3"(#loc739)) +#loc2830 = loc("multiply_799"(#loc740)) +#loc2831 = loc("index_800.dc.transpose.0"(#loc741)) +#loc2832 = loc("index_800.dc.matmul.2"(#loc742)) +#loc2833 = loc("index_800.dc.transpose.3"(#loc743)) +#loc2834 = loc("concatenate_801"(#loc744)) +#loc2835 = loc("sine_802"(#loc745)) +#loc2836 = loc("unsqueeze_803"(#loc746)) +#loc2837 = loc("multiply_804"(#loc747)) +#loc2838 = loc("add_805"(#loc748)) +#loc2839 = loc("reshape_806.dc.squeeze.0"(#loc749)) +#loc2840 = loc("matmul_808"(#loc750)) +#loc2841 = loc("reshape_809"(#loc751)) +#loc2842 = loc("transpose_810"(#loc752)) +#loc2843 = loc("multiply_811"(#loc753)) +#loc2844 = loc("index_812.dc.transpose.0"(#loc754)) +#loc2845 = loc("index_812.dc.matmul.2"(#loc755)) +#loc2846 = loc("index_812.dc.transpose.3"(#loc756)) +#loc2847 = loc("multiply_813"(#loc757)) +#loc2848 = loc("index_814.dc.transpose.0"(#loc758)) +#loc2849 = loc("index_814.dc.matmul.2"(#loc759)) +#loc2850 = loc("index_814.dc.transpose.3"(#loc760)) +#loc2851 = loc("concatenate_815"(#loc761)) +#loc2852 = loc("multiply_816"(#loc762)) +#loc2853 = loc("add_817"(#loc763)) +#loc2854 = loc("reshape_818.dc.squeeze.0"(#loc764)) +#loc2855 = loc("transpose_819"(#loc765)) +#loc2856 = loc("matmul_820"(#loc766)) +#loc2857 = loc("reshape_821.dc.unsqueeze.0"(#loc767)) +#loc2858 = loc("multiply_822"(#loc768)) +#loc2859 = loc("add_823"(#loc769)) +#loc2860 = loc("softmax_824"(#loc770)) +#loc2861 = loc("reshape_826.dc.squeeze.0"(#loc771)) +#loc2862 = loc("matmul_828"(#loc772)) +#loc2863 = loc("reshape_829"(#loc773)) +#loc2864 = loc("transpose_830"(#loc774)) +#loc2865 = loc("transpose_831"(#loc775)) +#loc2866 = loc("reshape_832.dc.squeeze.0"(#loc776)) +#loc2867 = loc("transpose_833"(#loc777)) +#loc2868 = loc("matmul_834"(#loc778)) +#loc2869 = loc("reshape_835.dc.unsqueeze.0"(#loc779)) +#loc2870 = loc("transpose_836"(#loc780)) +#loc2871 = loc("reshape_837"(#loc781)) +#loc2872 = loc("matmul_839"(#loc782)) +#loc2873 = loc("reshape_840.dc.unsqueeze.0"(#loc783)) +#loc2874 = loc("add_841"(#loc784)) +#loc2875 = loc("multiply_842"(#loc785)) +#loc2876 = loc("reduce_avg_843"(#loc786)) +#loc2877 = loc("add_844"(#loc787)) +#loc2878 = loc("sqrt_845"(#loc788)) +#loc2879 = loc("reciprocal_846"(#loc789)) +#loc2880 = loc("multiply_847"(#loc790)) +#loc2881 = loc("multiply_848"(#loc791)) +#loc2882 = loc("reshape_849.dc.squeeze.0"(#loc792)) +#loc2883 = loc("matmul_851"(#loc793)) +#loc2884 = loc("reshape_852.dc.unsqueeze.0"(#loc794)) +#loc2885 = loc("sigmoid_853"(#loc795)) +#loc2886 = loc("multiply_854"(#loc796)) +#loc2887 = loc("matmul_856"(#loc797)) +#loc2888 = loc("reshape_857.dc.unsqueeze.0"(#loc798)) +#loc2889 = loc("multiply_858"(#loc799)) +#loc2890 = loc("matmul_860"(#loc800)) +#loc2891 = loc("add_861"(#loc801)) +#loc2892 = loc("multiply_862"(#loc802)) +#loc2893 = loc("reduce_avg_863"(#loc803)) +#loc2894 = loc("add_864"(#loc804)) +#loc2895 = loc("sqrt_865"(#loc805)) +#loc2896 = loc("reciprocal_866"(#loc806)) +#loc2897 = loc("multiply_867"(#loc807)) +#loc2898 = loc("multiply_868"(#loc808)) +#loc2899 = loc("reshape_869.dc.squeeze.0"(#loc809)) +#loc2900 = loc("matmul_871"(#loc810)) +#loc2901 = loc("reshape_872"(#loc811)) +#loc2902 = loc("transpose_873"(#loc812)) +#loc2903 = loc("concatenate_880"(#loc813)) +#loc2904 = loc("cosine_881"(#loc814)) +#loc2905 = loc("unsqueeze_882"(#loc815)) +#loc2906 = loc("multiply_883"(#loc816)) +#loc2907 = loc("index_884.dc.transpose.0"(#loc817)) +#loc2908 = loc("index_884.dc.matmul.2"(#loc818)) +#loc2909 = loc("index_884.dc.transpose.3"(#loc819)) +#loc2910 = loc("multiply_885"(#loc820)) +#loc2911 = loc("index_886.dc.transpose.0"(#loc821)) +#loc2912 = loc("index_886.dc.matmul.2"(#loc822)) +#loc2913 = loc("index_886.dc.transpose.3"(#loc823)) +#loc2914 = loc("concatenate_887"(#loc824)) +#loc2915 = loc("sine_888"(#loc825)) +#loc2916 = loc("unsqueeze_889"(#loc826)) +#loc2917 = loc("multiply_890"(#loc827)) +#loc2918 = loc("add_891"(#loc828)) +#loc2919 = loc("reshape_892.dc.squeeze.0"(#loc829)) +#loc2920 = loc("matmul_894"(#loc830)) +#loc2921 = loc("reshape_895"(#loc831)) +#loc2922 = loc("transpose_896"(#loc832)) +#loc2923 = loc("multiply_897"(#loc833)) +#loc2924 = loc("index_898.dc.transpose.0"(#loc834)) +#loc2925 = loc("index_898.dc.matmul.2"(#loc835)) +#loc2926 = loc("index_898.dc.transpose.3"(#loc836)) +#loc2927 = loc("multiply_899"(#loc837)) +#loc2928 = loc("index_900.dc.transpose.0"(#loc838)) +#loc2929 = loc("index_900.dc.matmul.2"(#loc839)) +#loc2930 = loc("index_900.dc.transpose.3"(#loc840)) +#loc2931 = loc("concatenate_901"(#loc841)) +#loc2932 = loc("multiply_902"(#loc842)) +#loc2933 = loc("add_903"(#loc843)) +#loc2934 = loc("reshape_904.dc.squeeze.0"(#loc844)) +#loc2935 = loc("transpose_905"(#loc845)) +#loc2936 = loc("matmul_906"(#loc846)) +#loc2937 = loc("reshape_907.dc.unsqueeze.0"(#loc847)) +#loc2938 = loc("multiply_908"(#loc848)) +#loc2939 = loc("add_909"(#loc849)) +#loc2940 = loc("softmax_910"(#loc850)) +#loc2941 = loc("reshape_912.dc.squeeze.0"(#loc851)) +#loc2942 = loc("matmul_914"(#loc852)) +#loc2943 = loc("reshape_915"(#loc853)) +#loc2944 = loc("transpose_916"(#loc854)) +#loc2945 = loc("transpose_917"(#loc855)) +#loc2946 = loc("reshape_918.dc.squeeze.0"(#loc856)) +#loc2947 = loc("transpose_919"(#loc857)) +#loc2948 = loc("matmul_920"(#loc858)) +#loc2949 = loc("reshape_921.dc.unsqueeze.0"(#loc859)) +#loc2950 = loc("transpose_922"(#loc860)) +#loc2951 = loc("reshape_923"(#loc861)) +#loc2952 = loc("matmul_925"(#loc862)) +#loc2953 = loc("reshape_926.dc.unsqueeze.0"(#loc863)) +#loc2954 = loc("add_927"(#loc864)) +#loc2955 = loc("multiply_928"(#loc865)) +#loc2956 = loc("reduce_avg_929"(#loc866)) +#loc2957 = loc("add_930"(#loc867)) +#loc2958 = loc("sqrt_931"(#loc868)) +#loc2959 = loc("reciprocal_932"(#loc869)) +#loc2960 = loc("multiply_933"(#loc870)) +#loc2961 = loc("multiply_934"(#loc871)) +#loc2962 = loc("reshape_935.dc.squeeze.0"(#loc872)) +#loc2963 = loc("matmul_937"(#loc873)) +#loc2964 = loc("reshape_938.dc.unsqueeze.0"(#loc874)) +#loc2965 = loc("sigmoid_939"(#loc875)) +#loc2966 = loc("multiply_940"(#loc876)) +#loc2967 = loc("matmul_942"(#loc877)) +#loc2968 = loc("reshape_943.dc.unsqueeze.0"(#loc878)) +#loc2969 = loc("multiply_944"(#loc879)) +#loc2970 = loc("matmul_946"(#loc880)) +#loc2971 = loc("add_947"(#loc881)) +#loc2972 = loc("multiply_948"(#loc882)) +#loc2973 = loc("reduce_avg_949"(#loc883)) +#loc2974 = loc("add_950"(#loc884)) +#loc2975 = loc("sqrt_951"(#loc885)) +#loc2976 = loc("reciprocal_952"(#loc886)) +#loc2977 = loc("multiply_953"(#loc887)) +#loc2978 = loc("multiply_954"(#loc888)) +#loc2979 = loc("reshape_955.dc.squeeze.0"(#loc889)) +#loc2980 = loc("matmul_957"(#loc890)) +#loc2981 = loc("reshape_958"(#loc891)) +#loc2982 = loc("transpose_959"(#loc892)) +#loc2983 = loc("concatenate_966"(#loc893)) +#loc2984 = loc("cosine_967"(#loc894)) +#loc2985 = loc("unsqueeze_968"(#loc895)) +#loc2986 = loc("multiply_969"(#loc896)) +#loc2987 = loc("index_970.dc.transpose.0"(#loc897)) +#loc2988 = loc("index_970.dc.matmul.2"(#loc898)) +#loc2989 = loc("index_970.dc.transpose.3"(#loc899)) +#loc2990 = loc("multiply_971"(#loc900)) +#loc2991 = loc("index_972.dc.transpose.0"(#loc901)) +#loc2992 = loc("index_972.dc.matmul.2"(#loc902)) +#loc2993 = loc("index_972.dc.transpose.3"(#loc903)) +#loc2994 = loc("concatenate_973"(#loc904)) +#loc2995 = loc("sine_974"(#loc905)) +#loc2996 = loc("unsqueeze_975"(#loc906)) +#loc2997 = loc("multiply_976"(#loc907)) +#loc2998 = loc("add_977"(#loc908)) +#loc2999 = loc("reshape_978.dc.squeeze.0"(#loc909)) +#loc3000 = loc("matmul_980"(#loc910)) +#loc3001 = loc("reshape_981"(#loc911)) +#loc3002 = loc("transpose_982"(#loc912)) +#loc3003 = loc("multiply_983"(#loc913)) +#loc3004 = loc("index_984.dc.transpose.0"(#loc914)) +#loc3005 = loc("index_984.dc.matmul.2"(#loc915)) +#loc3006 = loc("index_984.dc.transpose.3"(#loc916)) +#loc3007 = loc("multiply_985"(#loc917)) +#loc3008 = loc("index_986.dc.transpose.0"(#loc918)) +#loc3009 = loc("index_986.dc.matmul.2"(#loc919)) +#loc3010 = loc("index_986.dc.transpose.3"(#loc920)) +#loc3011 = loc("concatenate_987"(#loc921)) +#loc3012 = loc("multiply_988"(#loc922)) +#loc3013 = loc("add_989"(#loc923)) +#loc3014 = loc("reshape_990.dc.squeeze.0"(#loc924)) +#loc3015 = loc("transpose_991"(#loc925)) +#loc3016 = loc("matmul_992"(#loc926)) +#loc3017 = loc("reshape_993.dc.unsqueeze.0"(#loc927)) +#loc3018 = loc("multiply_994"(#loc928)) +#loc3019 = loc("add_995"(#loc929)) +#loc3020 = loc("softmax_996"(#loc930)) +#loc3021 = loc("reshape_998.dc.squeeze.0"(#loc931)) +#loc3022 = loc("matmul_1000"(#loc932)) +#loc3023 = loc("reshape_1001"(#loc933)) +#loc3024 = loc("transpose_1002"(#loc934)) +#loc3025 = loc("transpose_1003"(#loc935)) +#loc3026 = loc("reshape_1004.dc.squeeze.0"(#loc936)) +#loc3027 = loc("transpose_1005"(#loc937)) +#loc3028 = loc("matmul_1006"(#loc938)) +#loc3029 = loc("reshape_1007.dc.unsqueeze.0"(#loc939)) +#loc3030 = loc("transpose_1008"(#loc940)) +#loc3031 = loc("reshape_1009"(#loc941)) +#loc3032 = loc("matmul_1011"(#loc942)) +#loc3033 = loc("reshape_1012.dc.unsqueeze.0"(#loc943)) +#loc3034 = loc("add_1013"(#loc944)) +#loc3035 = loc("multiply_1014"(#loc945)) +#loc3036 = loc("reduce_avg_1015"(#loc946)) +#loc3037 = loc("add_1016"(#loc947)) +#loc3038 = loc("sqrt_1017"(#loc948)) +#loc3039 = loc("reciprocal_1018"(#loc949)) +#loc3040 = loc("multiply_1019"(#loc950)) +#loc3041 = loc("multiply_1020"(#loc951)) +#loc3042 = loc("reshape_1021.dc.squeeze.0"(#loc952)) +#loc3043 = loc("matmul_1023"(#loc953)) +#loc3044 = loc("reshape_1024.dc.unsqueeze.0"(#loc954)) +#loc3045 = loc("sigmoid_1025"(#loc955)) +#loc3046 = loc("multiply_1026"(#loc956)) +#loc3047 = loc("matmul_1028"(#loc957)) +#loc3048 = loc("reshape_1029.dc.unsqueeze.0"(#loc958)) +#loc3049 = loc("multiply_1030"(#loc959)) +#loc3050 = loc("matmul_1032"(#loc960)) +#loc3051 = loc("add_1033"(#loc961)) +#loc3052 = loc("multiply_1034"(#loc962)) +#loc3053 = loc("reduce_avg_1035"(#loc963)) +#loc3054 = loc("add_1036"(#loc964)) +#loc3055 = loc("sqrt_1037"(#loc965)) +#loc3056 = loc("reciprocal_1038"(#loc966)) +#loc3057 = loc("multiply_1039"(#loc967)) +#loc3058 = loc("multiply_1040"(#loc968)) +#loc3059 = loc("reshape_1041.dc.squeeze.0"(#loc969)) +#loc3060 = loc("matmul_1043"(#loc970)) +#loc3061 = loc("reshape_1044"(#loc971)) +#loc3062 = loc("transpose_1045"(#loc972)) +#loc3063 = loc("concatenate_1052"(#loc973)) +#loc3064 = loc("cosine_1053"(#loc974)) +#loc3065 = loc("unsqueeze_1054"(#loc975)) +#loc3066 = loc("multiply_1055"(#loc976)) +#loc3067 = loc("index_1056.dc.transpose.0"(#loc977)) +#loc3068 = loc("index_1056.dc.matmul.2"(#loc978)) +#loc3069 = loc("index_1056.dc.transpose.3"(#loc979)) +#loc3070 = loc("multiply_1057"(#loc980)) +#loc3071 = loc("index_1058.dc.transpose.0"(#loc981)) +#loc3072 = loc("index_1058.dc.matmul.2"(#loc982)) +#loc3073 = loc("index_1058.dc.transpose.3"(#loc983)) +#loc3074 = loc("concatenate_1059"(#loc984)) +#loc3075 = loc("sine_1060"(#loc985)) +#loc3076 = loc("unsqueeze_1061"(#loc986)) +#loc3077 = loc("multiply_1062"(#loc987)) +#loc3078 = loc("add_1063"(#loc988)) +#loc3079 = loc("reshape_1064.dc.squeeze.0"(#loc989)) +#loc3080 = loc("matmul_1066"(#loc990)) +#loc3081 = loc("reshape_1067"(#loc991)) +#loc3082 = loc("transpose_1068"(#loc992)) +#loc3083 = loc("multiply_1069"(#loc993)) +#loc3084 = loc("index_1070.dc.transpose.0"(#loc994)) +#loc3085 = loc("index_1070.dc.matmul.2"(#loc995)) +#loc3086 = loc("index_1070.dc.transpose.3"(#loc996)) +#loc3087 = loc("multiply_1071"(#loc997)) +#loc3088 = loc("index_1072.dc.transpose.0"(#loc998)) +#loc3089 = loc("index_1072.dc.matmul.2"(#loc999)) +#loc3090 = loc("index_1072.dc.transpose.3"(#loc1000)) +#loc3091 = loc("concatenate_1073"(#loc1001)) +#loc3092 = loc("multiply_1074"(#loc1002)) +#loc3093 = loc("add_1075"(#loc1003)) +#loc3094 = loc("reshape_1076.dc.squeeze.0"(#loc1004)) +#loc3095 = loc("transpose_1077"(#loc1005)) +#loc3096 = loc("matmul_1078"(#loc1006)) +#loc3097 = loc("reshape_1079.dc.unsqueeze.0"(#loc1007)) +#loc3098 = loc("multiply_1080"(#loc1008)) +#loc3099 = loc("add_1081"(#loc1009)) +#loc3100 = loc("softmax_1082"(#loc1010)) +#loc3101 = loc("reshape_1084.dc.squeeze.0"(#loc1011)) +#loc3102 = loc("matmul_1086"(#loc1012)) +#loc3103 = loc("reshape_1087"(#loc1013)) +#loc3104 = loc("transpose_1088"(#loc1014)) +#loc3105 = loc("transpose_1089"(#loc1015)) +#loc3106 = loc("reshape_1090.dc.squeeze.0"(#loc1016)) +#loc3107 = loc("transpose_1091"(#loc1017)) +#loc3108 = loc("matmul_1092"(#loc1018)) +#loc3109 = loc("reshape_1093.dc.unsqueeze.0"(#loc1019)) +#loc3110 = loc("transpose_1094"(#loc1020)) +#loc3111 = loc("reshape_1095"(#loc1021)) +#loc3112 = loc("matmul_1097"(#loc1022)) +#loc3113 = loc("reshape_1098.dc.unsqueeze.0"(#loc1023)) +#loc3114 = loc("add_1099"(#loc1024)) +#loc3115 = loc("multiply_1100"(#loc1025)) +#loc3116 = loc("reduce_avg_1101"(#loc1026)) +#loc3117 = loc("add_1102"(#loc1027)) +#loc3118 = loc("sqrt_1103"(#loc1028)) +#loc3119 = loc("reciprocal_1104"(#loc1029)) +#loc3120 = loc("multiply_1105"(#loc1030)) +#loc3121 = loc("multiply_1106"(#loc1031)) +#loc3122 = loc("reshape_1107.dc.squeeze.0"(#loc1032)) +#loc3123 = loc("matmul_1109"(#loc1033)) +#loc3124 = loc("reshape_1110.dc.unsqueeze.0"(#loc1034)) +#loc3125 = loc("sigmoid_1111"(#loc1035)) +#loc3126 = loc("multiply_1112"(#loc1036)) +#loc3127 = loc("matmul_1114"(#loc1037)) +#loc3128 = loc("reshape_1115.dc.unsqueeze.0"(#loc1038)) +#loc3129 = loc("multiply_1116"(#loc1039)) +#loc3130 = loc("matmul_1118"(#loc1040)) +#loc3131 = loc("add_1119"(#loc1041)) +#loc3132 = loc("multiply_1120"(#loc1042)) +#loc3133 = loc("reduce_avg_1121"(#loc1043)) +#loc3134 = loc("add_1122"(#loc1044)) +#loc3135 = loc("sqrt_1123"(#loc1045)) +#loc3136 = loc("reciprocal_1124"(#loc1046)) +#loc3137 = loc("multiply_1125"(#loc1047)) +#loc3138 = loc("multiply_1126"(#loc1048)) +#loc3139 = loc("reshape_1127.dc.squeeze.0"(#loc1049)) +#loc3140 = loc("matmul_1129"(#loc1050)) +#loc3141 = loc("reshape_1130"(#loc1051)) +#loc3142 = loc("transpose_1131"(#loc1052)) +#loc3143 = loc("concatenate_1138"(#loc1053)) +#loc3144 = loc("cosine_1139"(#loc1054)) +#loc3145 = loc("unsqueeze_1140"(#loc1055)) +#loc3146 = loc("multiply_1141"(#loc1056)) +#loc3147 = loc("index_1142.dc.transpose.0"(#loc1057)) +#loc3148 = loc("index_1142.dc.matmul.2"(#loc1058)) +#loc3149 = loc("index_1142.dc.transpose.3"(#loc1059)) +#loc3150 = loc("multiply_1143"(#loc1060)) +#loc3151 = loc("index_1144.dc.transpose.0"(#loc1061)) +#loc3152 = loc("index_1144.dc.matmul.2"(#loc1062)) +#loc3153 = loc("index_1144.dc.transpose.3"(#loc1063)) +#loc3154 = loc("concatenate_1145"(#loc1064)) +#loc3155 = loc("sine_1146"(#loc1065)) +#loc3156 = loc("unsqueeze_1147"(#loc1066)) +#loc3157 = loc("multiply_1148"(#loc1067)) +#loc3158 = loc("add_1149"(#loc1068)) +#loc3159 = loc("reshape_1150.dc.squeeze.0"(#loc1069)) +#loc3160 = loc("matmul_1152"(#loc1070)) +#loc3161 = loc("reshape_1153"(#loc1071)) +#loc3162 = loc("transpose_1154"(#loc1072)) +#loc3163 = loc("multiply_1155"(#loc1073)) +#loc3164 = loc("index_1156.dc.transpose.0"(#loc1074)) +#loc3165 = loc("index_1156.dc.matmul.2"(#loc1075)) +#loc3166 = loc("index_1156.dc.transpose.3"(#loc1076)) +#loc3167 = loc("multiply_1157"(#loc1077)) +#loc3168 = loc("index_1158.dc.transpose.0"(#loc1078)) +#loc3169 = loc("index_1158.dc.matmul.2"(#loc1079)) +#loc3170 = loc("index_1158.dc.transpose.3"(#loc1080)) +#loc3171 = loc("concatenate_1159"(#loc1081)) +#loc3172 = loc("multiply_1160"(#loc1082)) +#loc3173 = loc("add_1161"(#loc1083)) +#loc3174 = loc("reshape_1162.dc.squeeze.0"(#loc1084)) +#loc3175 = loc("transpose_1163"(#loc1085)) +#loc3176 = loc("matmul_1164"(#loc1086)) +#loc3177 = loc("reshape_1165.dc.unsqueeze.0"(#loc1087)) +#loc3178 = loc("multiply_1166"(#loc1088)) +#loc3179 = loc("add_1167"(#loc1089)) +#loc3180 = loc("softmax_1168"(#loc1090)) +#loc3181 = loc("reshape_1170.dc.squeeze.0"(#loc1091)) +#loc3182 = loc("matmul_1172"(#loc1092)) +#loc3183 = loc("reshape_1173"(#loc1093)) +#loc3184 = loc("transpose_1174"(#loc1094)) +#loc3185 = loc("transpose_1175"(#loc1095)) +#loc3186 = loc("reshape_1176.dc.squeeze.0"(#loc1096)) +#loc3187 = loc("transpose_1177"(#loc1097)) +#loc3188 = loc("matmul_1178"(#loc1098)) +#loc3189 = loc("reshape_1179.dc.unsqueeze.0"(#loc1099)) +#loc3190 = loc("transpose_1180"(#loc1100)) +#loc3191 = loc("reshape_1181"(#loc1101)) +#loc3192 = loc("matmul_1183"(#loc1102)) +#loc3193 = loc("reshape_1184.dc.unsqueeze.0"(#loc1103)) +#loc3194 = loc("add_1185"(#loc1104)) +#loc3195 = loc("multiply_1186"(#loc1105)) +#loc3196 = loc("reduce_avg_1187"(#loc1106)) +#loc3197 = loc("add_1188"(#loc1107)) +#loc3198 = loc("sqrt_1189"(#loc1108)) +#loc3199 = loc("reciprocal_1190"(#loc1109)) +#loc3200 = loc("multiply_1191"(#loc1110)) +#loc3201 = loc("multiply_1192"(#loc1111)) +#loc3202 = loc("reshape_1193.dc.squeeze.0"(#loc1112)) +#loc3203 = loc("matmul_1195"(#loc1113)) +#loc3204 = loc("reshape_1196.dc.unsqueeze.0"(#loc1114)) +#loc3205 = loc("sigmoid_1197"(#loc1115)) +#loc3206 = loc("multiply_1198"(#loc1116)) +#loc3207 = loc("matmul_1200"(#loc1117)) +#loc3208 = loc("reshape_1201.dc.unsqueeze.0"(#loc1118)) +#loc3209 = loc("multiply_1202"(#loc1119)) +#loc3210 = loc("matmul_1204"(#loc1120)) +#loc3211 = loc("add_1205"(#loc1121)) +#loc3212 = loc("multiply_1206"(#loc1122)) +#loc3213 = loc("reduce_avg_1207"(#loc1123)) +#loc3214 = loc("add_1208"(#loc1124)) +#loc3215 = loc("sqrt_1209"(#loc1125)) +#loc3216 = loc("reciprocal_1210"(#loc1126)) +#loc3217 = loc("multiply_1211"(#loc1127)) +#loc3218 = loc("multiply_1212"(#loc1128)) +#loc3219 = loc("reshape_1213.dc.squeeze.0"(#loc1129)) +#loc3220 = loc("matmul_1215"(#loc1130)) +#loc3221 = loc("reshape_1216"(#loc1131)) +#loc3222 = loc("transpose_1217"(#loc1132)) +#loc3223 = loc("concatenate_1224"(#loc1133)) +#loc3224 = loc("cosine_1225"(#loc1134)) +#loc3225 = loc("unsqueeze_1226"(#loc1135)) +#loc3226 = loc("multiply_1227"(#loc1136)) +#loc3227 = loc("index_1228.dc.transpose.0"(#loc1137)) +#loc3228 = loc("index_1228.dc.matmul.2"(#loc1138)) +#loc3229 = loc("index_1228.dc.transpose.3"(#loc1139)) +#loc3230 = loc("multiply_1229"(#loc1140)) +#loc3231 = loc("index_1230.dc.transpose.0"(#loc1141)) +#loc3232 = loc("index_1230.dc.matmul.2"(#loc1142)) +#loc3233 = loc("index_1230.dc.transpose.3"(#loc1143)) +#loc3234 = loc("concatenate_1231"(#loc1144)) +#loc3235 = loc("sine_1232"(#loc1145)) +#loc3236 = loc("unsqueeze_1233"(#loc1146)) +#loc3237 = loc("multiply_1234"(#loc1147)) +#loc3238 = loc("add_1235"(#loc1148)) +#loc3239 = loc("reshape_1236.dc.squeeze.0"(#loc1149)) +#loc3240 = loc("matmul_1238"(#loc1150)) +#loc3241 = loc("reshape_1239"(#loc1151)) +#loc3242 = loc("transpose_1240"(#loc1152)) +#loc3243 = loc("multiply_1241"(#loc1153)) +#loc3244 = loc("index_1242.dc.transpose.0"(#loc1154)) +#loc3245 = loc("index_1242.dc.matmul.2"(#loc1155)) +#loc3246 = loc("index_1242.dc.transpose.3"(#loc1156)) +#loc3247 = loc("multiply_1243"(#loc1157)) +#loc3248 = loc("index_1244.dc.transpose.0"(#loc1158)) +#loc3249 = loc("index_1244.dc.matmul.2"(#loc1159)) +#loc3250 = loc("index_1244.dc.transpose.3"(#loc1160)) +#loc3251 = loc("concatenate_1245"(#loc1161)) +#loc3252 = loc("multiply_1246"(#loc1162)) +#loc3253 = loc("add_1247"(#loc1163)) +#loc3254 = loc("reshape_1248.dc.squeeze.0"(#loc1164)) +#loc3255 = loc("transpose_1249"(#loc1165)) +#loc3256 = loc("matmul_1250"(#loc1166)) +#loc3257 = loc("reshape_1251.dc.unsqueeze.0"(#loc1167)) +#loc3258 = loc("multiply_1252"(#loc1168)) +#loc3259 = loc("add_1253"(#loc1169)) +#loc3260 = loc("softmax_1254"(#loc1170)) +#loc3261 = loc("reshape_1256.dc.squeeze.0"(#loc1171)) +#loc3262 = loc("matmul_1258"(#loc1172)) +#loc3263 = loc("reshape_1259"(#loc1173)) +#loc3264 = loc("transpose_1260"(#loc1174)) +#loc3265 = loc("transpose_1261"(#loc1175)) +#loc3266 = loc("reshape_1262.dc.squeeze.0"(#loc1176)) +#loc3267 = loc("transpose_1263"(#loc1177)) +#loc3268 = loc("matmul_1264"(#loc1178)) +#loc3269 = loc("reshape_1265.dc.unsqueeze.0"(#loc1179)) +#loc3270 = loc("transpose_1266"(#loc1180)) +#loc3271 = loc("reshape_1267"(#loc1181)) +#loc3272 = loc("matmul_1269"(#loc1182)) +#loc3273 = loc("reshape_1270.dc.unsqueeze.0"(#loc1183)) +#loc3274 = loc("add_1271"(#loc1184)) +#loc3275 = loc("multiply_1272"(#loc1185)) +#loc3276 = loc("reduce_avg_1273"(#loc1186)) +#loc3277 = loc("add_1274"(#loc1187)) +#loc3278 = loc("sqrt_1275"(#loc1188)) +#loc3279 = loc("reciprocal_1276"(#loc1189)) +#loc3280 = loc("multiply_1277"(#loc1190)) +#loc3281 = loc("multiply_1278"(#loc1191)) +#loc3282 = loc("reshape_1279.dc.squeeze.0"(#loc1192)) +#loc3283 = loc("matmul_1281"(#loc1193)) +#loc3284 = loc("reshape_1282.dc.unsqueeze.0"(#loc1194)) +#loc3285 = loc("sigmoid_1283"(#loc1195)) +#loc3286 = loc("multiply_1284"(#loc1196)) +#loc3287 = loc("matmul_1286"(#loc1197)) +#loc3288 = loc("reshape_1287.dc.unsqueeze.0"(#loc1198)) +#loc3289 = loc("multiply_1288"(#loc1199)) +#loc3290 = loc("matmul_1290"(#loc1200)) +#loc3291 = loc("add_1291"(#loc1201)) +#loc3292 = loc("multiply_1292"(#loc1202)) +#loc3293 = loc("reduce_avg_1293"(#loc1203)) +#loc3294 = loc("add_1294"(#loc1204)) +#loc3295 = loc("sqrt_1295"(#loc1205)) +#loc3296 = loc("reciprocal_1296"(#loc1206)) +#loc3297 = loc("multiply_1297"(#loc1207)) +#loc3298 = loc("multiply_1298"(#loc1208)) +#loc3299 = loc("reshape_1299.dc.squeeze.0"(#loc1209)) +#loc3300 = loc("matmul_1301"(#loc1210)) +#loc3301 = loc("reshape_1302"(#loc1211)) +#loc3302 = loc("transpose_1303"(#loc1212)) +#loc3303 = loc("concatenate_1310"(#loc1213)) +#loc3304 = loc("cosine_1311"(#loc1214)) +#loc3305 = loc("unsqueeze_1312"(#loc1215)) +#loc3306 = loc("multiply_1313"(#loc1216)) +#loc3307 = loc("index_1314.dc.transpose.0"(#loc1217)) +#loc3308 = loc("index_1314.dc.matmul.2"(#loc1218)) +#loc3309 = loc("index_1314.dc.transpose.3"(#loc1219)) +#loc3310 = loc("multiply_1315"(#loc1220)) +#loc3311 = loc("index_1316.dc.transpose.0"(#loc1221)) +#loc3312 = loc("index_1316.dc.matmul.2"(#loc1222)) +#loc3313 = loc("index_1316.dc.transpose.3"(#loc1223)) +#loc3314 = loc("concatenate_1317"(#loc1224)) +#loc3315 = loc("sine_1318"(#loc1225)) +#loc3316 = loc("unsqueeze_1319"(#loc1226)) +#loc3317 = loc("multiply_1320"(#loc1227)) +#loc3318 = loc("add_1321"(#loc1228)) +#loc3319 = loc("reshape_1322.dc.squeeze.0"(#loc1229)) +#loc3320 = loc("matmul_1324"(#loc1230)) +#loc3321 = loc("reshape_1325"(#loc1231)) +#loc3322 = loc("transpose_1326"(#loc1232)) +#loc3323 = loc("multiply_1327"(#loc1233)) +#loc3324 = loc("index_1328.dc.transpose.0"(#loc1234)) +#loc3325 = loc("index_1328.dc.matmul.2"(#loc1235)) +#loc3326 = loc("index_1328.dc.transpose.3"(#loc1236)) +#loc3327 = loc("multiply_1329"(#loc1237)) +#loc3328 = loc("index_1330.dc.transpose.0"(#loc1238)) +#loc3329 = loc("index_1330.dc.matmul.2"(#loc1239)) +#loc3330 = loc("index_1330.dc.transpose.3"(#loc1240)) +#loc3331 = loc("concatenate_1331"(#loc1241)) +#loc3332 = loc("multiply_1332"(#loc1242)) +#loc3333 = loc("add_1333"(#loc1243)) +#loc3334 = loc("reshape_1334.dc.squeeze.0"(#loc1244)) +#loc3335 = loc("transpose_1335"(#loc1245)) +#loc3336 = loc("matmul_1336"(#loc1246)) +#loc3337 = loc("reshape_1337.dc.unsqueeze.0"(#loc1247)) +#loc3338 = loc("multiply_1338"(#loc1248)) +#loc3339 = loc("add_1339"(#loc1249)) +#loc3340 = loc("softmax_1340"(#loc1250)) +#loc3341 = loc("reshape_1342.dc.squeeze.0"(#loc1251)) +#loc3342 = loc("matmul_1344"(#loc1252)) +#loc3343 = loc("reshape_1345"(#loc1253)) +#loc3344 = loc("transpose_1346"(#loc1254)) +#loc3345 = loc("transpose_1347"(#loc1255)) +#loc3346 = loc("reshape_1348.dc.squeeze.0"(#loc1256)) +#loc3347 = loc("transpose_1349"(#loc1257)) +#loc3348 = loc("matmul_1350"(#loc1258)) +#loc3349 = loc("reshape_1351.dc.unsqueeze.0"(#loc1259)) +#loc3350 = loc("transpose_1352"(#loc1260)) +#loc3351 = loc("reshape_1353"(#loc1261)) +#loc3352 = loc("matmul_1355"(#loc1262)) +#loc3353 = loc("reshape_1356.dc.unsqueeze.0"(#loc1263)) +#loc3354 = loc("add_1357"(#loc1264)) +#loc3355 = loc("multiply_1358"(#loc1265)) +#loc3356 = loc("reduce_avg_1359"(#loc1266)) +#loc3357 = loc("add_1360"(#loc1267)) +#loc3358 = loc("sqrt_1361"(#loc1268)) +#loc3359 = loc("reciprocal_1362"(#loc1269)) +#loc3360 = loc("multiply_1363"(#loc1270)) +#loc3361 = loc("multiply_1364"(#loc1271)) +#loc3362 = loc("reshape_1365.dc.squeeze.0"(#loc1272)) +#loc3363 = loc("matmul_1367"(#loc1273)) +#loc3364 = loc("reshape_1368.dc.unsqueeze.0"(#loc1274)) +#loc3365 = loc("sigmoid_1369"(#loc1275)) +#loc3366 = loc("multiply_1370"(#loc1276)) +#loc3367 = loc("matmul_1372"(#loc1277)) +#loc3368 = loc("reshape_1373.dc.unsqueeze.0"(#loc1278)) +#loc3369 = loc("multiply_1374"(#loc1279)) +#loc3370 = loc("matmul_1376"(#loc1280)) +#loc3371 = loc("add_1377"(#loc1281)) +#loc3372 = loc("multiply_1378"(#loc1282)) +#loc3373 = loc("reduce_avg_1379"(#loc1283)) +#loc3374 = loc("add_1380"(#loc1284)) +#loc3375 = loc("sqrt_1381"(#loc1285)) +#loc3376 = loc("reciprocal_1382"(#loc1286)) +#loc3377 = loc("multiply_1383"(#loc1287)) +#loc3378 = loc("multiply_1384"(#loc1288)) +#loc3379 = loc("reshape_1385.dc.squeeze.0"(#loc1289)) +#loc3380 = loc("matmul_1387"(#loc1290)) +#loc3381 = loc("reshape_1388"(#loc1291)) +#loc3382 = loc("transpose_1389"(#loc1292)) +#loc3383 = loc("concatenate_1396"(#loc1293)) +#loc3384 = loc("cosine_1397"(#loc1294)) +#loc3385 = loc("unsqueeze_1398"(#loc1295)) +#loc3386 = loc("multiply_1399"(#loc1296)) +#loc3387 = loc("index_1400.dc.transpose.0"(#loc1297)) +#loc3388 = loc("index_1400.dc.matmul.2"(#loc1298)) +#loc3389 = loc("index_1400.dc.transpose.3"(#loc1299)) +#loc3390 = loc("multiply_1401"(#loc1300)) +#loc3391 = loc("index_1402.dc.transpose.0"(#loc1301)) +#loc3392 = loc("index_1402.dc.matmul.2"(#loc1302)) +#loc3393 = loc("index_1402.dc.transpose.3"(#loc1303)) +#loc3394 = loc("concatenate_1403"(#loc1304)) +#loc3395 = loc("sine_1404"(#loc1305)) +#loc3396 = loc("unsqueeze_1405"(#loc1306)) +#loc3397 = loc("multiply_1406"(#loc1307)) +#loc3398 = loc("add_1407"(#loc1308)) +#loc3399 = loc("reshape_1408.dc.squeeze.0"(#loc1309)) +#loc3400 = loc("matmul_1410"(#loc1310)) +#loc3401 = loc("reshape_1411"(#loc1311)) +#loc3402 = loc("transpose_1412"(#loc1312)) +#loc3403 = loc("multiply_1413"(#loc1313)) +#loc3404 = loc("index_1414.dc.transpose.0"(#loc1314)) +#loc3405 = loc("index_1414.dc.matmul.2"(#loc1315)) +#loc3406 = loc("index_1414.dc.transpose.3"(#loc1316)) +#loc3407 = loc("multiply_1415"(#loc1317)) +#loc3408 = loc("index_1416.dc.transpose.0"(#loc1318)) +#loc3409 = loc("index_1416.dc.matmul.2"(#loc1319)) +#loc3410 = loc("index_1416.dc.transpose.3"(#loc1320)) +#loc3411 = loc("concatenate_1417"(#loc1321)) +#loc3412 = loc("multiply_1418"(#loc1322)) +#loc3413 = loc("add_1419"(#loc1323)) +#loc3414 = loc("reshape_1420.dc.squeeze.0"(#loc1324)) +#loc3415 = loc("transpose_1421"(#loc1325)) +#loc3416 = loc("matmul_1422"(#loc1326)) +#loc3417 = loc("reshape_1423.dc.unsqueeze.0"(#loc1327)) +#loc3418 = loc("multiply_1424"(#loc1328)) +#loc3419 = loc("add_1425"(#loc1329)) +#loc3420 = loc("softmax_1426"(#loc1330)) +#loc3421 = loc("reshape_1428.dc.squeeze.0"(#loc1331)) +#loc3422 = loc("matmul_1430"(#loc1332)) +#loc3423 = loc("reshape_1431"(#loc1333)) +#loc3424 = loc("transpose_1432"(#loc1334)) +#loc3425 = loc("transpose_1433"(#loc1335)) +#loc3426 = loc("reshape_1434.dc.squeeze.0"(#loc1336)) +#loc3427 = loc("transpose_1435"(#loc1337)) +#loc3428 = loc("matmul_1436"(#loc1338)) +#loc3429 = loc("reshape_1437.dc.unsqueeze.0"(#loc1339)) +#loc3430 = loc("transpose_1438"(#loc1340)) +#loc3431 = loc("reshape_1439"(#loc1341)) +#loc3432 = loc("matmul_1441"(#loc1342)) +#loc3433 = loc("reshape_1442.dc.unsqueeze.0"(#loc1343)) +#loc3434 = loc("add_1443"(#loc1344)) +#loc3435 = loc("multiply_1444"(#loc1345)) +#loc3436 = loc("reduce_avg_1445"(#loc1346)) +#loc3437 = loc("add_1446"(#loc1347)) +#loc3438 = loc("sqrt_1447"(#loc1348)) +#loc3439 = loc("reciprocal_1448"(#loc1349)) +#loc3440 = loc("multiply_1449"(#loc1350)) +#loc3441 = loc("multiply_1450"(#loc1351)) +#loc3442 = loc("reshape_1451.dc.squeeze.0"(#loc1352)) +#loc3443 = loc("matmul_1453"(#loc1353)) +#loc3444 = loc("reshape_1454.dc.unsqueeze.0"(#loc1354)) +#loc3445 = loc("sigmoid_1455"(#loc1355)) +#loc3446 = loc("multiply_1456"(#loc1356)) +#loc3447 = loc("matmul_1458"(#loc1357)) +#loc3448 = loc("reshape_1459.dc.unsqueeze.0"(#loc1358)) +#loc3449 = loc("multiply_1460"(#loc1359)) +#loc3450 = loc("matmul_1462"(#loc1360)) +#loc3451 = loc("add_1463"(#loc1361)) +#loc3452 = loc("multiply_1464"(#loc1362)) +#loc3453 = loc("reduce_avg_1465"(#loc1363)) +#loc3454 = loc("add_1466"(#loc1364)) +#loc3455 = loc("sqrt_1467"(#loc1365)) +#loc3456 = loc("reciprocal_1468"(#loc1366)) +#loc3457 = loc("multiply_1469"(#loc1367)) +#loc3458 = loc("multiply_1470"(#loc1368)) +#loc3459 = loc("reshape_1471.dc.squeeze.0"(#loc1369)) +#loc3460 = loc("matmul_1473"(#loc1370)) +#loc3461 = loc("reshape_1474"(#loc1371)) +#loc3462 = loc("transpose_1475"(#loc1372)) +#loc3463 = loc("concatenate_1482"(#loc1373)) +#loc3464 = loc("cosine_1483"(#loc1374)) +#loc3465 = loc("unsqueeze_1484"(#loc1375)) +#loc3466 = loc("multiply_1485"(#loc1376)) +#loc3467 = loc("index_1486.dc.transpose.0"(#loc1377)) +#loc3468 = loc("index_1486.dc.matmul.2"(#loc1378)) +#loc3469 = loc("index_1486.dc.transpose.3"(#loc1379)) +#loc3470 = loc("multiply_1487"(#loc1380)) +#loc3471 = loc("index_1488.dc.transpose.0"(#loc1381)) +#loc3472 = loc("index_1488.dc.matmul.2"(#loc1382)) +#loc3473 = loc("index_1488.dc.transpose.3"(#loc1383)) +#loc3474 = loc("concatenate_1489"(#loc1384)) +#loc3475 = loc("sine_1490"(#loc1385)) +#loc3476 = loc("unsqueeze_1491"(#loc1386)) +#loc3477 = loc("multiply_1492"(#loc1387)) +#loc3478 = loc("add_1493"(#loc1388)) +#loc3479 = loc("reshape_1494.dc.squeeze.0"(#loc1389)) +#loc3480 = loc("matmul_1496"(#loc1390)) +#loc3481 = loc("reshape_1497"(#loc1391)) +#loc3482 = loc("transpose_1498"(#loc1392)) +#loc3483 = loc("multiply_1499"(#loc1393)) +#loc3484 = loc("index_1500.dc.transpose.0"(#loc1394)) +#loc3485 = loc("index_1500.dc.matmul.2"(#loc1395)) +#loc3486 = loc("index_1500.dc.transpose.3"(#loc1396)) +#loc3487 = loc("multiply_1501"(#loc1397)) +#loc3488 = loc("index_1502.dc.transpose.0"(#loc1398)) +#loc3489 = loc("index_1502.dc.matmul.2"(#loc1399)) +#loc3490 = loc("index_1502.dc.transpose.3"(#loc1400)) +#loc3491 = loc("concatenate_1503"(#loc1401)) +#loc3492 = loc("multiply_1504"(#loc1402)) +#loc3493 = loc("add_1505"(#loc1403)) +#loc3494 = loc("reshape_1506.dc.squeeze.0"(#loc1404)) +#loc3495 = loc("transpose_1507"(#loc1405)) +#loc3496 = loc("matmul_1508"(#loc1406)) +#loc3497 = loc("reshape_1509.dc.unsqueeze.0"(#loc1407)) +#loc3498 = loc("multiply_1510"(#loc1408)) +#loc3499 = loc("add_1511"(#loc1409)) +#loc3500 = loc("softmax_1512"(#loc1410)) +#loc3501 = loc("reshape_1514.dc.squeeze.0"(#loc1411)) +#loc3502 = loc("matmul_1516"(#loc1412)) +#loc3503 = loc("reshape_1517"(#loc1413)) +#loc3504 = loc("transpose_1518"(#loc1414)) +#loc3505 = loc("transpose_1519"(#loc1415)) +#loc3506 = loc("reshape_1520.dc.squeeze.0"(#loc1416)) +#loc3507 = loc("transpose_1521"(#loc1417)) +#loc3508 = loc("matmul_1522"(#loc1418)) +#loc3509 = loc("reshape_1523.dc.unsqueeze.0"(#loc1419)) +#loc3510 = loc("transpose_1524"(#loc1420)) +#loc3511 = loc("reshape_1525"(#loc1421)) +#loc3512 = loc("matmul_1527"(#loc1422)) +#loc3513 = loc("reshape_1528.dc.unsqueeze.0"(#loc1423)) +#loc3514 = loc("add_1529"(#loc1424)) +#loc3515 = loc("multiply_1530"(#loc1425)) +#loc3516 = loc("reduce_avg_1531"(#loc1426)) +#loc3517 = loc("add_1532"(#loc1427)) +#loc3518 = loc("sqrt_1533"(#loc1428)) +#loc3519 = loc("reciprocal_1534"(#loc1429)) +#loc3520 = loc("multiply_1535"(#loc1430)) +#loc3521 = loc("multiply_1536"(#loc1431)) +#loc3522 = loc("reshape_1537.dc.squeeze.0"(#loc1432)) +#loc3523 = loc("matmul_1539"(#loc1433)) +#loc3524 = loc("reshape_1540.dc.unsqueeze.0"(#loc1434)) +#loc3525 = loc("sigmoid_1541"(#loc1435)) +#loc3526 = loc("multiply_1542"(#loc1436)) +#loc3527 = loc("matmul_1544"(#loc1437)) +#loc3528 = loc("reshape_1545.dc.unsqueeze.0"(#loc1438)) +#loc3529 = loc("multiply_1546"(#loc1439)) +#loc3530 = loc("matmul_1548"(#loc1440)) +#loc3531 = loc("add_1549"(#loc1441)) +#loc3532 = loc("multiply_1550"(#loc1442)) +#loc3533 = loc("reduce_avg_1551"(#loc1443)) +#loc3534 = loc("add_1552"(#loc1444)) +#loc3535 = loc("sqrt_1553"(#loc1445)) +#loc3536 = loc("reciprocal_1554"(#loc1446)) +#loc3537 = loc("multiply_1555"(#loc1447)) +#loc3538 = loc("multiply_1556"(#loc1448)) +#loc3539 = loc("reshape_1557.dc.squeeze.0"(#loc1449)) +#loc3540 = loc("matmul_1559"(#loc1450)) +#loc3541 = loc("reshape_1560"(#loc1451)) +#loc3542 = loc("transpose_1561"(#loc1452)) +#loc3543 = loc("concatenate_1568"(#loc1453)) +#loc3544 = loc("cosine_1569"(#loc1454)) +#loc3545 = loc("unsqueeze_1570"(#loc1455)) +#loc3546 = loc("multiply_1571"(#loc1456)) +#loc3547 = loc("index_1572.dc.transpose.0"(#loc1457)) +#loc3548 = loc("index_1572.dc.matmul.2"(#loc1458)) +#loc3549 = loc("index_1572.dc.transpose.3"(#loc1459)) +#loc3550 = loc("multiply_1573"(#loc1460)) +#loc3551 = loc("index_1574.dc.transpose.0"(#loc1461)) +#loc3552 = loc("index_1574.dc.matmul.2"(#loc1462)) +#loc3553 = loc("index_1574.dc.transpose.3"(#loc1463)) +#loc3554 = loc("concatenate_1575"(#loc1464)) +#loc3555 = loc("sine_1576"(#loc1465)) +#loc3556 = loc("unsqueeze_1577"(#loc1466)) +#loc3557 = loc("multiply_1578"(#loc1467)) +#loc3558 = loc("add_1579"(#loc1468)) +#loc3559 = loc("reshape_1580.dc.squeeze.0"(#loc1469)) +#loc3560 = loc("matmul_1582"(#loc1470)) +#loc3561 = loc("reshape_1583"(#loc1471)) +#loc3562 = loc("transpose_1584"(#loc1472)) +#loc3563 = loc("multiply_1585"(#loc1473)) +#loc3564 = loc("index_1586.dc.transpose.0"(#loc1474)) +#loc3565 = loc("index_1586.dc.matmul.2"(#loc1475)) +#loc3566 = loc("index_1586.dc.transpose.3"(#loc1476)) +#loc3567 = loc("multiply_1587"(#loc1477)) +#loc3568 = loc("index_1588.dc.transpose.0"(#loc1478)) +#loc3569 = loc("index_1588.dc.matmul.2"(#loc1479)) +#loc3570 = loc("index_1588.dc.transpose.3"(#loc1480)) +#loc3571 = loc("concatenate_1589"(#loc1481)) +#loc3572 = loc("multiply_1590"(#loc1482)) +#loc3573 = loc("add_1591"(#loc1483)) +#loc3574 = loc("reshape_1592.dc.squeeze.0"(#loc1484)) +#loc3575 = loc("transpose_1593"(#loc1485)) +#loc3576 = loc("matmul_1594"(#loc1486)) +#loc3577 = loc("reshape_1595.dc.unsqueeze.0"(#loc1487)) +#loc3578 = loc("multiply_1596"(#loc1488)) +#loc3579 = loc("add_1597"(#loc1489)) +#loc3580 = loc("softmax_1598"(#loc1490)) +#loc3581 = loc("reshape_1600.dc.squeeze.0"(#loc1491)) +#loc3582 = loc("matmul_1602"(#loc1492)) +#loc3583 = loc("reshape_1603"(#loc1493)) +#loc3584 = loc("transpose_1604"(#loc1494)) +#loc3585 = loc("transpose_1605"(#loc1495)) +#loc3586 = loc("reshape_1606.dc.squeeze.0"(#loc1496)) +#loc3587 = loc("transpose_1607"(#loc1497)) +#loc3588 = loc("matmul_1608"(#loc1498)) +#loc3589 = loc("reshape_1609.dc.unsqueeze.0"(#loc1499)) +#loc3590 = loc("transpose_1610"(#loc1500)) +#loc3591 = loc("reshape_1611"(#loc1501)) +#loc3592 = loc("matmul_1613"(#loc1502)) +#loc3593 = loc("reshape_1614.dc.unsqueeze.0"(#loc1503)) +#loc3594 = loc("add_1615"(#loc1504)) +#loc3595 = loc("multiply_1616"(#loc1505)) +#loc3596 = loc("reduce_avg_1617"(#loc1506)) +#loc3597 = loc("add_1618"(#loc1507)) +#loc3598 = loc("sqrt_1619"(#loc1508)) +#loc3599 = loc("reciprocal_1620"(#loc1509)) +#loc3600 = loc("multiply_1621"(#loc1510)) +#loc3601 = loc("multiply_1622"(#loc1511)) +#loc3602 = loc("reshape_1623.dc.squeeze.0"(#loc1512)) +#loc3603 = loc("matmul_1625"(#loc1513)) +#loc3604 = loc("reshape_1626.dc.unsqueeze.0"(#loc1514)) +#loc3605 = loc("sigmoid_1627"(#loc1515)) +#loc3606 = loc("multiply_1628"(#loc1516)) +#loc3607 = loc("matmul_1630"(#loc1517)) +#loc3608 = loc("reshape_1631.dc.unsqueeze.0"(#loc1518)) +#loc3609 = loc("multiply_1632"(#loc1519)) +#loc3610 = loc("matmul_1634"(#loc1520)) +#loc3611 = loc("add_1635"(#loc1521)) +#loc3612 = loc("multiply_1636"(#loc1522)) +#loc3613 = loc("reduce_avg_1637"(#loc1523)) +#loc3614 = loc("add_1638"(#loc1524)) +#loc3615 = loc("sqrt_1639"(#loc1525)) +#loc3616 = loc("reciprocal_1640"(#loc1526)) +#loc3617 = loc("multiply_1641"(#loc1527)) +#loc3618 = loc("multiply_1642"(#loc1528)) +#loc3619 = loc("reshape_1643.dc.squeeze.0"(#loc1529)) +#loc3620 = loc("matmul_1645"(#loc1530)) +#loc3621 = loc("reshape_1646"(#loc1531)) +#loc3622 = loc("transpose_1647"(#loc1532)) +#loc3623 = loc("concatenate_1654"(#loc1533)) +#loc3624 = loc("cosine_1655"(#loc1534)) +#loc3625 = loc("unsqueeze_1656"(#loc1535)) +#loc3626 = loc("multiply_1657"(#loc1536)) +#loc3627 = loc("index_1658.dc.transpose.0"(#loc1537)) +#loc3628 = loc("index_1658.dc.matmul.2"(#loc1538)) +#loc3629 = loc("index_1658.dc.transpose.3"(#loc1539)) +#loc3630 = loc("multiply_1659"(#loc1540)) +#loc3631 = loc("index_1660.dc.transpose.0"(#loc1541)) +#loc3632 = loc("index_1660.dc.matmul.2"(#loc1542)) +#loc3633 = loc("index_1660.dc.transpose.3"(#loc1543)) +#loc3634 = loc("concatenate_1661"(#loc1544)) +#loc3635 = loc("sine_1662"(#loc1545)) +#loc3636 = loc("unsqueeze_1663"(#loc1546)) +#loc3637 = loc("multiply_1664"(#loc1547)) +#loc3638 = loc("add_1665"(#loc1548)) +#loc3639 = loc("reshape_1666.dc.squeeze.0"(#loc1549)) +#loc3640 = loc("matmul_1668"(#loc1550)) +#loc3641 = loc("reshape_1669"(#loc1551)) +#loc3642 = loc("transpose_1670"(#loc1552)) +#loc3643 = loc("multiply_1671"(#loc1553)) +#loc3644 = loc("index_1672.dc.transpose.0"(#loc1554)) +#loc3645 = loc("index_1672.dc.matmul.2"(#loc1555)) +#loc3646 = loc("index_1672.dc.transpose.3"(#loc1556)) +#loc3647 = loc("multiply_1673"(#loc1557)) +#loc3648 = loc("index_1674.dc.transpose.0"(#loc1558)) +#loc3649 = loc("index_1674.dc.matmul.2"(#loc1559)) +#loc3650 = loc("index_1674.dc.transpose.3"(#loc1560)) +#loc3651 = loc("concatenate_1675"(#loc1561)) +#loc3652 = loc("multiply_1676"(#loc1562)) +#loc3653 = loc("add_1677"(#loc1563)) +#loc3654 = loc("reshape_1678.dc.squeeze.0"(#loc1564)) +#loc3655 = loc("transpose_1679"(#loc1565)) +#loc3656 = loc("matmul_1680"(#loc1566)) +#loc3657 = loc("reshape_1681.dc.unsqueeze.0"(#loc1567)) +#loc3658 = loc("multiply_1682"(#loc1568)) +#loc3659 = loc("add_1683"(#loc1569)) +#loc3660 = loc("softmax_1684"(#loc1570)) +#loc3661 = loc("reshape_1686.dc.squeeze.0"(#loc1571)) +#loc3662 = loc("matmul_1688"(#loc1572)) +#loc3663 = loc("reshape_1689"(#loc1573)) +#loc3664 = loc("transpose_1690"(#loc1574)) +#loc3665 = loc("transpose_1691"(#loc1575)) +#loc3666 = loc("reshape_1692.dc.squeeze.0"(#loc1576)) +#loc3667 = loc("transpose_1693"(#loc1577)) +#loc3668 = loc("matmul_1694"(#loc1578)) +#loc3669 = loc("reshape_1695.dc.unsqueeze.0"(#loc1579)) +#loc3670 = loc("transpose_1696"(#loc1580)) +#loc3671 = loc("reshape_1697"(#loc1581)) +#loc3672 = loc("matmul_1699"(#loc1582)) +#loc3673 = loc("reshape_1700.dc.unsqueeze.0"(#loc1583)) +#loc3674 = loc("add_1701"(#loc1584)) +#loc3675 = loc("multiply_1702"(#loc1585)) +#loc3676 = loc("reduce_avg_1703"(#loc1586)) +#loc3677 = loc("add_1704"(#loc1587)) +#loc3678 = loc("sqrt_1705"(#loc1588)) +#loc3679 = loc("reciprocal_1706"(#loc1589)) +#loc3680 = loc("multiply_1707"(#loc1590)) +#loc3681 = loc("multiply_1708"(#loc1591)) +#loc3682 = loc("reshape_1709.dc.squeeze.0"(#loc1592)) +#loc3683 = loc("matmul_1711"(#loc1593)) +#loc3684 = loc("reshape_1712.dc.unsqueeze.0"(#loc1594)) +#loc3685 = loc("sigmoid_1713"(#loc1595)) +#loc3686 = loc("multiply_1714"(#loc1596)) +#loc3687 = loc("matmul_1716"(#loc1597)) +#loc3688 = loc("reshape_1717.dc.unsqueeze.0"(#loc1598)) +#loc3689 = loc("multiply_1718"(#loc1599)) +#loc3690 = loc("matmul_1720"(#loc1600)) +#loc3691 = loc("add_1721"(#loc1601)) +#loc3692 = loc("multiply_1722"(#loc1602)) +#loc3693 = loc("reduce_avg_1723"(#loc1603)) +#loc3694 = loc("add_1724"(#loc1604)) +#loc3695 = loc("sqrt_1725"(#loc1605)) +#loc3696 = loc("reciprocal_1726"(#loc1606)) +#loc3697 = loc("multiply_1727"(#loc1607)) +#loc3698 = loc("multiply_1728"(#loc1608)) +#loc3699 = loc("reshape_1729.dc.squeeze.0"(#loc1609)) +#loc3700 = loc("matmul_1731"(#loc1610)) +#loc3701 = loc("reshape_1732"(#loc1611)) +#loc3702 = loc("transpose_1733"(#loc1612)) +#loc3703 = loc("concatenate_1740"(#loc1613)) +#loc3704 = loc("cosine_1741"(#loc1614)) +#loc3705 = loc("unsqueeze_1742"(#loc1615)) +#loc3706 = loc("multiply_1743"(#loc1616)) +#loc3707 = loc("index_1744.dc.transpose.0"(#loc1617)) +#loc3708 = loc("index_1744.dc.matmul.2"(#loc1618)) +#loc3709 = loc("index_1744.dc.transpose.3"(#loc1619)) +#loc3710 = loc("multiply_1745"(#loc1620)) +#loc3711 = loc("index_1746.dc.transpose.0"(#loc1621)) +#loc3712 = loc("index_1746.dc.matmul.2"(#loc1622)) +#loc3713 = loc("index_1746.dc.transpose.3"(#loc1623)) +#loc3714 = loc("concatenate_1747"(#loc1624)) +#loc3715 = loc("sine_1748"(#loc1625)) +#loc3716 = loc("unsqueeze_1749"(#loc1626)) +#loc3717 = loc("multiply_1750"(#loc1627)) +#loc3718 = loc("add_1751"(#loc1628)) +#loc3719 = loc("reshape_1752.dc.squeeze.0"(#loc1629)) +#loc3720 = loc("matmul_1754"(#loc1630)) +#loc3721 = loc("reshape_1755"(#loc1631)) +#loc3722 = loc("transpose_1756"(#loc1632)) +#loc3723 = loc("multiply_1757"(#loc1633)) +#loc3724 = loc("index_1758.dc.transpose.0"(#loc1634)) +#loc3725 = loc("index_1758.dc.matmul.2"(#loc1635)) +#loc3726 = loc("index_1758.dc.transpose.3"(#loc1636)) +#loc3727 = loc("multiply_1759"(#loc1637)) +#loc3728 = loc("index_1760.dc.transpose.0"(#loc1638)) +#loc3729 = loc("index_1760.dc.matmul.2"(#loc1639)) +#loc3730 = loc("index_1760.dc.transpose.3"(#loc1640)) +#loc3731 = loc("concatenate_1761"(#loc1641)) +#loc3732 = loc("multiply_1762"(#loc1642)) +#loc3733 = loc("add_1763"(#loc1643)) +#loc3734 = loc("reshape_1764.dc.squeeze.0"(#loc1644)) +#loc3735 = loc("transpose_1765"(#loc1645)) +#loc3736 = loc("matmul_1766"(#loc1646)) +#loc3737 = loc("reshape_1767.dc.unsqueeze.0"(#loc1647)) +#loc3738 = loc("multiply_1768"(#loc1648)) +#loc3739 = loc("add_1769"(#loc1649)) +#loc3740 = loc("softmax_1770"(#loc1650)) +#loc3741 = loc("reshape_1772.dc.squeeze.0"(#loc1651)) +#loc3742 = loc("matmul_1774"(#loc1652)) +#loc3743 = loc("reshape_1775"(#loc1653)) +#loc3744 = loc("transpose_1776"(#loc1654)) +#loc3745 = loc("transpose_1777"(#loc1655)) +#loc3746 = loc("reshape_1778.dc.squeeze.0"(#loc1656)) +#loc3747 = loc("transpose_1779"(#loc1657)) +#loc3748 = loc("matmul_1780"(#loc1658)) +#loc3749 = loc("reshape_1781.dc.unsqueeze.0"(#loc1659)) +#loc3750 = loc("transpose_1782"(#loc1660)) +#loc3751 = loc("reshape_1783"(#loc1661)) +#loc3752 = loc("matmul_1785"(#loc1662)) +#loc3753 = loc("reshape_1786.dc.unsqueeze.0"(#loc1663)) +#loc3754 = loc("add_1787"(#loc1664)) +#loc3755 = loc("multiply_1788"(#loc1665)) +#loc3756 = loc("reduce_avg_1789"(#loc1666)) +#loc3757 = loc("add_1790"(#loc1667)) +#loc3758 = loc("sqrt_1791"(#loc1668)) +#loc3759 = loc("reciprocal_1792"(#loc1669)) +#loc3760 = loc("multiply_1793"(#loc1670)) +#loc3761 = loc("multiply_1794"(#loc1671)) +#loc3762 = loc("reshape_1795.dc.squeeze.0"(#loc1672)) +#loc3763 = loc("matmul_1797"(#loc1673)) +#loc3764 = loc("reshape_1798.dc.unsqueeze.0"(#loc1674)) +#loc3765 = loc("sigmoid_1799"(#loc1675)) +#loc3766 = loc("multiply_1800"(#loc1676)) +#loc3767 = loc("matmul_1802"(#loc1677)) +#loc3768 = loc("reshape_1803.dc.unsqueeze.0"(#loc1678)) +#loc3769 = loc("multiply_1804"(#loc1679)) +#loc3770 = loc("matmul_1806"(#loc1680)) +#loc3771 = loc("add_1807"(#loc1681)) +#loc3772 = loc("multiply_1808"(#loc1682)) +#loc3773 = loc("reduce_avg_1809"(#loc1683)) +#loc3774 = loc("add_1810"(#loc1684)) +#loc3775 = loc("sqrt_1811"(#loc1685)) +#loc3776 = loc("reciprocal_1812"(#loc1686)) +#loc3777 = loc("multiply_1813"(#loc1687)) +#loc3778 = loc("multiply_1814"(#loc1688)) +#loc3779 = loc("reshape_1815.dc.squeeze.0"(#loc1689)) +#loc3780 = loc("matmul_1817"(#loc1690)) +#loc3781 = loc("reshape_1818"(#loc1691)) +#loc3782 = loc("transpose_1819"(#loc1692)) +#loc3783 = loc("concatenate_1826"(#loc1693)) +#loc3784 = loc("cosine_1827"(#loc1694)) +#loc3785 = loc("unsqueeze_1828"(#loc1695)) +#loc3786 = loc("multiply_1829"(#loc1696)) +#loc3787 = loc("index_1830.dc.transpose.0"(#loc1697)) +#loc3788 = loc("index_1830.dc.matmul.2"(#loc1698)) +#loc3789 = loc("index_1830.dc.transpose.3"(#loc1699)) +#loc3790 = loc("multiply_1831"(#loc1700)) +#loc3791 = loc("index_1832.dc.transpose.0"(#loc1701)) +#loc3792 = loc("index_1832.dc.matmul.2"(#loc1702)) +#loc3793 = loc("index_1832.dc.transpose.3"(#loc1703)) +#loc3794 = loc("concatenate_1833"(#loc1704)) +#loc3795 = loc("sine_1834"(#loc1705)) +#loc3796 = loc("unsqueeze_1835"(#loc1706)) +#loc3797 = loc("multiply_1836"(#loc1707)) +#loc3798 = loc("add_1837"(#loc1708)) +#loc3799 = loc("reshape_1838.dc.squeeze.0"(#loc1709)) +#loc3800 = loc("matmul_1840"(#loc1710)) +#loc3801 = loc("reshape_1841"(#loc1711)) +#loc3802 = loc("transpose_1842"(#loc1712)) +#loc3803 = loc("multiply_1843"(#loc1713)) +#loc3804 = loc("index_1844.dc.transpose.0"(#loc1714)) +#loc3805 = loc("index_1844.dc.matmul.2"(#loc1715)) +#loc3806 = loc("index_1844.dc.transpose.3"(#loc1716)) +#loc3807 = loc("multiply_1845"(#loc1717)) +#loc3808 = loc("index_1846.dc.transpose.0"(#loc1718)) +#loc3809 = loc("index_1846.dc.matmul.2"(#loc1719)) +#loc3810 = loc("index_1846.dc.transpose.3"(#loc1720)) +#loc3811 = loc("concatenate_1847"(#loc1721)) +#loc3812 = loc("multiply_1848"(#loc1722)) +#loc3813 = loc("add_1849"(#loc1723)) +#loc3814 = loc("reshape_1850.dc.squeeze.0"(#loc1724)) +#loc3815 = loc("transpose_1851"(#loc1725)) +#loc3816 = loc("matmul_1852"(#loc1726)) +#loc3817 = loc("reshape_1853.dc.unsqueeze.0"(#loc1727)) +#loc3818 = loc("multiply_1854"(#loc1728)) +#loc3819 = loc("add_1855"(#loc1729)) +#loc3820 = loc("softmax_1856"(#loc1730)) +#loc3821 = loc("reshape_1858.dc.squeeze.0"(#loc1731)) +#loc3822 = loc("matmul_1860"(#loc1732)) +#loc3823 = loc("reshape_1861"(#loc1733)) +#loc3824 = loc("transpose_1862"(#loc1734)) +#loc3825 = loc("transpose_1863"(#loc1735)) +#loc3826 = loc("reshape_1864.dc.squeeze.0"(#loc1736)) +#loc3827 = loc("transpose_1865"(#loc1737)) +#loc3828 = loc("matmul_1866"(#loc1738)) +#loc3829 = loc("reshape_1867.dc.unsqueeze.0"(#loc1739)) +#loc3830 = loc("transpose_1868"(#loc1740)) +#loc3831 = loc("reshape_1869"(#loc1741)) +#loc3832 = loc("matmul_1871"(#loc1742)) +#loc3833 = loc("reshape_1872.dc.unsqueeze.0"(#loc1743)) +#loc3834 = loc("add_1873"(#loc1744)) +#loc3835 = loc("multiply_1874"(#loc1745)) +#loc3836 = loc("reduce_avg_1875"(#loc1746)) +#loc3837 = loc("add_1876"(#loc1747)) +#loc3838 = loc("sqrt_1877"(#loc1748)) +#loc3839 = loc("reciprocal_1878"(#loc1749)) +#loc3840 = loc("multiply_1879"(#loc1750)) +#loc3841 = loc("multiply_1880"(#loc1751)) +#loc3842 = loc("reshape_1881.dc.squeeze.0"(#loc1752)) +#loc3843 = loc("matmul_1883"(#loc1753)) +#loc3844 = loc("reshape_1884.dc.unsqueeze.0"(#loc1754)) +#loc3845 = loc("sigmoid_1885"(#loc1755)) +#loc3846 = loc("multiply_1886"(#loc1756)) +#loc3847 = loc("matmul_1888"(#loc1757)) +#loc3848 = loc("reshape_1889.dc.unsqueeze.0"(#loc1758)) +#loc3849 = loc("multiply_1890"(#loc1759)) +#loc3850 = loc("matmul_1892"(#loc1760)) +#loc3851 = loc("add_1893"(#loc1761)) +#loc3852 = loc("multiply_1894"(#loc1762)) +#loc3853 = loc("reduce_avg_1895"(#loc1763)) +#loc3854 = loc("add_1896"(#loc1764)) +#loc3855 = loc("sqrt_1897"(#loc1765)) +#loc3856 = loc("reciprocal_1898"(#loc1766)) +#loc3857 = loc("multiply_1899"(#loc1767)) +#loc3858 = loc("multiply_1900"(#loc1768)) +#loc3859 = loc("reshape_1901.dc.squeeze.0"(#loc1769)) +#loc3860 = loc("matmul_1903"(#loc1770)) +#loc3861 = loc("reshape_1904"(#loc1771)) +#loc3862 = loc("transpose_1905"(#loc1772)) +#loc3863 = loc("concatenate_1912"(#loc1773)) +#loc3864 = loc("cosine_1913"(#loc1774)) +#loc3865 = loc("unsqueeze_1914"(#loc1775)) +#loc3866 = loc("multiply_1915"(#loc1776)) +#loc3867 = loc("index_1916.dc.transpose.0"(#loc1777)) +#loc3868 = loc("index_1916.dc.matmul.2"(#loc1778)) +#loc3869 = loc("index_1916.dc.transpose.3"(#loc1779)) +#loc3870 = loc("multiply_1917"(#loc1780)) +#loc3871 = loc("index_1918.dc.transpose.0"(#loc1781)) +#loc3872 = loc("index_1918.dc.matmul.2"(#loc1782)) +#loc3873 = loc("index_1918.dc.transpose.3"(#loc1783)) +#loc3874 = loc("concatenate_1919"(#loc1784)) +#loc3875 = loc("sine_1920"(#loc1785)) +#loc3876 = loc("unsqueeze_1921"(#loc1786)) +#loc3877 = loc("multiply_1922"(#loc1787)) +#loc3878 = loc("add_1923"(#loc1788)) +#loc3879 = loc("reshape_1924.dc.squeeze.0"(#loc1789)) +#loc3880 = loc("matmul_1926"(#loc1790)) +#loc3881 = loc("reshape_1927"(#loc1791)) +#loc3882 = loc("transpose_1928"(#loc1792)) +#loc3883 = loc("multiply_1929"(#loc1793)) +#loc3884 = loc("index_1930.dc.transpose.0"(#loc1794)) +#loc3885 = loc("index_1930.dc.matmul.2"(#loc1795)) +#loc3886 = loc("index_1930.dc.transpose.3"(#loc1796)) +#loc3887 = loc("multiply_1931"(#loc1797)) +#loc3888 = loc("index_1932.dc.transpose.0"(#loc1798)) +#loc3889 = loc("index_1932.dc.matmul.2"(#loc1799)) +#loc3890 = loc("index_1932.dc.transpose.3"(#loc1800)) +#loc3891 = loc("concatenate_1933"(#loc1801)) +#loc3892 = loc("multiply_1934"(#loc1802)) +#loc3893 = loc("add_1935"(#loc1803)) +#loc3894 = loc("reshape_1936.dc.squeeze.0"(#loc1804)) +#loc3895 = loc("transpose_1937"(#loc1805)) +#loc3896 = loc("matmul_1938"(#loc1806)) +#loc3897 = loc("reshape_1939.dc.unsqueeze.0"(#loc1807)) +#loc3898 = loc("multiply_1940"(#loc1808)) +#loc3899 = loc("add_1941"(#loc1809)) +#loc3900 = loc("softmax_1942"(#loc1810)) +#loc3901 = loc("reshape_1944.dc.squeeze.0"(#loc1811)) +#loc3902 = loc("matmul_1946"(#loc1812)) +#loc3903 = loc("reshape_1947"(#loc1813)) +#loc3904 = loc("transpose_1948"(#loc1814)) +#loc3905 = loc("transpose_1949"(#loc1815)) +#loc3906 = loc("reshape_1950.dc.squeeze.0"(#loc1816)) +#loc3907 = loc("transpose_1951"(#loc1817)) +#loc3908 = loc("matmul_1952"(#loc1818)) +#loc3909 = loc("reshape_1953.dc.unsqueeze.0"(#loc1819)) +#loc3910 = loc("transpose_1954"(#loc1820)) +#loc3911 = loc("reshape_1955"(#loc1821)) +#loc3912 = loc("matmul_1957"(#loc1822)) +#loc3913 = loc("reshape_1958.dc.unsqueeze.0"(#loc1823)) +#loc3914 = loc("add_1959"(#loc1824)) +#loc3915 = loc("multiply_1960"(#loc1825)) +#loc3916 = loc("reduce_avg_1961"(#loc1826)) +#loc3917 = loc("add_1962"(#loc1827)) +#loc3918 = loc("sqrt_1963"(#loc1828)) +#loc3919 = loc("reciprocal_1964"(#loc1829)) +#loc3920 = loc("multiply_1965"(#loc1830)) +#loc3921 = loc("multiply_1966"(#loc1831)) +#loc3922 = loc("reshape_1967.dc.squeeze.0"(#loc1832)) +#loc3923 = loc("matmul_1969"(#loc1833)) +#loc3924 = loc("reshape_1970.dc.unsqueeze.0"(#loc1834)) +#loc3925 = loc("sigmoid_1971"(#loc1835)) +#loc3926 = loc("multiply_1972"(#loc1836)) +#loc3927 = loc("matmul_1974"(#loc1837)) +#loc3928 = loc("reshape_1975.dc.unsqueeze.0"(#loc1838)) +#loc3929 = loc("multiply_1976"(#loc1839)) +#loc3930 = loc("matmul_1978"(#loc1840)) +#loc3931 = loc("add_1979"(#loc1841)) +#loc3932 = loc("multiply_1980"(#loc1842)) +#loc3933 = loc("reduce_avg_1981"(#loc1843)) +#loc3934 = loc("add_1982"(#loc1844)) +#loc3935 = loc("sqrt_1983"(#loc1845)) +#loc3936 = loc("reciprocal_1984"(#loc1846)) +#loc3937 = loc("multiply_1985"(#loc1847)) +#loc3938 = loc("multiply_1986"(#loc1848)) +#loc3939 = loc("reshape_1987.dc.squeeze.0"(#loc1849)) +#loc3940 = loc("matmul_1989"(#loc1850)) +#loc3941 = loc("reshape_1990"(#loc1851)) +#loc3942 = loc("transpose_1991"(#loc1852)) +#loc3943 = loc("concatenate_1998"(#loc1853)) +#loc3944 = loc("cosine_1999"(#loc1854)) +#loc3945 = loc("unsqueeze_2000"(#loc1855)) +#loc3946 = loc("multiply_2001"(#loc1856)) +#loc3947 = loc("index_2002.dc.transpose.0"(#loc1857)) +#loc3948 = loc("index_2002.dc.matmul.2"(#loc1858)) +#loc3949 = loc("index_2002.dc.transpose.3"(#loc1859)) +#loc3950 = loc("multiply_2003"(#loc1860)) +#loc3951 = loc("index_2004.dc.transpose.0"(#loc1861)) +#loc3952 = loc("index_2004.dc.matmul.2"(#loc1862)) +#loc3953 = loc("index_2004.dc.transpose.3"(#loc1863)) +#loc3954 = loc("concatenate_2005"(#loc1864)) +#loc3955 = loc("sine_2006"(#loc1865)) +#loc3956 = loc("unsqueeze_2007"(#loc1866)) +#loc3957 = loc("multiply_2008"(#loc1867)) +#loc3958 = loc("add_2009"(#loc1868)) +#loc3959 = loc("reshape_2010.dc.squeeze.0"(#loc1869)) +#loc3960 = loc("matmul_2012"(#loc1870)) +#loc3961 = loc("reshape_2013"(#loc1871)) +#loc3962 = loc("transpose_2014"(#loc1872)) +#loc3963 = loc("multiply_2015"(#loc1873)) +#loc3964 = loc("index_2016.dc.transpose.0"(#loc1874)) +#loc3965 = loc("index_2016.dc.matmul.2"(#loc1875)) +#loc3966 = loc("index_2016.dc.transpose.3"(#loc1876)) +#loc3967 = loc("multiply_2017"(#loc1877)) +#loc3968 = loc("index_2018.dc.transpose.0"(#loc1878)) +#loc3969 = loc("index_2018.dc.matmul.2"(#loc1879)) +#loc3970 = loc("index_2018.dc.transpose.3"(#loc1880)) +#loc3971 = loc("concatenate_2019"(#loc1881)) +#loc3972 = loc("multiply_2020"(#loc1882)) +#loc3973 = loc("add_2021"(#loc1883)) +#loc3974 = loc("reshape_2022.dc.squeeze.0"(#loc1884)) +#loc3975 = loc("transpose_2023"(#loc1885)) +#loc3976 = loc("matmul_2024"(#loc1886)) +#loc3977 = loc("reshape_2025.dc.unsqueeze.0"(#loc1887)) +#loc3978 = loc("multiply_2026"(#loc1888)) +#loc3979 = loc("add_2027"(#loc1889)) +#loc3980 = loc("softmax_2028"(#loc1890)) +#loc3981 = loc("reshape_2030.dc.squeeze.0"(#loc1891)) +#loc3982 = loc("matmul_2032"(#loc1892)) +#loc3983 = loc("reshape_2033"(#loc1893)) +#loc3984 = loc("transpose_2034"(#loc1894)) +#loc3985 = loc("transpose_2035"(#loc1895)) +#loc3986 = loc("reshape_2036.dc.squeeze.0"(#loc1896)) +#loc3987 = loc("transpose_2037"(#loc1897)) +#loc3988 = loc("matmul_2038"(#loc1898)) +#loc3989 = loc("reshape_2039.dc.unsqueeze.0"(#loc1899)) +#loc3990 = loc("transpose_2040"(#loc1900)) +#loc3991 = loc("reshape_2041"(#loc1901)) +#loc3992 = loc("matmul_2043"(#loc1902)) +#loc3993 = loc("reshape_2044.dc.unsqueeze.0"(#loc1903)) +#loc3994 = loc("add_2045"(#loc1904)) +#loc3995 = loc("multiply_2046"(#loc1905)) +#loc3996 = loc("reduce_avg_2047"(#loc1906)) +#loc3997 = loc("add_2048"(#loc1907)) +#loc3998 = loc("sqrt_2049"(#loc1908)) +#loc3999 = loc("reciprocal_2050"(#loc1909)) +#loc4000 = loc("multiply_2051"(#loc1910)) +#loc4001 = loc("multiply_2052"(#loc1911)) +#loc4002 = loc("reshape_2053.dc.squeeze.0"(#loc1912)) +#loc4003 = loc("matmul_2055"(#loc1913)) +#loc4004 = loc("reshape_2056.dc.unsqueeze.0"(#loc1914)) +#loc4005 = loc("sigmoid_2057"(#loc1915)) +#loc4006 = loc("multiply_2058"(#loc1916)) +#loc4007 = loc("matmul_2060"(#loc1917)) +#loc4008 = loc("reshape_2061.dc.unsqueeze.0"(#loc1918)) +#loc4009 = loc("multiply_2062"(#loc1919)) +#loc4010 = loc("matmul_2064"(#loc1920)) +#loc4011 = loc("add_2065"(#loc1921)) +#loc4012 = loc("multiply_2066"(#loc1922)) +#loc4013 = loc("reduce_avg_2067"(#loc1923)) +#loc4014 = loc("add_2068"(#loc1924)) +#loc4015 = loc("sqrt_2069"(#loc1925)) +#loc4016 = loc("reciprocal_2070"(#loc1926)) +#loc4017 = loc("multiply_2071"(#loc1927)) +#loc4018 = loc("multiply_2072"(#loc1928)) +#loc4019 = loc("reshape_2073.dc.squeeze.0"(#loc1929)) +#loc4020 = loc("matmul_2075"(#loc1930)) +#loc4021 = loc("reshape_2076"(#loc1931)) +#loc4022 = loc("transpose_2077"(#loc1932)) +#loc4023 = loc("concatenate_2084"(#loc1933)) +#loc4024 = loc("cosine_2085"(#loc1934)) +#loc4025 = loc("unsqueeze_2086"(#loc1935)) +#loc4026 = loc("multiply_2087"(#loc1936)) +#loc4027 = loc("index_2088.dc.transpose.0"(#loc1937)) +#loc4028 = loc("index_2088.dc.matmul.2"(#loc1938)) +#loc4029 = loc("index_2088.dc.transpose.3"(#loc1939)) +#loc4030 = loc("multiply_2089"(#loc1940)) +#loc4031 = loc("index_2090.dc.transpose.0"(#loc1941)) +#loc4032 = loc("index_2090.dc.matmul.2"(#loc1942)) +#loc4033 = loc("index_2090.dc.transpose.3"(#loc1943)) +#loc4034 = loc("concatenate_2091"(#loc1944)) +#loc4035 = loc("sine_2092"(#loc1945)) +#loc4036 = loc("unsqueeze_2093"(#loc1946)) +#loc4037 = loc("multiply_2094"(#loc1947)) +#loc4038 = loc("add_2095"(#loc1948)) +#loc4039 = loc("reshape_2096.dc.squeeze.0"(#loc1949)) +#loc4040 = loc("matmul_2098"(#loc1950)) +#loc4041 = loc("reshape_2099"(#loc1951)) +#loc4042 = loc("transpose_2100"(#loc1952)) +#loc4043 = loc("multiply_2101"(#loc1953)) +#loc4044 = loc("index_2102.dc.transpose.0"(#loc1954)) +#loc4045 = loc("index_2102.dc.matmul.2"(#loc1955)) +#loc4046 = loc("index_2102.dc.transpose.3"(#loc1956)) +#loc4047 = loc("multiply_2103"(#loc1957)) +#loc4048 = loc("index_2104.dc.transpose.0"(#loc1958)) +#loc4049 = loc("index_2104.dc.matmul.2"(#loc1959)) +#loc4050 = loc("index_2104.dc.transpose.3"(#loc1960)) +#loc4051 = loc("concatenate_2105"(#loc1961)) +#loc4052 = loc("multiply_2106"(#loc1962)) +#loc4053 = loc("add_2107"(#loc1963)) +#loc4054 = loc("reshape_2108.dc.squeeze.0"(#loc1964)) +#loc4055 = loc("transpose_2109"(#loc1965)) +#loc4056 = loc("matmul_2110"(#loc1966)) +#loc4057 = loc("reshape_2111.dc.unsqueeze.0"(#loc1967)) +#loc4058 = loc("multiply_2112"(#loc1968)) +#loc4059 = loc("add_2113"(#loc1969)) +#loc4060 = loc("softmax_2114"(#loc1970)) +#loc4061 = loc("reshape_2116.dc.squeeze.0"(#loc1971)) +#loc4062 = loc("matmul_2118"(#loc1972)) +#loc4063 = loc("reshape_2119"(#loc1973)) +#loc4064 = loc("transpose_2120"(#loc1974)) +#loc4065 = loc("transpose_2121"(#loc1975)) +#loc4066 = loc("reshape_2122.dc.squeeze.0"(#loc1976)) +#loc4067 = loc("transpose_2123"(#loc1977)) +#loc4068 = loc("matmul_2124"(#loc1978)) +#loc4069 = loc("reshape_2125.dc.unsqueeze.0"(#loc1979)) +#loc4070 = loc("transpose_2126"(#loc1980)) +#loc4071 = loc("reshape_2127"(#loc1981)) +#loc4072 = loc("matmul_2129"(#loc1982)) +#loc4073 = loc("reshape_2130.dc.unsqueeze.0"(#loc1983)) +#loc4074 = loc("add_2131"(#loc1984)) +#loc4075 = loc("multiply_2132"(#loc1985)) +#loc4076 = loc("reduce_avg_2133"(#loc1986)) +#loc4077 = loc("add_2134"(#loc1987)) +#loc4078 = loc("sqrt_2135"(#loc1988)) +#loc4079 = loc("reciprocal_2136"(#loc1989)) +#loc4080 = loc("multiply_2137"(#loc1990)) +#loc4081 = loc("multiply_2138"(#loc1991)) +#loc4082 = loc("reshape_2139.dc.squeeze.0"(#loc1992)) +#loc4083 = loc("matmul_2141"(#loc1993)) +#loc4084 = loc("reshape_2142.dc.unsqueeze.0"(#loc1994)) +#loc4085 = loc("sigmoid_2143"(#loc1995)) +#loc4086 = loc("multiply_2144"(#loc1996)) +#loc4087 = loc("matmul_2146"(#loc1997)) +#loc4088 = loc("reshape_2147.dc.unsqueeze.0"(#loc1998)) +#loc4089 = loc("multiply_2148"(#loc1999)) +#loc4090 = loc("matmul_2150"(#loc2000)) +#loc4091 = loc("add_2151"(#loc2001)) +#loc4092 = loc("multiply_2152"(#loc2002)) +#loc4093 = loc("reduce_avg_2153"(#loc2003)) +#loc4094 = loc("add_2154"(#loc2004)) +#loc4095 = loc("sqrt_2155"(#loc2005)) +#loc4096 = loc("reciprocal_2156"(#loc2006)) +#loc4097 = loc("multiply_2157"(#loc2007)) +#loc4098 = loc("multiply_2158"(#loc2008)) +#loc4099 = loc("reshape_2159.dc.squeeze.0"(#loc2009)) +#loc4100 = loc("matmul_2161"(#loc2010)) +#loc4101 = loc("reshape_2162"(#loc2011)) +#loc4102 = loc("transpose_2163"(#loc2012)) +#loc4103 = loc("concatenate_2170"(#loc2013)) +#loc4104 = loc("cosine_2171"(#loc2014)) +#loc4105 = loc("unsqueeze_2172"(#loc2015)) +#loc4106 = loc("multiply_2173"(#loc2016)) +#loc4107 = loc("index_2174.dc.transpose.0"(#loc2017)) +#loc4108 = loc("index_2174.dc.matmul.2"(#loc2018)) +#loc4109 = loc("index_2174.dc.transpose.3"(#loc2019)) +#loc4110 = loc("multiply_2175"(#loc2020)) +#loc4111 = loc("index_2176.dc.transpose.0"(#loc2021)) +#loc4112 = loc("index_2176.dc.matmul.2"(#loc2022)) +#loc4113 = loc("index_2176.dc.transpose.3"(#loc2023)) +#loc4114 = loc("concatenate_2177"(#loc2024)) +#loc4115 = loc("sine_2178"(#loc2025)) +#loc4116 = loc("unsqueeze_2179"(#loc2026)) +#loc4117 = loc("multiply_2180"(#loc2027)) +#loc4118 = loc("add_2181"(#loc2028)) +#loc4119 = loc("reshape_2182.dc.squeeze.0"(#loc2029)) +#loc4120 = loc("matmul_2184"(#loc2030)) +#loc4121 = loc("reshape_2185"(#loc2031)) +#loc4122 = loc("transpose_2186"(#loc2032)) +#loc4123 = loc("multiply_2187"(#loc2033)) +#loc4124 = loc("index_2188.dc.transpose.0"(#loc2034)) +#loc4125 = loc("index_2188.dc.matmul.2"(#loc2035)) +#loc4126 = loc("index_2188.dc.transpose.3"(#loc2036)) +#loc4127 = loc("multiply_2189"(#loc2037)) +#loc4128 = loc("index_2190.dc.transpose.0"(#loc2038)) +#loc4129 = loc("index_2190.dc.matmul.2"(#loc2039)) +#loc4130 = loc("index_2190.dc.transpose.3"(#loc2040)) +#loc4131 = loc("concatenate_2191"(#loc2041)) +#loc4132 = loc("multiply_2192"(#loc2042)) +#loc4133 = loc("add_2193"(#loc2043)) +#loc4134 = loc("reshape_2194.dc.squeeze.0"(#loc2044)) +#loc4135 = loc("transpose_2195"(#loc2045)) +#loc4136 = loc("matmul_2196"(#loc2046)) +#loc4137 = loc("reshape_2197.dc.unsqueeze.0"(#loc2047)) +#loc4138 = loc("multiply_2198"(#loc2048)) +#loc4139 = loc("add_2199"(#loc2049)) +#loc4140 = loc("softmax_2200"(#loc2050)) +#loc4141 = loc("reshape_2202.dc.squeeze.0"(#loc2051)) +#loc4142 = loc("matmul_2204"(#loc2052)) +#loc4143 = loc("reshape_2205"(#loc2053)) +#loc4144 = loc("transpose_2206"(#loc2054)) +#loc4145 = loc("transpose_2207"(#loc2055)) +#loc4146 = loc("reshape_2208.dc.squeeze.0"(#loc2056)) +#loc4147 = loc("transpose_2209"(#loc2057)) +#loc4148 = loc("matmul_2210"(#loc2058)) +#loc4149 = loc("reshape_2211.dc.unsqueeze.0"(#loc2059)) +#loc4150 = loc("transpose_2212"(#loc2060)) +#loc4151 = loc("reshape_2213"(#loc2061)) +#loc4152 = loc("matmul_2215"(#loc2062)) +#loc4153 = loc("reshape_2216.dc.unsqueeze.0"(#loc2063)) +#loc4154 = loc("add_2217"(#loc2064)) +#loc4155 = loc("multiply_2218"(#loc2065)) +#loc4156 = loc("reduce_avg_2219"(#loc2066)) +#loc4157 = loc("add_2220"(#loc2067)) +#loc4158 = loc("sqrt_2221"(#loc2068)) +#loc4159 = loc("reciprocal_2222"(#loc2069)) +#loc4160 = loc("multiply_2223"(#loc2070)) +#loc4161 = loc("multiply_2224"(#loc2071)) +#loc4162 = loc("reshape_2225.dc.squeeze.0"(#loc2072)) +#loc4163 = loc("matmul_2227"(#loc2073)) +#loc4164 = loc("reshape_2228.dc.unsqueeze.0"(#loc2074)) +#loc4165 = loc("sigmoid_2229"(#loc2075)) +#loc4166 = loc("multiply_2230"(#loc2076)) +#loc4167 = loc("matmul_2232"(#loc2077)) +#loc4168 = loc("reshape_2233.dc.unsqueeze.0"(#loc2078)) +#loc4169 = loc("multiply_2234"(#loc2079)) +#loc4170 = loc("matmul_2236"(#loc2080)) +#loc4171 = loc("add_2237"(#loc2081)) +#loc4172 = loc("multiply_2238"(#loc2082)) +#loc4173 = loc("reduce_avg_2239"(#loc2083)) +#loc4174 = loc("add_2240"(#loc2084)) +#loc4175 = loc("sqrt_2241"(#loc2085)) +#loc4176 = loc("reciprocal_2242"(#loc2086)) +#loc4177 = loc("multiply_2243"(#loc2087)) +#loc4178 = loc("multiply_2244"(#loc2088)) +#loc4179 = loc("matmul_2246"(#loc2089)) diff --git a/tools/explorer/test/models/resnet_ttir.mlir b/tools/explorer/test/models/resnet_ttir.mlir new file mode 100644 index 000000000..e94c3d83d --- /dev/null +++ b/tools/explorer/test/models/resnet_ttir.mlir @@ -0,0 +1,1788 @@ +#any_device = #tt.operand_constraint +#loc = loc("ResNet":0:0) +module @ResNet attributes {} { + func.func @forward(%arg0: tensor<1x3x224x224xf32> {ttir.name = "input_1"} loc("ResNet":0:0), %arg1: tensor<64x1x1xf32> {ttir.name = "input_1_add_1"} loc("ResNet":0:0), %arg2: tensor<64x1x1xf32> {ttir.name = "input_1_add_1_fork_clone1229"} loc("ResNet":0:0), %arg3: tensor<64x1x1xf32> {ttir.name = "input_1_add_18"} loc("ResNet":0:0), %arg4: tensor<64x1x1xf32> {ttir.name = "input_1_add_18_fork_clone1271"} loc("ResNet":0:0), %arg5: tensor<64x1x1xf32> {ttir.name = "input_1_add_34"} loc("ResNet":0:0), %arg6: tensor<64x1x1xf32> {ttir.name = "input_1_add_34_fork_clone1204"} loc("ResNet":0:0), %arg7: tensor<256x1x1xf32> {ttir.name = "input_1_add_50"} loc("ResNet":0:0), %arg8: tensor<256x1x1xf32> {ttir.name = "input_1_add_50_fork_clone1108"} loc("ResNet":0:0), %arg9: tensor<256x1x1xf32> {ttir.name = "input_1_add_65"} loc("ResNet":0:0), %arg10: tensor<256x1x1xf32> {ttir.name = "input_1_add_65_fork_clone1112"} loc("ResNet":0:0), %arg11: tensor<64x1x1xf32> {ttir.name = "input_1_add_82"} loc("ResNet":0:0), %arg12: tensor<64x1x1xf32> {ttir.name = "input_1_add_82_fork_clone1238"} loc("ResNet":0:0), %arg13: tensor<64x1x1xf32> {ttir.name = "input_1_add_98"} loc("ResNet":0:0), %arg14: tensor<64x1x1xf32> {ttir.name = "input_1_add_98_fork_clone1152"} loc("ResNet":0:0), %arg15: tensor<256x1x1xf32> {ttir.name = "input_1_add_114"} loc("ResNet":0:0), %arg16: tensor<256x1x1xf32> {ttir.name = "input_1_add_114_fork_clone1051"} loc("ResNet":0:0), %arg17: tensor<64x1x1xf32> {ttir.name = "input_1_add_131"} loc("ResNet":0:0), %arg18: tensor<64x1x1xf32> {ttir.name = "input_1_add_131_fork_clone1192"} loc("ResNet":0:0), %arg19: tensor<64x1x1xf32> {ttir.name = "input_1_add_147"} loc("ResNet":0:0), %arg20: tensor<64x1x1xf32> {ttir.name = "input_1_add_147_fork_clone1096"} loc("ResNet":0:0), %arg21: tensor<256x1x1xf32> {ttir.name = "input_1_add_163"} loc("ResNet":0:0), %arg22: tensor<256x1x1xf32> {ttir.name = "input_1_add_163_fork_clone992"} loc("ResNet":0:0), %arg23: tensor<128x1x1xf32> {ttir.name = "input_1_add_180"} loc("ResNet":0:0), %arg24: tensor<128x1x1xf32> {ttir.name = "input_1_add_180_fork_clone1065"} loc("ResNet":0:0), %arg25: tensor<128x1x1xf32> {ttir.name = "input_1_add_196"} loc("ResNet":0:0), %arg26: tensor<128x1x1xf32> {ttir.name = "input_1_add_196_fork_clone962"} loc("ResNet":0:0), %arg27: tensor<512x1x1xf32> {ttir.name = "input_1_add_212"} loc("ResNet":0:0), %arg28: tensor<512x1x1xf32> {ttir.name = "input_1_add_212_fork_clone853"} loc("ResNet":0:0), %arg29: tensor<512x1x1xf32> {ttir.name = "input_1_add_227"} loc("ResNet":0:0), %arg30: tensor<512x1x1xf32> {ttir.name = "input_1_add_227_fork_clone857"} loc("ResNet":0:0), %arg31: tensor<128x1x1xf32> {ttir.name = "input_1_add_244"} loc("ResNet":0:0), %arg32: tensor<128x1x1xf32> {ttir.name = "input_1_add_244_fork_clone1007"} loc("ResNet":0:0), %arg33: tensor<128x1x1xf32> {ttir.name = "input_1_add_260"} loc("ResNet":0:0), %arg34: tensor<128x1x1xf32> {ttir.name = "input_1_add_260_fork_clone901"} loc("ResNet":0:0), %arg35: tensor<512x1x1xf32> {ttir.name = "input_1_add_276"} loc("ResNet":0:0), %arg36: tensor<512x1x1xf32> {ttir.name = "input_1_add_276_fork_clone791"} loc("ResNet":0:0), %arg37: tensor<128x1x1xf32> {ttir.name = "input_1_add_293"} loc("ResNet":0:0), %arg38: tensor<128x1x1xf32> {ttir.name = "input_1_add_293_fork_clone950"} loc("ResNet":0:0), %arg39: tensor<128x1x1xf32> {ttir.name = "input_1_add_309"} loc("ResNet":0:0), %arg40: tensor<128x1x1xf32> {ttir.name = "input_1_add_309_fork_clone841"} loc("ResNet":0:0), %arg41: tensor<512x1x1xf32> {ttir.name = "input_1_add_325"} loc("ResNet":0:0), %arg42: tensor<512x1x1xf32> {ttir.name = "input_1_add_325_fork_clone735"} loc("ResNet":0:0), %arg43: tensor<128x1x1xf32> {ttir.name = "input_1_add_342"} loc("ResNet":0:0), %arg44: tensor<128x1x1xf32> {ttir.name = "input_1_add_342_fork_clone889"} loc("ResNet":0:0), %arg45: tensor<128x1x1xf32> {ttir.name = "input_1_add_358"} loc("ResNet":0:0), %arg46: tensor<128x1x1xf32> {ttir.name = "input_1_add_358_fork_clone779"} loc("ResNet":0:0), %arg47: tensor<512x1x1xf32> {ttir.name = "input_1_add_374"} loc("ResNet":0:0), %arg48: tensor<512x1x1xf32> {ttir.name = "input_1_add_374_fork_clone677"} loc("ResNet":0:0), %arg49: tensor<256x1x1xf32> {ttir.name = "input_1_add_391"} loc("ResNet":0:0), %arg50: tensor<256x1x1xf32> {ttir.name = "input_1_add_391_fork_clone748"} loc("ResNet":0:0), %arg51: tensor<256x1x1xf32> {ttir.name = "input_1_add_407"} loc("ResNet":0:0), %arg52: tensor<256x1x1xf32> {ttir.name = "input_1_add_407_fork_clone645"} loc("ResNet":0:0), %arg53: tensor<1024x1x1xf32> {ttir.name = "input_1_add_423"} loc("ResNet":0:0), %arg54: tensor<1024x1x1xf32> {ttir.name = "input_1_add_423_fork_clone524"} loc("ResNet":0:0), %arg55: tensor<1024x1x1xf32> {ttir.name = "input_1_add_438"} loc("ResNet":0:0), %arg56: tensor<1024x1x1xf32> {ttir.name = "input_1_add_438_fork_clone528"} loc("ResNet":0:0), %arg57: tensor<256x1x1xf32> {ttir.name = "input_1_add_455"} loc("ResNet":0:0), %arg58: tensor<256x1x1xf32> {ttir.name = "input_1_add_455_fork_clone692"} loc("ResNet":0:0), %arg59: tensor<256x1x1xf32> {ttir.name = "input_1_add_471"} loc("ResNet":0:0), %arg60: tensor<256x1x1xf32> {ttir.name = "input_1_add_471_fork_clone580"} loc("ResNet":0:0), %arg61: tensor<1024x1x1xf32> {ttir.name = "input_1_add_487"} loc("ResNet":0:0), %arg62: tensor<1024x1x1xf32> {ttir.name = "input_1_add_487_fork_clone453"} loc("ResNet":0:0), %arg63: tensor<256x1x1xf32> {ttir.name = "input_1_add_504"} loc("ResNet":0:0), %arg64: tensor<256x1x1xf32> {ttir.name = "input_1_add_504_fork_clone633"} loc("ResNet":0:0), %arg65: tensor<256x1x1xf32> {ttir.name = "input_1_add_520"} loc("ResNet":0:0), %arg66: tensor<256x1x1xf32> {ttir.name = "input_1_add_520_fork_clone512"} loc("ResNet":0:0), %arg67: tensor<1024x1x1xf32> {ttir.name = "input_1_add_536"} loc("ResNet":0:0), %arg68: tensor<1024x1x1xf32> {ttir.name = "input_1_add_536_fork_clone389"} loc("ResNet":0:0), %arg69: tensor<256x1x1xf32> {ttir.name = "input_1_add_553"} loc("ResNet":0:0), %arg70: tensor<256x1x1xf32> {ttir.name = "input_1_add_553_fork_clone568"} loc("ResNet":0:0), %arg71: tensor<256x1x1xf32> {ttir.name = "input_1_add_569"} loc("ResNet":0:0), %arg72: tensor<256x1x1xf32> {ttir.name = "input_1_add_569_fork_clone441"} loc("ResNet":0:0), %arg73: tensor<1024x1x1xf32> {ttir.name = "input_1_add_585"} loc("ResNet":0:0), %arg74: tensor<1024x1x1xf32> {ttir.name = "input_1_add_585_fork_clone329"} loc("ResNet":0:0), %arg75: tensor<256x1x1xf32> {ttir.name = "input_1_add_602"} loc("ResNet":0:0), %arg76: tensor<256x1x1xf32> {ttir.name = "input_1_add_602_fork_clone500"} loc("ResNet":0:0), %arg77: tensor<256x1x1xf32> {ttir.name = "input_1_add_618"} loc("ResNet":0:0), %arg78: tensor<256x1x1xf32> {ttir.name = "input_1_add_618_fork_clone377"} loc("ResNet":0:0), %arg79: tensor<1024x1x1xf32> {ttir.name = "input_1_add_634"} loc("ResNet":0:0), %arg80: tensor<1024x1x1xf32> {ttir.name = "input_1_add_634_fork_clone274"} loc("ResNet":0:0), %arg81: tensor<256x1x1xf32> {ttir.name = "input_1_add_651"} loc("ResNet":0:0), %arg82: tensor<256x1x1xf32> {ttir.name = "input_1_add_651_fork_clone429"} loc("ResNet":0:0), %arg83: tensor<256x1x1xf32> {ttir.name = "input_1_add_667"} loc("ResNet":0:0), %arg84: tensor<256x1x1xf32> {ttir.name = "input_1_add_667_fork_clone317"} loc("ResNet":0:0), %arg85: tensor<1024x1x1xf32> {ttir.name = "input_1_add_683"} loc("ResNet":0:0), %arg86: tensor<1024x1x1xf32> {ttir.name = "input_1_add_683_fork_clone219"} loc("ResNet":0:0), %arg87: tensor<512x1x1xf32> {ttir.name = "input_1_add_700"} loc("ResNet":0:0), %arg88: tensor<512x1x1xf32> {ttir.name = "input_1_add_700_fork_clone287"} loc("ResNet":0:0), %arg89: tensor<512x1x1xf32> {ttir.name = "input_1_add_716"} loc("ResNet":0:0), %arg90: tensor<512x1x1xf32> {ttir.name = "input_1_add_716_fork_clone190"} loc("ResNet":0:0), %arg91: tensor<2048x1x1xf32> {ttir.name = "input_1_add_732"} loc("ResNet":0:0), %arg92: tensor<2048x1x1xf32> {ttir.name = "input_1_add_732_fork_clone101"} loc("ResNet":0:0), %arg93: tensor<2048x1x1xf32> {ttir.name = "input_1_add_747"} loc("ResNet":0:0), %arg94: tensor<2048x1x1xf32> {ttir.name = "input_1_add_747_fork_clone105"} loc("ResNet":0:0), %arg95: tensor<512x1x1xf32> {ttir.name = "input_1_add_764"} loc("ResNet":0:0), %arg96: tensor<512x1x1xf32> {ttir.name = "input_1_add_764_fork_clone233"} loc("ResNet":0:0), %arg97: tensor<512x1x1xf32> {ttir.name = "input_1_add_780"} loc("ResNet":0:0), %arg98: tensor<512x1x1xf32> {ttir.name = "input_1_add_780_fork_clone138"} loc("ResNet":0:0), %arg99: tensor<2048x1x1xf32> {ttir.name = "input_1_add_796"} loc("ResNet":0:0), %arg100: tensor<2048x1x1xf32> {ttir.name = "input_1_add_796_fork_clone61"} loc("ResNet":0:0), %arg101: tensor<512x1x1xf32> {ttir.name = "input_1_add_813"} loc("ResNet":0:0), %arg102: tensor<512x1x1xf32> {ttir.name = "input_1_add_813_fork_clone178"} loc("ResNet":0:0), %arg103: tensor<512x1x1xf32> {ttir.name = "input_1_add_829"} loc("ResNet":0:0), %arg104: tensor<512x1x1xf32> {ttir.name = "input_1_add_829_fork_clone89"} loc("ResNet":0:0), %arg105: tensor<2048x1x1xf32> {ttir.name = "input_1_add_845"} loc("ResNet":0:0), %arg106: tensor<2048x1x1xf32> {ttir.name = "input_1_add_845_fork_clone32"} loc("ResNet":0:0), %arg107: tensor<64x3x7x7xf32> {ttir.name = "conv1.weight"} loc("ResNet":0:0), %arg108: tensor<64x64x1x1xf32> {ttir.name = "layer1.0.conv1.weight"} loc("ResNet":0:0), %arg109: tensor<64x64x3x3xf32> {ttir.name = "layer1.0.conv2.weight"} loc("ResNet":0:0), %arg110: tensor<256x64x1x1xf32> {ttir.name = "layer1.0.conv3.weight"} loc("ResNet":0:0), %arg111: tensor<256x64x1x1xf32> {ttir.name = "layer1.0.downsample.0.weight"} loc("ResNet":0:0), %arg112: tensor<64x256x1x1xf32> {ttir.name = "layer1.1.conv1.weight"} loc("ResNet":0:0), %arg113: tensor<64x64x3x3xf32> {ttir.name = "layer1.1.conv2.weight"} loc("ResNet":0:0), %arg114: tensor<256x64x1x1xf32> {ttir.name = "layer1.1.conv3.weight"} loc("ResNet":0:0), %arg115: tensor<64x256x1x1xf32> {ttir.name = "layer1.2.conv1.weight"} loc("ResNet":0:0), %arg116: tensor<64x64x3x3xf32> {ttir.name = "layer1.2.conv2.weight"} loc("ResNet":0:0), %arg117: tensor<256x64x1x1xf32> {ttir.name = "layer1.2.conv3.weight"} loc("ResNet":0:0), %arg118: tensor<128x256x1x1xf32> {ttir.name = "layer2.0.conv1.weight"} loc("ResNet":0:0), %arg119: tensor<128x128x3x3xf32> {ttir.name = "layer2.0.conv2.weight"} loc("ResNet":0:0), %arg120: tensor<512x128x1x1xf32> {ttir.name = "layer2.0.conv3.weight"} loc("ResNet":0:0), %arg121: tensor<512x256x1x1xf32> {ttir.name = "layer2.0.downsample.0.weight"} loc("ResNet":0:0), %arg122: tensor<128x512x1x1xf32> {ttir.name = "layer2.1.conv1.weight"} loc("ResNet":0:0), %arg123: tensor<128x128x3x3xf32> {ttir.name = "layer2.1.conv2.weight"} loc("ResNet":0:0), %arg124: tensor<512x128x1x1xf32> {ttir.name = "layer2.1.conv3.weight"} loc("ResNet":0:0), %arg125: tensor<128x512x1x1xf32> {ttir.name = "layer2.2.conv1.weight"} loc("ResNet":0:0), %arg126: tensor<128x128x3x3xf32> {ttir.name = "layer2.2.conv2.weight"} loc("ResNet":0:0), %arg127: tensor<512x128x1x1xf32> {ttir.name = "layer2.2.conv3.weight"} loc("ResNet":0:0), %arg128: tensor<128x512x1x1xf32> {ttir.name = "layer2.3.conv1.weight"} loc("ResNet":0:0), %arg129: tensor<128x128x3x3xf32> {ttir.name = "layer2.3.conv2.weight"} loc("ResNet":0:0), %arg130: tensor<512x128x1x1xf32> {ttir.name = "layer2.3.conv3.weight"} loc("ResNet":0:0), %arg131: tensor<256x512x1x1xf32> {ttir.name = "layer3.0.conv1.weight"} loc("ResNet":0:0), %arg132: tensor<256x256x3x3xf32> {ttir.name = "layer3.0.conv2.weight"} loc("ResNet":0:0), %arg133: tensor<1024x256x1x1xf32> {ttir.name = "layer3.0.conv3.weight"} loc("ResNet":0:0), %arg134: tensor<1024x512x1x1xf32> {ttir.name = "layer3.0.downsample.0.weight"} loc("ResNet":0:0), %arg135: tensor<256x1024x1x1xf32> {ttir.name = "layer3.1.conv1.weight"} loc("ResNet":0:0), %arg136: tensor<256x256x3x3xf32> {ttir.name = "layer3.1.conv2.weight"} loc("ResNet":0:0), %arg137: tensor<1024x256x1x1xf32> {ttir.name = "layer3.1.conv3.weight"} loc("ResNet":0:0), %arg138: tensor<256x1024x1x1xf32> {ttir.name = "layer3.2.conv1.weight"} loc("ResNet":0:0), %arg139: tensor<256x256x3x3xf32> {ttir.name = "layer3.2.conv2.weight"} loc("ResNet":0:0), %arg140: tensor<1024x256x1x1xf32> {ttir.name = "layer3.2.conv3.weight"} loc("ResNet":0:0), %arg141: tensor<256x1024x1x1xf32> {ttir.name = "layer3.3.conv1.weight"} loc("ResNet":0:0), %arg142: tensor<256x256x3x3xf32> {ttir.name = "layer3.3.conv2.weight"} loc("ResNet":0:0), %arg143: tensor<1024x256x1x1xf32> {ttir.name = "layer3.3.conv3.weight"} loc("ResNet":0:0), %arg144: tensor<256x1024x1x1xf32> {ttir.name = "layer3.4.conv1.weight"} loc("ResNet":0:0), %arg145: tensor<256x256x3x3xf32> {ttir.name = "layer3.4.conv2.weight"} loc("ResNet":0:0), %arg146: tensor<1024x256x1x1xf32> {ttir.name = "layer3.4.conv3.weight"} loc("ResNet":0:0), %arg147: tensor<256x1024x1x1xf32> {ttir.name = "layer3.5.conv1.weight"} loc("ResNet":0:0), %arg148: tensor<256x256x3x3xf32> {ttir.name = "layer3.5.conv2.weight"} loc("ResNet":0:0), %arg149: tensor<1024x256x1x1xf32> {ttir.name = "layer3.5.conv3.weight"} loc("ResNet":0:0), %arg150: tensor<512x1024x1x1xf32> {ttir.name = "layer4.0.conv1.weight"} loc("ResNet":0:0), %arg151: tensor<512x512x3x3xf32> {ttir.name = "layer4.0.conv2.weight"} loc("ResNet":0:0), %arg152: tensor<2048x512x1x1xf32> {ttir.name = "layer4.0.conv3.weight"} loc("ResNet":0:0), %arg153: tensor<2048x1024x1x1xf32> {ttir.name = "layer4.0.downsample.0.weight"} loc("ResNet":0:0), %arg154: tensor<512x2048x1x1xf32> {ttir.name = "layer4.1.conv1.weight"} loc("ResNet":0:0), %arg155: tensor<512x512x3x3xf32> {ttir.name = "layer4.1.conv2.weight"} loc("ResNet":0:0), %arg156: tensor<2048x512x1x1xf32> {ttir.name = "layer4.1.conv3.weight"} loc("ResNet":0:0), %arg157: tensor<512x2048x1x1xf32> {ttir.name = "layer4.2.conv1.weight"} loc("ResNet":0:0), %arg158: tensor<512x512x3x3xf32> {ttir.name = "layer4.2.conv2.weight"} loc("ResNet":0:0), %arg159: tensor<2048x512x1x1xf32> {ttir.name = "layer4.2.conv3.weight"} loc("ResNet":0:0), %arg160: tensor<2048x1000xf32> {ttir.name = "fc.weight"} loc("ResNet":0:0), %arg161: tensor<1000xf32> {ttir.name = "fc.bias"} loc("ResNet":0:0)) -> (tensor<1x1000xf32> {ttir.name = "ResNet.output_add_867"}) { + %0 = tensor.empty() : tensor<1x224x3x224xf32> loc(#loc447) + %1 = "ttir.transpose"(%arg0, %0) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x3x224x224xf32>, tensor<1x224x3x224xf32>) -> tensor<1x224x3x224xf32> loc(#loc447) + %2 = tensor.empty() : tensor<1x224x224x3xf32> loc(#loc448) + %3 = "ttir.transpose"(%1, %2) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x224x3x224xf32>, tensor<1x224x224x3xf32>) -> tensor<1x224x224x3xf32> loc(#loc448) + %4 = tensor.empty() : tensor<1x112x112x64xf32> loc(#loc449) + %5 = "ttir.conv2d"(%3, %arg107, %4) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 3 : si32, padding_left = 3 : si32, padding_right = 3 : si32, padding_top = 3 : si32, stride_height = 2 : si32, stride_width = 2 : si32}> {channel_last = 1 : si32} : (tensor<1x224x224x3xf32>, tensor<64x3x7x7xf32>, tensor<1x112x112x64xf32>) -> tensor<1x112x112x64xf32> loc(#loc449) + %6 = tensor.empty() : tensor<1x112x64x112xf32> loc(#loc450) + %7 = "ttir.transpose"(%5, %6) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x112x112x64xf32>, tensor<1x112x64x112xf32>) -> tensor<1x112x64x112xf32> loc(#loc450) + %8 = tensor.empty() : tensor<1x64x112x112xf32> loc(#loc451) + %9 = "ttir.transpose"(%7, %8) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x112x64x112xf32>, tensor<1x64x112x112xf32>) -> tensor<1x64x112x112xf32> loc(#loc451) + %10 = tensor.empty() : tensor<1x64x112x112xf32> loc(#loc452) + %11 = "ttir.multiply"(%9, %arg1, %10) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x64x112x112xf32>, tensor<64x1x1xf32>, tensor<1x64x112x112xf32>) -> tensor<1x64x112x112xf32> loc(#loc452) + %12 = tensor.empty() : tensor<1x64x112x112xf32> loc(#loc453) + %13 = "ttir.add"(%11, %arg2, %12) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x64x112x112xf32>, tensor<64x1x1xf32>, tensor<1x64x112x112xf32>) -> tensor<1x64x112x112xf32> loc(#loc453) + %14 = tensor.empty() : tensor<1x64x112x112xf32> loc(#loc454) + %15 = "ttir.relu"(%13, %14) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x64x112x112xf32>, tensor<1x64x112x112xf32>) -> tensor<1x64x112x112xf32> loc(#loc454) + %16 = tensor.empty() : tensor<1x64x56x56xf32> loc(#loc455) + %17 = "ttir.max_pool2d"(%15, %16) <{ceil_mode = false, dilation_height = 1 : si32, dilation_width = 1 : si32, kernel_height = 3 : si32, kernel_width = 3 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 1 : si32, padding_left = 1 : si32, padding_right = 1 : si32, padding_top = 1 : si32, stride_height = 2 : si32, stride_width = 2 : si32}> {channel_last = 0 : si32} : (tensor<1x64x112x112xf32>, tensor<1x64x56x56xf32>) -> tensor<1x64x56x56xf32> loc(#loc455) + %18 = tensor.empty() : tensor<1x56x64x56xf32> loc(#loc456) + %19 = "ttir.transpose"(%17, %18) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<1x56x64x56xf32>) -> tensor<1x56x64x56xf32> loc(#loc456) + %20 = tensor.empty() : tensor<1x56x56x64xf32> loc(#loc457) + %21 = "ttir.transpose"(%19, %20) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x64x56xf32>, tensor<1x56x56x64xf32>) -> tensor<1x56x56x64xf32> loc(#loc457) + %22 = tensor.empty() : tensor<1x56x56x64xf32> loc(#loc458) + %23 = "ttir.conv2d"(%21, %arg108, %22) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x56x56x64xf32>, tensor<64x64x1x1xf32>, tensor<1x56x56x64xf32>) -> tensor<1x56x56x64xf32> loc(#loc458) + %24 = tensor.empty() : tensor<1x56x64x56xf32> loc(#loc459) + %25 = "ttir.transpose"(%23, %24) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x56x64xf32>, tensor<1x56x64x56xf32>) -> tensor<1x56x64x56xf32> loc(#loc459) + %26 = tensor.empty() : tensor<1x64x56x56xf32> loc(#loc460) + %27 = "ttir.transpose"(%25, %26) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x64x56xf32>, tensor<1x64x56x56xf32>) -> tensor<1x64x56x56xf32> loc(#loc460) + %28 = tensor.empty() : tensor<1x64x56x56xf32> loc(#loc461) + %29 = "ttir.multiply"(%27, %arg3, %28) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<64x1x1xf32>, tensor<1x64x56x56xf32>) -> tensor<1x64x56x56xf32> loc(#loc461) + %30 = tensor.empty() : tensor<1x64x56x56xf32> loc(#loc462) + %31 = "ttir.add"(%29, %arg4, %30) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<64x1x1xf32>, tensor<1x64x56x56xf32>) -> tensor<1x64x56x56xf32> loc(#loc462) + %32 = tensor.empty() : tensor<1x64x56x56xf32> loc(#loc463) + %33 = "ttir.relu"(%31, %32) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<1x64x56x56xf32>) -> tensor<1x64x56x56xf32> loc(#loc463) + %34 = tensor.empty() : tensor<1x56x64x56xf32> loc(#loc464) + %35 = "ttir.transpose"(%33, %34) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<1x56x64x56xf32>) -> tensor<1x56x64x56xf32> loc(#loc464) + %36 = tensor.empty() : tensor<1x56x56x64xf32> loc(#loc465) + %37 = "ttir.transpose"(%35, %36) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x64x56xf32>, tensor<1x56x56x64xf32>) -> tensor<1x56x56x64xf32> loc(#loc465) + %38 = tensor.empty() : tensor<1x56x56x64xf32> loc(#loc466) + %39 = "ttir.conv2d"(%37, %arg109, %38) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 1 : si32, padding_left = 1 : si32, padding_right = 1 : si32, padding_top = 1 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x56x56x64xf32>, tensor<64x64x3x3xf32>, tensor<1x56x56x64xf32>) -> tensor<1x56x56x64xf32> loc(#loc466) + %40 = tensor.empty() : tensor<1x56x64x56xf32> loc(#loc467) + %41 = "ttir.transpose"(%39, %40) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x56x64xf32>, tensor<1x56x64x56xf32>) -> tensor<1x56x64x56xf32> loc(#loc467) + %42 = tensor.empty() : tensor<1x64x56x56xf32> loc(#loc468) + %43 = "ttir.transpose"(%41, %42) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x64x56xf32>, tensor<1x64x56x56xf32>) -> tensor<1x64x56x56xf32> loc(#loc468) + %44 = tensor.empty() : tensor<1x64x56x56xf32> loc(#loc469) + %45 = "ttir.multiply"(%43, %arg5, %44) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<64x1x1xf32>, tensor<1x64x56x56xf32>) -> tensor<1x64x56x56xf32> loc(#loc469) + %46 = tensor.empty() : tensor<1x64x56x56xf32> loc(#loc470) + %47 = "ttir.add"(%45, %arg6, %46) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<64x1x1xf32>, tensor<1x64x56x56xf32>) -> tensor<1x64x56x56xf32> loc(#loc470) + %48 = tensor.empty() : tensor<1x64x56x56xf32> loc(#loc471) + %49 = "ttir.relu"(%47, %48) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<1x64x56x56xf32>) -> tensor<1x64x56x56xf32> loc(#loc471) + %50 = tensor.empty() : tensor<1x56x64x56xf32> loc(#loc472) + %51 = "ttir.transpose"(%49, %50) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<1x56x64x56xf32>) -> tensor<1x56x64x56xf32> loc(#loc472) + %52 = tensor.empty() : tensor<1x56x56x64xf32> loc(#loc473) + %53 = "ttir.transpose"(%51, %52) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x64x56xf32>, tensor<1x56x56x64xf32>) -> tensor<1x56x56x64xf32> loc(#loc473) + %54 = tensor.empty() : tensor<1x56x56x256xf32> loc(#loc474) + %55 = "ttir.conv2d"(%53, %arg110, %54) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x56x56x64xf32>, tensor<256x64x1x1xf32>, tensor<1x56x56x256xf32>) -> tensor<1x56x56x256xf32> loc(#loc474) + %56 = tensor.empty() : tensor<1x56x256x56xf32> loc(#loc475) + %57 = "ttir.transpose"(%55, %56) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x56x256xf32>, tensor<1x56x256x56xf32>) -> tensor<1x56x256x56xf32> loc(#loc475) + %58 = tensor.empty() : tensor<1x256x56x56xf32> loc(#loc476) + %59 = "ttir.transpose"(%57, %58) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x256x56xf32>, tensor<1x256x56x56xf32>) -> tensor<1x256x56x56xf32> loc(#loc476) + %60 = tensor.empty() : tensor<1x256x56x56xf32> loc(#loc477) + %61 = "ttir.multiply"(%59, %arg7, %60) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x56x56xf32>, tensor<256x1x1xf32>, tensor<1x256x56x56xf32>) -> tensor<1x256x56x56xf32> loc(#loc477) + %62 = tensor.empty() : tensor<1x256x56x56xf32> loc(#loc478) + %63 = "ttir.add"(%61, %arg8, %62) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x56x56xf32>, tensor<256x1x1xf32>, tensor<1x256x56x56xf32>) -> tensor<1x256x56x56xf32> loc(#loc478) + %64 = tensor.empty() : tensor<1x56x64x56xf32> loc(#loc479) + %65 = "ttir.transpose"(%17, %64) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<1x56x64x56xf32>) -> tensor<1x56x64x56xf32> loc(#loc479) + %66 = tensor.empty() : tensor<1x56x56x64xf32> loc(#loc480) + %67 = "ttir.transpose"(%65, %66) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x64x56xf32>, tensor<1x56x56x64xf32>) -> tensor<1x56x56x64xf32> loc(#loc480) + %68 = tensor.empty() : tensor<1x56x56x256xf32> loc(#loc481) + %69 = "ttir.conv2d"(%67, %arg111, %68) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x56x56x64xf32>, tensor<256x64x1x1xf32>, tensor<1x56x56x256xf32>) -> tensor<1x56x56x256xf32> loc(#loc481) + %70 = tensor.empty() : tensor<1x56x256x56xf32> loc(#loc482) + %71 = "ttir.transpose"(%69, %70) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x56x256xf32>, tensor<1x56x256x56xf32>) -> tensor<1x56x256x56xf32> loc(#loc482) + %72 = tensor.empty() : tensor<1x256x56x56xf32> loc(#loc483) + %73 = "ttir.transpose"(%71, %72) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x256x56xf32>, tensor<1x256x56x56xf32>) -> tensor<1x256x56x56xf32> loc(#loc483) + %74 = tensor.empty() : tensor<1x256x56x56xf32> loc(#loc484) + %75 = "ttir.multiply"(%73, %arg9, %74) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x56x56xf32>, tensor<256x1x1xf32>, tensor<1x256x56x56xf32>) -> tensor<1x256x56x56xf32> loc(#loc484) + %76 = tensor.empty() : tensor<1x256x56x56xf32> loc(#loc485) + %77 = "ttir.add"(%75, %arg10, %76) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x56x56xf32>, tensor<256x1x1xf32>, tensor<1x256x56x56xf32>) -> tensor<1x256x56x56xf32> loc(#loc485) + %78 = tensor.empty() : tensor<1x256x56x56xf32> loc(#loc486) + %79 = "ttir.add"(%63, %77, %78) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x56x56xf32>, tensor<1x256x56x56xf32>, tensor<1x256x56x56xf32>) -> tensor<1x256x56x56xf32> loc(#loc486) + %80 = tensor.empty() : tensor<1x256x56x56xf32> loc(#loc487) + %81 = "ttir.relu"(%79, %80) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x56x56xf32>, tensor<1x256x56x56xf32>) -> tensor<1x256x56x56xf32> loc(#loc487) + %82 = tensor.empty() : tensor<1x56x256x56xf32> loc(#loc488) + %83 = "ttir.transpose"(%81, %82) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x56x56xf32>, tensor<1x56x256x56xf32>) -> tensor<1x56x256x56xf32> loc(#loc488) + %84 = tensor.empty() : tensor<1x56x56x256xf32> loc(#loc489) + %85 = "ttir.transpose"(%83, %84) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x256x56xf32>, tensor<1x56x56x256xf32>) -> tensor<1x56x56x256xf32> loc(#loc489) + %86 = tensor.empty() : tensor<1x56x56x64xf32> loc(#loc490) + %87 = "ttir.conv2d"(%85, %arg112, %86) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x56x56x256xf32>, tensor<64x256x1x1xf32>, tensor<1x56x56x64xf32>) -> tensor<1x56x56x64xf32> loc(#loc490) + %88 = tensor.empty() : tensor<1x56x64x56xf32> loc(#loc491) + %89 = "ttir.transpose"(%87, %88) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x56x64xf32>, tensor<1x56x64x56xf32>) -> tensor<1x56x64x56xf32> loc(#loc491) + %90 = tensor.empty() : tensor<1x64x56x56xf32> loc(#loc492) + %91 = "ttir.transpose"(%89, %90) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x64x56xf32>, tensor<1x64x56x56xf32>) -> tensor<1x64x56x56xf32> loc(#loc492) + %92 = tensor.empty() : tensor<1x64x56x56xf32> loc(#loc493) + %93 = "ttir.multiply"(%91, %arg11, %92) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<64x1x1xf32>, tensor<1x64x56x56xf32>) -> tensor<1x64x56x56xf32> loc(#loc493) + %94 = tensor.empty() : tensor<1x64x56x56xf32> loc(#loc494) + %95 = "ttir.add"(%93, %arg12, %94) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<64x1x1xf32>, tensor<1x64x56x56xf32>) -> tensor<1x64x56x56xf32> loc(#loc494) + %96 = tensor.empty() : tensor<1x64x56x56xf32> loc(#loc495) + %97 = "ttir.relu"(%95, %96) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<1x64x56x56xf32>) -> tensor<1x64x56x56xf32> loc(#loc495) + %98 = tensor.empty() : tensor<1x56x64x56xf32> loc(#loc496) + %99 = "ttir.transpose"(%97, %98) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<1x56x64x56xf32>) -> tensor<1x56x64x56xf32> loc(#loc496) + %100 = tensor.empty() : tensor<1x56x56x64xf32> loc(#loc497) + %101 = "ttir.transpose"(%99, %100) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x64x56xf32>, tensor<1x56x56x64xf32>) -> tensor<1x56x56x64xf32> loc(#loc497) + %102 = tensor.empty() : tensor<1x56x56x64xf32> loc(#loc498) + %103 = "ttir.conv2d"(%101, %arg113, %102) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 1 : si32, padding_left = 1 : si32, padding_right = 1 : si32, padding_top = 1 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x56x56x64xf32>, tensor<64x64x3x3xf32>, tensor<1x56x56x64xf32>) -> tensor<1x56x56x64xf32> loc(#loc498) + %104 = tensor.empty() : tensor<1x56x64x56xf32> loc(#loc499) + %105 = "ttir.transpose"(%103, %104) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x56x64xf32>, tensor<1x56x64x56xf32>) -> tensor<1x56x64x56xf32> loc(#loc499) + %106 = tensor.empty() : tensor<1x64x56x56xf32> loc(#loc500) + %107 = "ttir.transpose"(%105, %106) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x64x56xf32>, tensor<1x64x56x56xf32>) -> tensor<1x64x56x56xf32> loc(#loc500) + %108 = tensor.empty() : tensor<1x64x56x56xf32> loc(#loc501) + %109 = "ttir.multiply"(%107, %arg13, %108) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<64x1x1xf32>, tensor<1x64x56x56xf32>) -> tensor<1x64x56x56xf32> loc(#loc501) + %110 = tensor.empty() : tensor<1x64x56x56xf32> loc(#loc502) + %111 = "ttir.add"(%109, %arg14, %110) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<64x1x1xf32>, tensor<1x64x56x56xf32>) -> tensor<1x64x56x56xf32> loc(#loc502) + %112 = tensor.empty() : tensor<1x64x56x56xf32> loc(#loc503) + %113 = "ttir.relu"(%111, %112) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<1x64x56x56xf32>) -> tensor<1x64x56x56xf32> loc(#loc503) + %114 = tensor.empty() : tensor<1x56x64x56xf32> loc(#loc504) + %115 = "ttir.transpose"(%113, %114) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<1x56x64x56xf32>) -> tensor<1x56x64x56xf32> loc(#loc504) + %116 = tensor.empty() : tensor<1x56x56x64xf32> loc(#loc505) + %117 = "ttir.transpose"(%115, %116) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x64x56xf32>, tensor<1x56x56x64xf32>) -> tensor<1x56x56x64xf32> loc(#loc505) + %118 = tensor.empty() : tensor<1x56x56x256xf32> loc(#loc506) + %119 = "ttir.conv2d"(%117, %arg114, %118) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x56x56x64xf32>, tensor<256x64x1x1xf32>, tensor<1x56x56x256xf32>) -> tensor<1x56x56x256xf32> loc(#loc506) + %120 = tensor.empty() : tensor<1x56x256x56xf32> loc(#loc507) + %121 = "ttir.transpose"(%119, %120) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x56x256xf32>, tensor<1x56x256x56xf32>) -> tensor<1x56x256x56xf32> loc(#loc507) + %122 = tensor.empty() : tensor<1x256x56x56xf32> loc(#loc508) + %123 = "ttir.transpose"(%121, %122) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x256x56xf32>, tensor<1x256x56x56xf32>) -> tensor<1x256x56x56xf32> loc(#loc508) + %124 = tensor.empty() : tensor<1x256x56x56xf32> loc(#loc509) + %125 = "ttir.multiply"(%123, %arg15, %124) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x56x56xf32>, tensor<256x1x1xf32>, tensor<1x256x56x56xf32>) -> tensor<1x256x56x56xf32> loc(#loc509) + %126 = tensor.empty() : tensor<1x256x56x56xf32> loc(#loc510) + %127 = "ttir.add"(%125, %arg16, %126) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x56x56xf32>, tensor<256x1x1xf32>, tensor<1x256x56x56xf32>) -> tensor<1x256x56x56xf32> loc(#loc510) + %128 = tensor.empty() : tensor<1x256x56x56xf32> loc(#loc511) + %129 = "ttir.add"(%127, %81, %128) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x56x56xf32>, tensor<1x256x56x56xf32>, tensor<1x256x56x56xf32>) -> tensor<1x256x56x56xf32> loc(#loc511) + %130 = tensor.empty() : tensor<1x256x56x56xf32> loc(#loc512) + %131 = "ttir.relu"(%129, %130) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x56x56xf32>, tensor<1x256x56x56xf32>) -> tensor<1x256x56x56xf32> loc(#loc512) + %132 = tensor.empty() : tensor<1x56x256x56xf32> loc(#loc513) + %133 = "ttir.transpose"(%131, %132) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x56x56xf32>, tensor<1x56x256x56xf32>) -> tensor<1x56x256x56xf32> loc(#loc513) + %134 = tensor.empty() : tensor<1x56x56x256xf32> loc(#loc514) + %135 = "ttir.transpose"(%133, %134) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x256x56xf32>, tensor<1x56x56x256xf32>) -> tensor<1x56x56x256xf32> loc(#loc514) + %136 = tensor.empty() : tensor<1x56x56x64xf32> loc(#loc515) + %137 = "ttir.conv2d"(%135, %arg115, %136) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x56x56x256xf32>, tensor<64x256x1x1xf32>, tensor<1x56x56x64xf32>) -> tensor<1x56x56x64xf32> loc(#loc515) + %138 = tensor.empty() : tensor<1x56x64x56xf32> loc(#loc516) + %139 = "ttir.transpose"(%137, %138) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x56x64xf32>, tensor<1x56x64x56xf32>) -> tensor<1x56x64x56xf32> loc(#loc516) + %140 = tensor.empty() : tensor<1x64x56x56xf32> loc(#loc517) + %141 = "ttir.transpose"(%139, %140) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x64x56xf32>, tensor<1x64x56x56xf32>) -> tensor<1x64x56x56xf32> loc(#loc517) + %142 = tensor.empty() : tensor<1x64x56x56xf32> loc(#loc518) + %143 = "ttir.multiply"(%141, %arg17, %142) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<64x1x1xf32>, tensor<1x64x56x56xf32>) -> tensor<1x64x56x56xf32> loc(#loc518) + %144 = tensor.empty() : tensor<1x64x56x56xf32> loc(#loc519) + %145 = "ttir.add"(%143, %arg18, %144) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<64x1x1xf32>, tensor<1x64x56x56xf32>) -> tensor<1x64x56x56xf32> loc(#loc519) + %146 = tensor.empty() : tensor<1x64x56x56xf32> loc(#loc520) + %147 = "ttir.relu"(%145, %146) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<1x64x56x56xf32>) -> tensor<1x64x56x56xf32> loc(#loc520) + %148 = tensor.empty() : tensor<1x56x64x56xf32> loc(#loc521) + %149 = "ttir.transpose"(%147, %148) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<1x56x64x56xf32>) -> tensor<1x56x64x56xf32> loc(#loc521) + %150 = tensor.empty() : tensor<1x56x56x64xf32> loc(#loc522) + %151 = "ttir.transpose"(%149, %150) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x64x56xf32>, tensor<1x56x56x64xf32>) -> tensor<1x56x56x64xf32> loc(#loc522) + %152 = tensor.empty() : tensor<1x56x56x64xf32> loc(#loc523) + %153 = "ttir.conv2d"(%151, %arg116, %152) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 1 : si32, padding_left = 1 : si32, padding_right = 1 : si32, padding_top = 1 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x56x56x64xf32>, tensor<64x64x3x3xf32>, tensor<1x56x56x64xf32>) -> tensor<1x56x56x64xf32> loc(#loc523) + %154 = tensor.empty() : tensor<1x56x64x56xf32> loc(#loc524) + %155 = "ttir.transpose"(%153, %154) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x56x64xf32>, tensor<1x56x64x56xf32>) -> tensor<1x56x64x56xf32> loc(#loc524) + %156 = tensor.empty() : tensor<1x64x56x56xf32> loc(#loc525) + %157 = "ttir.transpose"(%155, %156) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x64x56xf32>, tensor<1x64x56x56xf32>) -> tensor<1x64x56x56xf32> loc(#loc525) + %158 = tensor.empty() : tensor<1x64x56x56xf32> loc(#loc526) + %159 = "ttir.multiply"(%157, %arg19, %158) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<64x1x1xf32>, tensor<1x64x56x56xf32>) -> tensor<1x64x56x56xf32> loc(#loc526) + %160 = tensor.empty() : tensor<1x64x56x56xf32> loc(#loc527) + %161 = "ttir.add"(%159, %arg20, %160) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<64x1x1xf32>, tensor<1x64x56x56xf32>) -> tensor<1x64x56x56xf32> loc(#loc527) + %162 = tensor.empty() : tensor<1x64x56x56xf32> loc(#loc528) + %163 = "ttir.relu"(%161, %162) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<1x64x56x56xf32>) -> tensor<1x64x56x56xf32> loc(#loc528) + %164 = tensor.empty() : tensor<1x56x64x56xf32> loc(#loc529) + %165 = "ttir.transpose"(%163, %164) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x64x56x56xf32>, tensor<1x56x64x56xf32>) -> tensor<1x56x64x56xf32> loc(#loc529) + %166 = tensor.empty() : tensor<1x56x56x64xf32> loc(#loc530) + %167 = "ttir.transpose"(%165, %166) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x64x56xf32>, tensor<1x56x56x64xf32>) -> tensor<1x56x56x64xf32> loc(#loc530) + %168 = tensor.empty() : tensor<1x56x56x256xf32> loc(#loc531) + %169 = "ttir.conv2d"(%167, %arg117, %168) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x56x56x64xf32>, tensor<256x64x1x1xf32>, tensor<1x56x56x256xf32>) -> tensor<1x56x56x256xf32> loc(#loc531) + %170 = tensor.empty() : tensor<1x56x256x56xf32> loc(#loc532) + %171 = "ttir.transpose"(%169, %170) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x56x256xf32>, tensor<1x56x256x56xf32>) -> tensor<1x56x256x56xf32> loc(#loc532) + %172 = tensor.empty() : tensor<1x256x56x56xf32> loc(#loc533) + %173 = "ttir.transpose"(%171, %172) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x256x56xf32>, tensor<1x256x56x56xf32>) -> tensor<1x256x56x56xf32> loc(#loc533) + %174 = tensor.empty() : tensor<1x256x56x56xf32> loc(#loc534) + %175 = "ttir.multiply"(%173, %arg21, %174) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x56x56xf32>, tensor<256x1x1xf32>, tensor<1x256x56x56xf32>) -> tensor<1x256x56x56xf32> loc(#loc534) + %176 = tensor.empty() : tensor<1x256x56x56xf32> loc(#loc535) + %177 = "ttir.add"(%175, %arg22, %176) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x56x56xf32>, tensor<256x1x1xf32>, tensor<1x256x56x56xf32>) -> tensor<1x256x56x56xf32> loc(#loc535) + %178 = tensor.empty() : tensor<1x256x56x56xf32> loc(#loc536) + %179 = "ttir.add"(%177, %131, %178) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x56x56xf32>, tensor<1x256x56x56xf32>, tensor<1x256x56x56xf32>) -> tensor<1x256x56x56xf32> loc(#loc536) + %180 = tensor.empty() : tensor<1x256x56x56xf32> loc(#loc537) + %181 = "ttir.relu"(%179, %180) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x56x56xf32>, tensor<1x256x56x56xf32>) -> tensor<1x256x56x56xf32> loc(#loc537) + %182 = tensor.empty() : tensor<1x56x256x56xf32> loc(#loc538) + %183 = "ttir.transpose"(%181, %182) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x56x56xf32>, tensor<1x56x256x56xf32>) -> tensor<1x56x256x56xf32> loc(#loc538) + %184 = tensor.empty() : tensor<1x56x56x256xf32> loc(#loc539) + %185 = "ttir.transpose"(%183, %184) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x256x56xf32>, tensor<1x56x56x256xf32>) -> tensor<1x56x56x256xf32> loc(#loc539) + %186 = tensor.empty() : tensor<1x56x56x128xf32> loc(#loc540) + %187 = "ttir.conv2d"(%185, %arg118, %186) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x56x56x256xf32>, tensor<128x256x1x1xf32>, tensor<1x56x56x128xf32>) -> tensor<1x56x56x128xf32> loc(#loc540) + %188 = tensor.empty() : tensor<1x56x128x56xf32> loc(#loc541) + %189 = "ttir.transpose"(%187, %188) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x56x128xf32>, tensor<1x56x128x56xf32>) -> tensor<1x56x128x56xf32> loc(#loc541) + %190 = tensor.empty() : tensor<1x128x56x56xf32> loc(#loc542) + %191 = "ttir.transpose"(%189, %190) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x128x56xf32>, tensor<1x128x56x56xf32>) -> tensor<1x128x56x56xf32> loc(#loc542) + %192 = tensor.empty() : tensor<1x128x56x56xf32> loc(#loc543) + %193 = "ttir.multiply"(%191, %arg23, %192) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x128x56x56xf32>, tensor<128x1x1xf32>, tensor<1x128x56x56xf32>) -> tensor<1x128x56x56xf32> loc(#loc543) + %194 = tensor.empty() : tensor<1x128x56x56xf32> loc(#loc544) + %195 = "ttir.add"(%193, %arg24, %194) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x128x56x56xf32>, tensor<128x1x1xf32>, tensor<1x128x56x56xf32>) -> tensor<1x128x56x56xf32> loc(#loc544) + %196 = tensor.empty() : tensor<1x128x56x56xf32> loc(#loc545) + %197 = "ttir.relu"(%195, %196) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x128x56x56xf32>, tensor<1x128x56x56xf32>) -> tensor<1x128x56x56xf32> loc(#loc545) + %198 = tensor.empty() : tensor<1x56x128x56xf32> loc(#loc546) + %199 = "ttir.transpose"(%197, %198) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x128x56x56xf32>, tensor<1x56x128x56xf32>) -> tensor<1x56x128x56xf32> loc(#loc546) + %200 = tensor.empty() : tensor<1x56x56x128xf32> loc(#loc547) + %201 = "ttir.transpose"(%199, %200) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x128x56xf32>, tensor<1x56x56x128xf32>) -> tensor<1x56x56x128xf32> loc(#loc547) + %202 = tensor.empty() : tensor<1x28x28x128xf32> loc(#loc548) + %203 = "ttir.conv2d"(%201, %arg119, %202) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 1 : si32, padding_left = 1 : si32, padding_right = 1 : si32, padding_top = 1 : si32, stride_height = 2 : si32, stride_width = 2 : si32}> {channel_last = 1 : si32} : (tensor<1x56x56x128xf32>, tensor<128x128x3x3xf32>, tensor<1x28x28x128xf32>) -> tensor<1x28x28x128xf32> loc(#loc548) + %204 = tensor.empty() : tensor<1x28x128x28xf32> loc(#loc549) + %205 = "ttir.transpose"(%203, %204) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x28x128xf32>, tensor<1x28x128x28xf32>) -> tensor<1x28x128x28xf32> loc(#loc549) + %206 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc550) + %207 = "ttir.transpose"(%205, %206) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x128x28xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc550) + %208 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc551) + %209 = "ttir.multiply"(%207, %arg25, %208) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<128x1x1xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc551) + %210 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc552) + %211 = "ttir.add"(%209, %arg26, %210) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<128x1x1xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc552) + %212 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc553) + %213 = "ttir.relu"(%211, %212) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc553) + %214 = tensor.empty() : tensor<1x28x128x28xf32> loc(#loc554) + %215 = "ttir.transpose"(%213, %214) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<1x28x128x28xf32>) -> tensor<1x28x128x28xf32> loc(#loc554) + %216 = tensor.empty() : tensor<1x28x28x128xf32> loc(#loc555) + %217 = "ttir.transpose"(%215, %216) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x128x28xf32>, tensor<1x28x28x128xf32>) -> tensor<1x28x28x128xf32> loc(#loc555) + %218 = tensor.empty() : tensor<1x28x28x512xf32> loc(#loc556) + %219 = "ttir.conv2d"(%217, %arg120, %218) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x28x28x128xf32>, tensor<512x128x1x1xf32>, tensor<1x28x28x512xf32>) -> tensor<1x28x28x512xf32> loc(#loc556) + %220 = tensor.empty() : tensor<1x28x512x28xf32> loc(#loc557) + %221 = "ttir.transpose"(%219, %220) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x28x512xf32>, tensor<1x28x512x28xf32>) -> tensor<1x28x512x28xf32> loc(#loc557) + %222 = tensor.empty() : tensor<1x512x28x28xf32> loc(#loc558) + %223 = "ttir.transpose"(%221, %222) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x512x28xf32>, tensor<1x512x28x28xf32>) -> tensor<1x512x28x28xf32> loc(#loc558) + %224 = tensor.empty() : tensor<1x512x28x28xf32> loc(#loc559) + %225 = "ttir.multiply"(%223, %arg27, %224) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x28x28xf32>, tensor<512x1x1xf32>, tensor<1x512x28x28xf32>) -> tensor<1x512x28x28xf32> loc(#loc559) + %226 = tensor.empty() : tensor<1x512x28x28xf32> loc(#loc560) + %227 = "ttir.add"(%225, %arg28, %226) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x28x28xf32>, tensor<512x1x1xf32>, tensor<1x512x28x28xf32>) -> tensor<1x512x28x28xf32> loc(#loc560) + %228 = tensor.empty() : tensor<1x56x256x56xf32> loc(#loc561) + %229 = "ttir.transpose"(%181, %228) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x56x56xf32>, tensor<1x56x256x56xf32>) -> tensor<1x56x256x56xf32> loc(#loc561) + %230 = tensor.empty() : tensor<1x56x56x256xf32> loc(#loc562) + %231 = "ttir.transpose"(%229, %230) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x56x256x56xf32>, tensor<1x56x56x256xf32>) -> tensor<1x56x56x256xf32> loc(#loc562) + %232 = tensor.empty() : tensor<1x28x28x512xf32> loc(#loc563) + %233 = "ttir.conv2d"(%231, %arg121, %232) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 2 : si32, stride_width = 2 : si32}> {channel_last = 1 : si32} : (tensor<1x56x56x256xf32>, tensor<512x256x1x1xf32>, tensor<1x28x28x512xf32>) -> tensor<1x28x28x512xf32> loc(#loc563) + %234 = tensor.empty() : tensor<1x28x512x28xf32> loc(#loc564) + %235 = "ttir.transpose"(%233, %234) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x28x512xf32>, tensor<1x28x512x28xf32>) -> tensor<1x28x512x28xf32> loc(#loc564) + %236 = tensor.empty() : tensor<1x512x28x28xf32> loc(#loc565) + %237 = "ttir.transpose"(%235, %236) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x512x28xf32>, tensor<1x512x28x28xf32>) -> tensor<1x512x28x28xf32> loc(#loc565) + %238 = tensor.empty() : tensor<1x512x28x28xf32> loc(#loc566) + %239 = "ttir.multiply"(%237, %arg29, %238) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x28x28xf32>, tensor<512x1x1xf32>, tensor<1x512x28x28xf32>) -> tensor<1x512x28x28xf32> loc(#loc566) + %240 = tensor.empty() : tensor<1x512x28x28xf32> loc(#loc567) + %241 = "ttir.add"(%239, %arg30, %240) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x28x28xf32>, tensor<512x1x1xf32>, tensor<1x512x28x28xf32>) -> tensor<1x512x28x28xf32> loc(#loc567) + %242 = tensor.empty() : tensor<1x512x28x28xf32> loc(#loc568) + %243 = "ttir.add"(%227, %241, %242) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x28x28xf32>, tensor<1x512x28x28xf32>, tensor<1x512x28x28xf32>) -> tensor<1x512x28x28xf32> loc(#loc568) + %244 = tensor.empty() : tensor<1x512x28x28xf32> loc(#loc569) + %245 = "ttir.relu"(%243, %244) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x28x28xf32>, tensor<1x512x28x28xf32>) -> tensor<1x512x28x28xf32> loc(#loc569) + %246 = tensor.empty() : tensor<1x28x512x28xf32> loc(#loc570) + %247 = "ttir.transpose"(%245, %246) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x512x28x28xf32>, tensor<1x28x512x28xf32>) -> tensor<1x28x512x28xf32> loc(#loc570) + %248 = tensor.empty() : tensor<1x28x28x512xf32> loc(#loc571) + %249 = "ttir.transpose"(%247, %248) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x512x28xf32>, tensor<1x28x28x512xf32>) -> tensor<1x28x28x512xf32> loc(#loc571) + %250 = tensor.empty() : tensor<1x28x28x128xf32> loc(#loc572) + %251 = "ttir.conv2d"(%249, %arg122, %250) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x28x28x512xf32>, tensor<128x512x1x1xf32>, tensor<1x28x28x128xf32>) -> tensor<1x28x28x128xf32> loc(#loc572) + %252 = tensor.empty() : tensor<1x28x128x28xf32> loc(#loc573) + %253 = "ttir.transpose"(%251, %252) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x28x128xf32>, tensor<1x28x128x28xf32>) -> tensor<1x28x128x28xf32> loc(#loc573) + %254 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc574) + %255 = "ttir.transpose"(%253, %254) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x128x28xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc574) + %256 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc575) + %257 = "ttir.multiply"(%255, %arg31, %256) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<128x1x1xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc575) + %258 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc576) + %259 = "ttir.add"(%257, %arg32, %258) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<128x1x1xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc576) + %260 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc577) + %261 = "ttir.relu"(%259, %260) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc577) + %262 = tensor.empty() : tensor<1x28x128x28xf32> loc(#loc578) + %263 = "ttir.transpose"(%261, %262) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<1x28x128x28xf32>) -> tensor<1x28x128x28xf32> loc(#loc578) + %264 = tensor.empty() : tensor<1x28x28x128xf32> loc(#loc579) + %265 = "ttir.transpose"(%263, %264) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x128x28xf32>, tensor<1x28x28x128xf32>) -> tensor<1x28x28x128xf32> loc(#loc579) + %266 = tensor.empty() : tensor<1x28x28x128xf32> loc(#loc580) + %267 = "ttir.conv2d"(%265, %arg123, %266) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 1 : si32, padding_left = 1 : si32, padding_right = 1 : si32, padding_top = 1 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x28x28x128xf32>, tensor<128x128x3x3xf32>, tensor<1x28x28x128xf32>) -> tensor<1x28x28x128xf32> loc(#loc580) + %268 = tensor.empty() : tensor<1x28x128x28xf32> loc(#loc581) + %269 = "ttir.transpose"(%267, %268) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x28x128xf32>, tensor<1x28x128x28xf32>) -> tensor<1x28x128x28xf32> loc(#loc581) + %270 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc582) + %271 = "ttir.transpose"(%269, %270) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x128x28xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc582) + %272 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc583) + %273 = "ttir.multiply"(%271, %arg33, %272) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<128x1x1xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc583) + %274 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc584) + %275 = "ttir.add"(%273, %arg34, %274) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<128x1x1xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc584) + %276 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc585) + %277 = "ttir.relu"(%275, %276) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc585) + %278 = tensor.empty() : tensor<1x28x128x28xf32> loc(#loc586) + %279 = "ttir.transpose"(%277, %278) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<1x28x128x28xf32>) -> tensor<1x28x128x28xf32> loc(#loc586) + %280 = tensor.empty() : tensor<1x28x28x128xf32> loc(#loc587) + %281 = "ttir.transpose"(%279, %280) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x128x28xf32>, tensor<1x28x28x128xf32>) -> tensor<1x28x28x128xf32> loc(#loc587) + %282 = tensor.empty() : tensor<1x28x28x512xf32> loc(#loc588) + %283 = "ttir.conv2d"(%281, %arg124, %282) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x28x28x128xf32>, tensor<512x128x1x1xf32>, tensor<1x28x28x512xf32>) -> tensor<1x28x28x512xf32> loc(#loc588) + %284 = tensor.empty() : tensor<1x28x512x28xf32> loc(#loc589) + %285 = "ttir.transpose"(%283, %284) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x28x512xf32>, tensor<1x28x512x28xf32>) -> tensor<1x28x512x28xf32> loc(#loc589) + %286 = tensor.empty() : tensor<1x512x28x28xf32> loc(#loc590) + %287 = "ttir.transpose"(%285, %286) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x512x28xf32>, tensor<1x512x28x28xf32>) -> tensor<1x512x28x28xf32> loc(#loc590) + %288 = tensor.empty() : tensor<1x512x28x28xf32> loc(#loc591) + %289 = "ttir.multiply"(%287, %arg35, %288) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x28x28xf32>, tensor<512x1x1xf32>, tensor<1x512x28x28xf32>) -> tensor<1x512x28x28xf32> loc(#loc591) + %290 = tensor.empty() : tensor<1x512x28x28xf32> loc(#loc592) + %291 = "ttir.add"(%289, %arg36, %290) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x28x28xf32>, tensor<512x1x1xf32>, tensor<1x512x28x28xf32>) -> tensor<1x512x28x28xf32> loc(#loc592) + %292 = tensor.empty() : tensor<1x512x28x28xf32> loc(#loc593) + %293 = "ttir.add"(%291, %245, %292) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x28x28xf32>, tensor<1x512x28x28xf32>, tensor<1x512x28x28xf32>) -> tensor<1x512x28x28xf32> loc(#loc593) + %294 = tensor.empty() : tensor<1x512x28x28xf32> loc(#loc594) + %295 = "ttir.relu"(%293, %294) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x28x28xf32>, tensor<1x512x28x28xf32>) -> tensor<1x512x28x28xf32> loc(#loc594) + %296 = tensor.empty() : tensor<1x28x512x28xf32> loc(#loc595) + %297 = "ttir.transpose"(%295, %296) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x512x28x28xf32>, tensor<1x28x512x28xf32>) -> tensor<1x28x512x28xf32> loc(#loc595) + %298 = tensor.empty() : tensor<1x28x28x512xf32> loc(#loc596) + %299 = "ttir.transpose"(%297, %298) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x512x28xf32>, tensor<1x28x28x512xf32>) -> tensor<1x28x28x512xf32> loc(#loc596) + %300 = tensor.empty() : tensor<1x28x28x128xf32> loc(#loc597) + %301 = "ttir.conv2d"(%299, %arg125, %300) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x28x28x512xf32>, tensor<128x512x1x1xf32>, tensor<1x28x28x128xf32>) -> tensor<1x28x28x128xf32> loc(#loc597) + %302 = tensor.empty() : tensor<1x28x128x28xf32> loc(#loc598) + %303 = "ttir.transpose"(%301, %302) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x28x128xf32>, tensor<1x28x128x28xf32>) -> tensor<1x28x128x28xf32> loc(#loc598) + %304 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc599) + %305 = "ttir.transpose"(%303, %304) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x128x28xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc599) + %306 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc600) + %307 = "ttir.multiply"(%305, %arg37, %306) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<128x1x1xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc600) + %308 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc601) + %309 = "ttir.add"(%307, %arg38, %308) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<128x1x1xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc601) + %310 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc602) + %311 = "ttir.relu"(%309, %310) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc602) + %312 = tensor.empty() : tensor<1x28x128x28xf32> loc(#loc603) + %313 = "ttir.transpose"(%311, %312) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<1x28x128x28xf32>) -> tensor<1x28x128x28xf32> loc(#loc603) + %314 = tensor.empty() : tensor<1x28x28x128xf32> loc(#loc604) + %315 = "ttir.transpose"(%313, %314) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x128x28xf32>, tensor<1x28x28x128xf32>) -> tensor<1x28x28x128xf32> loc(#loc604) + %316 = tensor.empty() : tensor<1x28x28x128xf32> loc(#loc605) + %317 = "ttir.conv2d"(%315, %arg126, %316) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 1 : si32, padding_left = 1 : si32, padding_right = 1 : si32, padding_top = 1 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x28x28x128xf32>, tensor<128x128x3x3xf32>, tensor<1x28x28x128xf32>) -> tensor<1x28x28x128xf32> loc(#loc605) + %318 = tensor.empty() : tensor<1x28x128x28xf32> loc(#loc606) + %319 = "ttir.transpose"(%317, %318) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x28x128xf32>, tensor<1x28x128x28xf32>) -> tensor<1x28x128x28xf32> loc(#loc606) + %320 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc607) + %321 = "ttir.transpose"(%319, %320) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x128x28xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc607) + %322 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc608) + %323 = "ttir.multiply"(%321, %arg39, %322) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<128x1x1xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc608) + %324 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc609) + %325 = "ttir.add"(%323, %arg40, %324) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<128x1x1xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc609) + %326 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc610) + %327 = "ttir.relu"(%325, %326) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc610) + %328 = tensor.empty() : tensor<1x28x128x28xf32> loc(#loc611) + %329 = "ttir.transpose"(%327, %328) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<1x28x128x28xf32>) -> tensor<1x28x128x28xf32> loc(#loc611) + %330 = tensor.empty() : tensor<1x28x28x128xf32> loc(#loc612) + %331 = "ttir.transpose"(%329, %330) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x128x28xf32>, tensor<1x28x28x128xf32>) -> tensor<1x28x28x128xf32> loc(#loc612) + %332 = tensor.empty() : tensor<1x28x28x512xf32> loc(#loc613) + %333 = "ttir.conv2d"(%331, %arg127, %332) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x28x28x128xf32>, tensor<512x128x1x1xf32>, tensor<1x28x28x512xf32>) -> tensor<1x28x28x512xf32> loc(#loc613) + %334 = tensor.empty() : tensor<1x28x512x28xf32> loc(#loc614) + %335 = "ttir.transpose"(%333, %334) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x28x512xf32>, tensor<1x28x512x28xf32>) -> tensor<1x28x512x28xf32> loc(#loc614) + %336 = tensor.empty() : tensor<1x512x28x28xf32> loc(#loc615) + %337 = "ttir.transpose"(%335, %336) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x512x28xf32>, tensor<1x512x28x28xf32>) -> tensor<1x512x28x28xf32> loc(#loc615) + %338 = tensor.empty() : tensor<1x512x28x28xf32> loc(#loc616) + %339 = "ttir.multiply"(%337, %arg41, %338) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x28x28xf32>, tensor<512x1x1xf32>, tensor<1x512x28x28xf32>) -> tensor<1x512x28x28xf32> loc(#loc616) + %340 = tensor.empty() : tensor<1x512x28x28xf32> loc(#loc617) + %341 = "ttir.add"(%339, %arg42, %340) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x28x28xf32>, tensor<512x1x1xf32>, tensor<1x512x28x28xf32>) -> tensor<1x512x28x28xf32> loc(#loc617) + %342 = tensor.empty() : tensor<1x512x28x28xf32> loc(#loc618) + %343 = "ttir.add"(%341, %295, %342) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x28x28xf32>, tensor<1x512x28x28xf32>, tensor<1x512x28x28xf32>) -> tensor<1x512x28x28xf32> loc(#loc618) + %344 = tensor.empty() : tensor<1x512x28x28xf32> loc(#loc619) + %345 = "ttir.relu"(%343, %344) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x28x28xf32>, tensor<1x512x28x28xf32>) -> tensor<1x512x28x28xf32> loc(#loc619) + %346 = tensor.empty() : tensor<1x28x512x28xf32> loc(#loc620) + %347 = "ttir.transpose"(%345, %346) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x512x28x28xf32>, tensor<1x28x512x28xf32>) -> tensor<1x28x512x28xf32> loc(#loc620) + %348 = tensor.empty() : tensor<1x28x28x512xf32> loc(#loc621) + %349 = "ttir.transpose"(%347, %348) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x512x28xf32>, tensor<1x28x28x512xf32>) -> tensor<1x28x28x512xf32> loc(#loc621) + %350 = tensor.empty() : tensor<1x28x28x128xf32> loc(#loc622) + %351 = "ttir.conv2d"(%349, %arg128, %350) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x28x28x512xf32>, tensor<128x512x1x1xf32>, tensor<1x28x28x128xf32>) -> tensor<1x28x28x128xf32> loc(#loc622) + %352 = tensor.empty() : tensor<1x28x128x28xf32> loc(#loc623) + %353 = "ttir.transpose"(%351, %352) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x28x128xf32>, tensor<1x28x128x28xf32>) -> tensor<1x28x128x28xf32> loc(#loc623) + %354 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc624) + %355 = "ttir.transpose"(%353, %354) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x128x28xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc624) + %356 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc625) + %357 = "ttir.multiply"(%355, %arg43, %356) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<128x1x1xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc625) + %358 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc626) + %359 = "ttir.add"(%357, %arg44, %358) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<128x1x1xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc626) + %360 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc627) + %361 = "ttir.relu"(%359, %360) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc627) + %362 = tensor.empty() : tensor<1x28x128x28xf32> loc(#loc628) + %363 = "ttir.transpose"(%361, %362) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<1x28x128x28xf32>) -> tensor<1x28x128x28xf32> loc(#loc628) + %364 = tensor.empty() : tensor<1x28x28x128xf32> loc(#loc629) + %365 = "ttir.transpose"(%363, %364) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x128x28xf32>, tensor<1x28x28x128xf32>) -> tensor<1x28x28x128xf32> loc(#loc629) + %366 = tensor.empty() : tensor<1x28x28x128xf32> loc(#loc630) + %367 = "ttir.conv2d"(%365, %arg129, %366) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 1 : si32, padding_left = 1 : si32, padding_right = 1 : si32, padding_top = 1 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x28x28x128xf32>, tensor<128x128x3x3xf32>, tensor<1x28x28x128xf32>) -> tensor<1x28x28x128xf32> loc(#loc630) + %368 = tensor.empty() : tensor<1x28x128x28xf32> loc(#loc631) + %369 = "ttir.transpose"(%367, %368) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x28x128xf32>, tensor<1x28x128x28xf32>) -> tensor<1x28x128x28xf32> loc(#loc631) + %370 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc632) + %371 = "ttir.transpose"(%369, %370) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x128x28xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc632) + %372 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc633) + %373 = "ttir.multiply"(%371, %arg45, %372) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<128x1x1xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc633) + %374 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc634) + %375 = "ttir.add"(%373, %arg46, %374) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<128x1x1xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc634) + %376 = tensor.empty() : tensor<1x128x28x28xf32> loc(#loc635) + %377 = "ttir.relu"(%375, %376) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<1x128x28x28xf32>) -> tensor<1x128x28x28xf32> loc(#loc635) + %378 = tensor.empty() : tensor<1x28x128x28xf32> loc(#loc636) + %379 = "ttir.transpose"(%377, %378) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x128x28x28xf32>, tensor<1x28x128x28xf32>) -> tensor<1x28x128x28xf32> loc(#loc636) + %380 = tensor.empty() : tensor<1x28x28x128xf32> loc(#loc637) + %381 = "ttir.transpose"(%379, %380) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x128x28xf32>, tensor<1x28x28x128xf32>) -> tensor<1x28x28x128xf32> loc(#loc637) + %382 = tensor.empty() : tensor<1x28x28x512xf32> loc(#loc638) + %383 = "ttir.conv2d"(%381, %arg130, %382) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x28x28x128xf32>, tensor<512x128x1x1xf32>, tensor<1x28x28x512xf32>) -> tensor<1x28x28x512xf32> loc(#loc638) + %384 = tensor.empty() : tensor<1x28x512x28xf32> loc(#loc639) + %385 = "ttir.transpose"(%383, %384) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x28x512xf32>, tensor<1x28x512x28xf32>) -> tensor<1x28x512x28xf32> loc(#loc639) + %386 = tensor.empty() : tensor<1x512x28x28xf32> loc(#loc640) + %387 = "ttir.transpose"(%385, %386) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x512x28xf32>, tensor<1x512x28x28xf32>) -> tensor<1x512x28x28xf32> loc(#loc640) + %388 = tensor.empty() : tensor<1x512x28x28xf32> loc(#loc641) + %389 = "ttir.multiply"(%387, %arg47, %388) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x28x28xf32>, tensor<512x1x1xf32>, tensor<1x512x28x28xf32>) -> tensor<1x512x28x28xf32> loc(#loc641) + %390 = tensor.empty() : tensor<1x512x28x28xf32> loc(#loc642) + %391 = "ttir.add"(%389, %arg48, %390) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x28x28xf32>, tensor<512x1x1xf32>, tensor<1x512x28x28xf32>) -> tensor<1x512x28x28xf32> loc(#loc642) + %392 = tensor.empty() : tensor<1x512x28x28xf32> loc(#loc643) + %393 = "ttir.add"(%391, %345, %392) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x28x28xf32>, tensor<1x512x28x28xf32>, tensor<1x512x28x28xf32>) -> tensor<1x512x28x28xf32> loc(#loc643) + %394 = tensor.empty() : tensor<1x512x28x28xf32> loc(#loc644) + %395 = "ttir.relu"(%393, %394) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x28x28xf32>, tensor<1x512x28x28xf32>) -> tensor<1x512x28x28xf32> loc(#loc644) + %396 = tensor.empty() : tensor<1x28x512x28xf32> loc(#loc645) + %397 = "ttir.transpose"(%395, %396) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x512x28x28xf32>, tensor<1x28x512x28xf32>) -> tensor<1x28x512x28xf32> loc(#loc645) + %398 = tensor.empty() : tensor<1x28x28x512xf32> loc(#loc646) + %399 = "ttir.transpose"(%397, %398) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x512x28xf32>, tensor<1x28x28x512xf32>) -> tensor<1x28x28x512xf32> loc(#loc646) + %400 = tensor.empty() : tensor<1x28x28x256xf32> loc(#loc647) + %401 = "ttir.conv2d"(%399, %arg131, %400) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x28x28x512xf32>, tensor<256x512x1x1xf32>, tensor<1x28x28x256xf32>) -> tensor<1x28x28x256xf32> loc(#loc647) + %402 = tensor.empty() : tensor<1x28x256x28xf32> loc(#loc648) + %403 = "ttir.transpose"(%401, %402) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x28x256xf32>, tensor<1x28x256x28xf32>) -> tensor<1x28x256x28xf32> loc(#loc648) + %404 = tensor.empty() : tensor<1x256x28x28xf32> loc(#loc649) + %405 = "ttir.transpose"(%403, %404) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x256x28xf32>, tensor<1x256x28x28xf32>) -> tensor<1x256x28x28xf32> loc(#loc649) + %406 = tensor.empty() : tensor<1x256x28x28xf32> loc(#loc650) + %407 = "ttir.multiply"(%405, %arg49, %406) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x28x28xf32>, tensor<256x1x1xf32>, tensor<1x256x28x28xf32>) -> tensor<1x256x28x28xf32> loc(#loc650) + %408 = tensor.empty() : tensor<1x256x28x28xf32> loc(#loc651) + %409 = "ttir.add"(%407, %arg50, %408) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x28x28xf32>, tensor<256x1x1xf32>, tensor<1x256x28x28xf32>) -> tensor<1x256x28x28xf32> loc(#loc651) + %410 = tensor.empty() : tensor<1x256x28x28xf32> loc(#loc652) + %411 = "ttir.relu"(%409, %410) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x28x28xf32>, tensor<1x256x28x28xf32>) -> tensor<1x256x28x28xf32> loc(#loc652) + %412 = tensor.empty() : tensor<1x28x256x28xf32> loc(#loc653) + %413 = "ttir.transpose"(%411, %412) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x28x28xf32>, tensor<1x28x256x28xf32>) -> tensor<1x28x256x28xf32> loc(#loc653) + %414 = tensor.empty() : tensor<1x28x28x256xf32> loc(#loc654) + %415 = "ttir.transpose"(%413, %414) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x256x28xf32>, tensor<1x28x28x256xf32>) -> tensor<1x28x28x256xf32> loc(#loc654) + %416 = tensor.empty() : tensor<1x14x14x256xf32> loc(#loc655) + %417 = "ttir.conv2d"(%415, %arg132, %416) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 1 : si32, padding_left = 1 : si32, padding_right = 1 : si32, padding_top = 1 : si32, stride_height = 2 : si32, stride_width = 2 : si32}> {channel_last = 1 : si32} : (tensor<1x28x28x256xf32>, tensor<256x256x3x3xf32>, tensor<1x14x14x256xf32>) -> tensor<1x14x14x256xf32> loc(#loc655) + %418 = tensor.empty() : tensor<1x14x256x14xf32> loc(#loc656) + %419 = "ttir.transpose"(%417, %418) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x14x256xf32>, tensor<1x14x256x14xf32>) -> tensor<1x14x256x14xf32> loc(#loc656) + %420 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc657) + %421 = "ttir.transpose"(%419, %420) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x256x14xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc657) + %422 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc658) + %423 = "ttir.multiply"(%421, %arg51, %422) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<256x1x1xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc658) + %424 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc659) + %425 = "ttir.add"(%423, %arg52, %424) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<256x1x1xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc659) + %426 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc660) + %427 = "ttir.relu"(%425, %426) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc660) + %428 = tensor.empty() : tensor<1x14x256x14xf32> loc(#loc661) + %429 = "ttir.transpose"(%427, %428) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<1x14x256x14xf32>) -> tensor<1x14x256x14xf32> loc(#loc661) + %430 = tensor.empty() : tensor<1x14x14x256xf32> loc(#loc662) + %431 = "ttir.transpose"(%429, %430) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x256x14xf32>, tensor<1x14x14x256xf32>) -> tensor<1x14x14x256xf32> loc(#loc662) + %432 = tensor.empty() : tensor<1x14x14x1024xf32> loc(#loc663) + %433 = "ttir.conv2d"(%431, %arg133, %432) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x14x14x256xf32>, tensor<1024x256x1x1xf32>, tensor<1x14x14x1024xf32>) -> tensor<1x14x14x1024xf32> loc(#loc663) + %434 = tensor.empty() : tensor<1x14x1024x14xf32> loc(#loc664) + %435 = "ttir.transpose"(%433, %434) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x14x1024xf32>, tensor<1x14x1024x14xf32>) -> tensor<1x14x1024x14xf32> loc(#loc664) + %436 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc665) + %437 = "ttir.transpose"(%435, %436) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x1024x14xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc665) + %438 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc666) + %439 = "ttir.multiply"(%437, %arg53, %438) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1024x1x1xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc666) + %440 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc667) + %441 = "ttir.add"(%439, %arg54, %440) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1024x1x1xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc667) + %442 = tensor.empty() : tensor<1x28x512x28xf32> loc(#loc668) + %443 = "ttir.transpose"(%395, %442) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x512x28x28xf32>, tensor<1x28x512x28xf32>) -> tensor<1x28x512x28xf32> loc(#loc668) + %444 = tensor.empty() : tensor<1x28x28x512xf32> loc(#loc669) + %445 = "ttir.transpose"(%443, %444) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x28x512x28xf32>, tensor<1x28x28x512xf32>) -> tensor<1x28x28x512xf32> loc(#loc669) + %446 = tensor.empty() : tensor<1x14x14x1024xf32> loc(#loc670) + %447 = "ttir.conv2d"(%445, %arg134, %446) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 2 : si32, stride_width = 2 : si32}> {channel_last = 1 : si32} : (tensor<1x28x28x512xf32>, tensor<1024x512x1x1xf32>, tensor<1x14x14x1024xf32>) -> tensor<1x14x14x1024xf32> loc(#loc670) + %448 = tensor.empty() : tensor<1x14x1024x14xf32> loc(#loc671) + %449 = "ttir.transpose"(%447, %448) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x14x1024xf32>, tensor<1x14x1024x14xf32>) -> tensor<1x14x1024x14xf32> loc(#loc671) + %450 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc672) + %451 = "ttir.transpose"(%449, %450) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x1024x14xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc672) + %452 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc673) + %453 = "ttir.multiply"(%451, %arg55, %452) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1024x1x1xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc673) + %454 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc674) + %455 = "ttir.add"(%453, %arg56, %454) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1024x1x1xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc674) + %456 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc675) + %457 = "ttir.add"(%441, %455, %456) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1x1024x14x14xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc675) + %458 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc676) + %459 = "ttir.relu"(%457, %458) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc676) + %460 = tensor.empty() : tensor<1x14x1024x14xf32> loc(#loc677) + %461 = "ttir.transpose"(%459, %460) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1x14x1024x14xf32>) -> tensor<1x14x1024x14xf32> loc(#loc677) + %462 = tensor.empty() : tensor<1x14x14x1024xf32> loc(#loc678) + %463 = "ttir.transpose"(%461, %462) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x1024x14xf32>, tensor<1x14x14x1024xf32>) -> tensor<1x14x14x1024xf32> loc(#loc678) + %464 = tensor.empty() : tensor<1x14x14x256xf32> loc(#loc679) + %465 = "ttir.conv2d"(%463, %arg135, %464) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x14x14x1024xf32>, tensor<256x1024x1x1xf32>, tensor<1x14x14x256xf32>) -> tensor<1x14x14x256xf32> loc(#loc679) + %466 = tensor.empty() : tensor<1x14x256x14xf32> loc(#loc680) + %467 = "ttir.transpose"(%465, %466) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x14x256xf32>, tensor<1x14x256x14xf32>) -> tensor<1x14x256x14xf32> loc(#loc680) + %468 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc681) + %469 = "ttir.transpose"(%467, %468) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x256x14xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc681) + %470 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc682) + %471 = "ttir.multiply"(%469, %arg57, %470) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<256x1x1xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc682) + %472 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc683) + %473 = "ttir.add"(%471, %arg58, %472) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<256x1x1xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc683) + %474 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc684) + %475 = "ttir.relu"(%473, %474) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc684) + %476 = tensor.empty() : tensor<1x14x256x14xf32> loc(#loc685) + %477 = "ttir.transpose"(%475, %476) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<1x14x256x14xf32>) -> tensor<1x14x256x14xf32> loc(#loc685) + %478 = tensor.empty() : tensor<1x14x14x256xf32> loc(#loc686) + %479 = "ttir.transpose"(%477, %478) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x256x14xf32>, tensor<1x14x14x256xf32>) -> tensor<1x14x14x256xf32> loc(#loc686) + %480 = tensor.empty() : tensor<1x14x14x256xf32> loc(#loc687) + %481 = "ttir.conv2d"(%479, %arg136, %480) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 1 : si32, padding_left = 1 : si32, padding_right = 1 : si32, padding_top = 1 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x14x14x256xf32>, tensor<256x256x3x3xf32>, tensor<1x14x14x256xf32>) -> tensor<1x14x14x256xf32> loc(#loc687) + %482 = tensor.empty() : tensor<1x14x256x14xf32> loc(#loc688) + %483 = "ttir.transpose"(%481, %482) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x14x256xf32>, tensor<1x14x256x14xf32>) -> tensor<1x14x256x14xf32> loc(#loc688) + %484 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc689) + %485 = "ttir.transpose"(%483, %484) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x256x14xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc689) + %486 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc690) + %487 = "ttir.multiply"(%485, %arg59, %486) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<256x1x1xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc690) + %488 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc691) + %489 = "ttir.add"(%487, %arg60, %488) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<256x1x1xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc691) + %490 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc692) + %491 = "ttir.relu"(%489, %490) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc692) + %492 = tensor.empty() : tensor<1x14x256x14xf32> loc(#loc693) + %493 = "ttir.transpose"(%491, %492) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<1x14x256x14xf32>) -> tensor<1x14x256x14xf32> loc(#loc693) + %494 = tensor.empty() : tensor<1x14x14x256xf32> loc(#loc694) + %495 = "ttir.transpose"(%493, %494) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x256x14xf32>, tensor<1x14x14x256xf32>) -> tensor<1x14x14x256xf32> loc(#loc694) + %496 = tensor.empty() : tensor<1x14x14x1024xf32> loc(#loc695) + %497 = "ttir.conv2d"(%495, %arg137, %496) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x14x14x256xf32>, tensor<1024x256x1x1xf32>, tensor<1x14x14x1024xf32>) -> tensor<1x14x14x1024xf32> loc(#loc695) + %498 = tensor.empty() : tensor<1x14x1024x14xf32> loc(#loc696) + %499 = "ttir.transpose"(%497, %498) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x14x1024xf32>, tensor<1x14x1024x14xf32>) -> tensor<1x14x1024x14xf32> loc(#loc696) + %500 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc697) + %501 = "ttir.transpose"(%499, %500) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x1024x14xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc697) + %502 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc698) + %503 = "ttir.multiply"(%501, %arg61, %502) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1024x1x1xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc698) + %504 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc699) + %505 = "ttir.add"(%503, %arg62, %504) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1024x1x1xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc699) + %506 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc700) + %507 = "ttir.add"(%505, %459, %506) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1x1024x14x14xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc700) + %508 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc701) + %509 = "ttir.relu"(%507, %508) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc701) + %510 = tensor.empty() : tensor<1x14x1024x14xf32> loc(#loc702) + %511 = "ttir.transpose"(%509, %510) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1x14x1024x14xf32>) -> tensor<1x14x1024x14xf32> loc(#loc702) + %512 = tensor.empty() : tensor<1x14x14x1024xf32> loc(#loc703) + %513 = "ttir.transpose"(%511, %512) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x1024x14xf32>, tensor<1x14x14x1024xf32>) -> tensor<1x14x14x1024xf32> loc(#loc703) + %514 = tensor.empty() : tensor<1x14x14x256xf32> loc(#loc704) + %515 = "ttir.conv2d"(%513, %arg138, %514) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x14x14x1024xf32>, tensor<256x1024x1x1xf32>, tensor<1x14x14x256xf32>) -> tensor<1x14x14x256xf32> loc(#loc704) + %516 = tensor.empty() : tensor<1x14x256x14xf32> loc(#loc705) + %517 = "ttir.transpose"(%515, %516) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x14x256xf32>, tensor<1x14x256x14xf32>) -> tensor<1x14x256x14xf32> loc(#loc705) + %518 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc706) + %519 = "ttir.transpose"(%517, %518) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x256x14xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc706) + %520 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc707) + %521 = "ttir.multiply"(%519, %arg63, %520) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<256x1x1xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc707) + %522 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc708) + %523 = "ttir.add"(%521, %arg64, %522) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<256x1x1xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc708) + %524 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc709) + %525 = "ttir.relu"(%523, %524) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc709) + %526 = tensor.empty() : tensor<1x14x256x14xf32> loc(#loc710) + %527 = "ttir.transpose"(%525, %526) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<1x14x256x14xf32>) -> tensor<1x14x256x14xf32> loc(#loc710) + %528 = tensor.empty() : tensor<1x14x14x256xf32> loc(#loc711) + %529 = "ttir.transpose"(%527, %528) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x256x14xf32>, tensor<1x14x14x256xf32>) -> tensor<1x14x14x256xf32> loc(#loc711) + %530 = tensor.empty() : tensor<1x14x14x256xf32> loc(#loc712) + %531 = "ttir.conv2d"(%529, %arg139, %530) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 1 : si32, padding_left = 1 : si32, padding_right = 1 : si32, padding_top = 1 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x14x14x256xf32>, tensor<256x256x3x3xf32>, tensor<1x14x14x256xf32>) -> tensor<1x14x14x256xf32> loc(#loc712) + %532 = tensor.empty() : tensor<1x14x256x14xf32> loc(#loc713) + %533 = "ttir.transpose"(%531, %532) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x14x256xf32>, tensor<1x14x256x14xf32>) -> tensor<1x14x256x14xf32> loc(#loc713) + %534 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc714) + %535 = "ttir.transpose"(%533, %534) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x256x14xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc714) + %536 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc715) + %537 = "ttir.multiply"(%535, %arg65, %536) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<256x1x1xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc715) + %538 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc716) + %539 = "ttir.add"(%537, %arg66, %538) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<256x1x1xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc716) + %540 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc717) + %541 = "ttir.relu"(%539, %540) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc717) + %542 = tensor.empty() : tensor<1x14x256x14xf32> loc(#loc718) + %543 = "ttir.transpose"(%541, %542) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<1x14x256x14xf32>) -> tensor<1x14x256x14xf32> loc(#loc718) + %544 = tensor.empty() : tensor<1x14x14x256xf32> loc(#loc719) + %545 = "ttir.transpose"(%543, %544) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x256x14xf32>, tensor<1x14x14x256xf32>) -> tensor<1x14x14x256xf32> loc(#loc719) + %546 = tensor.empty() : tensor<1x14x14x1024xf32> loc(#loc720) + %547 = "ttir.conv2d"(%545, %arg140, %546) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x14x14x256xf32>, tensor<1024x256x1x1xf32>, tensor<1x14x14x1024xf32>) -> tensor<1x14x14x1024xf32> loc(#loc720) + %548 = tensor.empty() : tensor<1x14x1024x14xf32> loc(#loc721) + %549 = "ttir.transpose"(%547, %548) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x14x1024xf32>, tensor<1x14x1024x14xf32>) -> tensor<1x14x1024x14xf32> loc(#loc721) + %550 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc722) + %551 = "ttir.transpose"(%549, %550) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x1024x14xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc722) + %552 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc723) + %553 = "ttir.multiply"(%551, %arg67, %552) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1024x1x1xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc723) + %554 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc724) + %555 = "ttir.add"(%553, %arg68, %554) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1024x1x1xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc724) + %556 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc725) + %557 = "ttir.add"(%555, %509, %556) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1x1024x14x14xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc725) + %558 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc726) + %559 = "ttir.relu"(%557, %558) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc726) + %560 = tensor.empty() : tensor<1x14x1024x14xf32> loc(#loc727) + %561 = "ttir.transpose"(%559, %560) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1x14x1024x14xf32>) -> tensor<1x14x1024x14xf32> loc(#loc727) + %562 = tensor.empty() : tensor<1x14x14x1024xf32> loc(#loc728) + %563 = "ttir.transpose"(%561, %562) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x1024x14xf32>, tensor<1x14x14x1024xf32>) -> tensor<1x14x14x1024xf32> loc(#loc728) + %564 = tensor.empty() : tensor<1x14x14x256xf32> loc(#loc729) + %565 = "ttir.conv2d"(%563, %arg141, %564) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x14x14x1024xf32>, tensor<256x1024x1x1xf32>, tensor<1x14x14x256xf32>) -> tensor<1x14x14x256xf32> loc(#loc729) + %566 = tensor.empty() : tensor<1x14x256x14xf32> loc(#loc730) + %567 = "ttir.transpose"(%565, %566) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x14x256xf32>, tensor<1x14x256x14xf32>) -> tensor<1x14x256x14xf32> loc(#loc730) + %568 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc731) + %569 = "ttir.transpose"(%567, %568) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x256x14xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc731) + %570 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc732) + %571 = "ttir.multiply"(%569, %arg69, %570) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<256x1x1xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc732) + %572 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc733) + %573 = "ttir.add"(%571, %arg70, %572) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<256x1x1xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc733) + %574 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc734) + %575 = "ttir.relu"(%573, %574) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc734) + %576 = tensor.empty() : tensor<1x14x256x14xf32> loc(#loc735) + %577 = "ttir.transpose"(%575, %576) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<1x14x256x14xf32>) -> tensor<1x14x256x14xf32> loc(#loc735) + %578 = tensor.empty() : tensor<1x14x14x256xf32> loc(#loc736) + %579 = "ttir.transpose"(%577, %578) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x256x14xf32>, tensor<1x14x14x256xf32>) -> tensor<1x14x14x256xf32> loc(#loc736) + %580 = tensor.empty() : tensor<1x14x14x256xf32> loc(#loc737) + %581 = "ttir.conv2d"(%579, %arg142, %580) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 1 : si32, padding_left = 1 : si32, padding_right = 1 : si32, padding_top = 1 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x14x14x256xf32>, tensor<256x256x3x3xf32>, tensor<1x14x14x256xf32>) -> tensor<1x14x14x256xf32> loc(#loc737) + %582 = tensor.empty() : tensor<1x14x256x14xf32> loc(#loc738) + %583 = "ttir.transpose"(%581, %582) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x14x256xf32>, tensor<1x14x256x14xf32>) -> tensor<1x14x256x14xf32> loc(#loc738) + %584 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc739) + %585 = "ttir.transpose"(%583, %584) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x256x14xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc739) + %586 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc740) + %587 = "ttir.multiply"(%585, %arg71, %586) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<256x1x1xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc740) + %588 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc741) + %589 = "ttir.add"(%587, %arg72, %588) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<256x1x1xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc741) + %590 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc742) + %591 = "ttir.relu"(%589, %590) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc742) + %592 = tensor.empty() : tensor<1x14x256x14xf32> loc(#loc743) + %593 = "ttir.transpose"(%591, %592) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<1x14x256x14xf32>) -> tensor<1x14x256x14xf32> loc(#loc743) + %594 = tensor.empty() : tensor<1x14x14x256xf32> loc(#loc744) + %595 = "ttir.transpose"(%593, %594) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x256x14xf32>, tensor<1x14x14x256xf32>) -> tensor<1x14x14x256xf32> loc(#loc744) + %596 = tensor.empty() : tensor<1x14x14x1024xf32> loc(#loc745) + %597 = "ttir.conv2d"(%595, %arg143, %596) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x14x14x256xf32>, tensor<1024x256x1x1xf32>, tensor<1x14x14x1024xf32>) -> tensor<1x14x14x1024xf32> loc(#loc745) + %598 = tensor.empty() : tensor<1x14x1024x14xf32> loc(#loc746) + %599 = "ttir.transpose"(%597, %598) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x14x1024xf32>, tensor<1x14x1024x14xf32>) -> tensor<1x14x1024x14xf32> loc(#loc746) + %600 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc747) + %601 = "ttir.transpose"(%599, %600) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x1024x14xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc747) + %602 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc748) + %603 = "ttir.multiply"(%601, %arg73, %602) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1024x1x1xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc748) + %604 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc749) + %605 = "ttir.add"(%603, %arg74, %604) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1024x1x1xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc749) + %606 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc750) + %607 = "ttir.add"(%605, %559, %606) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1x1024x14x14xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc750) + %608 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc751) + %609 = "ttir.relu"(%607, %608) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc751) + %610 = tensor.empty() : tensor<1x14x1024x14xf32> loc(#loc752) + %611 = "ttir.transpose"(%609, %610) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1x14x1024x14xf32>) -> tensor<1x14x1024x14xf32> loc(#loc752) + %612 = tensor.empty() : tensor<1x14x14x1024xf32> loc(#loc753) + %613 = "ttir.transpose"(%611, %612) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x1024x14xf32>, tensor<1x14x14x1024xf32>) -> tensor<1x14x14x1024xf32> loc(#loc753) + %614 = tensor.empty() : tensor<1x14x14x256xf32> loc(#loc754) + %615 = "ttir.conv2d"(%613, %arg144, %614) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x14x14x1024xf32>, tensor<256x1024x1x1xf32>, tensor<1x14x14x256xf32>) -> tensor<1x14x14x256xf32> loc(#loc754) + %616 = tensor.empty() : tensor<1x14x256x14xf32> loc(#loc755) + %617 = "ttir.transpose"(%615, %616) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x14x256xf32>, tensor<1x14x256x14xf32>) -> tensor<1x14x256x14xf32> loc(#loc755) + %618 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc756) + %619 = "ttir.transpose"(%617, %618) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x256x14xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc756) + %620 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc757) + %621 = "ttir.multiply"(%619, %arg75, %620) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<256x1x1xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc757) + %622 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc758) + %623 = "ttir.add"(%621, %arg76, %622) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<256x1x1xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc758) + %624 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc759) + %625 = "ttir.relu"(%623, %624) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc759) + %626 = tensor.empty() : tensor<1x14x256x14xf32> loc(#loc760) + %627 = "ttir.transpose"(%625, %626) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<1x14x256x14xf32>) -> tensor<1x14x256x14xf32> loc(#loc760) + %628 = tensor.empty() : tensor<1x14x14x256xf32> loc(#loc761) + %629 = "ttir.transpose"(%627, %628) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x256x14xf32>, tensor<1x14x14x256xf32>) -> tensor<1x14x14x256xf32> loc(#loc761) + %630 = tensor.empty() : tensor<1x14x14x256xf32> loc(#loc762) + %631 = "ttir.conv2d"(%629, %arg145, %630) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 1 : si32, padding_left = 1 : si32, padding_right = 1 : si32, padding_top = 1 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x14x14x256xf32>, tensor<256x256x3x3xf32>, tensor<1x14x14x256xf32>) -> tensor<1x14x14x256xf32> loc(#loc762) + %632 = tensor.empty() : tensor<1x14x256x14xf32> loc(#loc763) + %633 = "ttir.transpose"(%631, %632) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x14x256xf32>, tensor<1x14x256x14xf32>) -> tensor<1x14x256x14xf32> loc(#loc763) + %634 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc764) + %635 = "ttir.transpose"(%633, %634) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x256x14xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc764) + %636 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc765) + %637 = "ttir.multiply"(%635, %arg77, %636) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<256x1x1xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc765) + %638 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc766) + %639 = "ttir.add"(%637, %arg78, %638) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<256x1x1xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc766) + %640 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc767) + %641 = "ttir.relu"(%639, %640) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc767) + %642 = tensor.empty() : tensor<1x14x256x14xf32> loc(#loc768) + %643 = "ttir.transpose"(%641, %642) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<1x14x256x14xf32>) -> tensor<1x14x256x14xf32> loc(#loc768) + %644 = tensor.empty() : tensor<1x14x14x256xf32> loc(#loc769) + %645 = "ttir.transpose"(%643, %644) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x256x14xf32>, tensor<1x14x14x256xf32>) -> tensor<1x14x14x256xf32> loc(#loc769) + %646 = tensor.empty() : tensor<1x14x14x1024xf32> loc(#loc770) + %647 = "ttir.conv2d"(%645, %arg146, %646) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x14x14x256xf32>, tensor<1024x256x1x1xf32>, tensor<1x14x14x1024xf32>) -> tensor<1x14x14x1024xf32> loc(#loc770) + %648 = tensor.empty() : tensor<1x14x1024x14xf32> loc(#loc771) + %649 = "ttir.transpose"(%647, %648) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x14x1024xf32>, tensor<1x14x1024x14xf32>) -> tensor<1x14x1024x14xf32> loc(#loc771) + %650 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc772) + %651 = "ttir.transpose"(%649, %650) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x1024x14xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc772) + %652 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc773) + %653 = "ttir.multiply"(%651, %arg79, %652) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1024x1x1xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc773) + %654 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc774) + %655 = "ttir.add"(%653, %arg80, %654) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1024x1x1xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc774) + %656 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc775) + %657 = "ttir.add"(%655, %609, %656) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1x1024x14x14xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc775) + %658 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc776) + %659 = "ttir.relu"(%657, %658) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc776) + %660 = tensor.empty() : tensor<1x14x1024x14xf32> loc(#loc777) + %661 = "ttir.transpose"(%659, %660) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1x14x1024x14xf32>) -> tensor<1x14x1024x14xf32> loc(#loc777) + %662 = tensor.empty() : tensor<1x14x14x1024xf32> loc(#loc778) + %663 = "ttir.transpose"(%661, %662) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x1024x14xf32>, tensor<1x14x14x1024xf32>) -> tensor<1x14x14x1024xf32> loc(#loc778) + %664 = tensor.empty() : tensor<1x14x14x256xf32> loc(#loc779) + %665 = "ttir.conv2d"(%663, %arg147, %664) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x14x14x1024xf32>, tensor<256x1024x1x1xf32>, tensor<1x14x14x256xf32>) -> tensor<1x14x14x256xf32> loc(#loc779) + %666 = tensor.empty() : tensor<1x14x256x14xf32> loc(#loc780) + %667 = "ttir.transpose"(%665, %666) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x14x256xf32>, tensor<1x14x256x14xf32>) -> tensor<1x14x256x14xf32> loc(#loc780) + %668 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc781) + %669 = "ttir.transpose"(%667, %668) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x256x14xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc781) + %670 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc782) + %671 = "ttir.multiply"(%669, %arg81, %670) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<256x1x1xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc782) + %672 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc783) + %673 = "ttir.add"(%671, %arg82, %672) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<256x1x1xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc783) + %674 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc784) + %675 = "ttir.relu"(%673, %674) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc784) + %676 = tensor.empty() : tensor<1x14x256x14xf32> loc(#loc785) + %677 = "ttir.transpose"(%675, %676) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<1x14x256x14xf32>) -> tensor<1x14x256x14xf32> loc(#loc785) + %678 = tensor.empty() : tensor<1x14x14x256xf32> loc(#loc786) + %679 = "ttir.transpose"(%677, %678) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x256x14xf32>, tensor<1x14x14x256xf32>) -> tensor<1x14x14x256xf32> loc(#loc786) + %680 = tensor.empty() : tensor<1x14x14x256xf32> loc(#loc787) + %681 = "ttir.conv2d"(%679, %arg148, %680) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 1 : si32, padding_left = 1 : si32, padding_right = 1 : si32, padding_top = 1 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x14x14x256xf32>, tensor<256x256x3x3xf32>, tensor<1x14x14x256xf32>) -> tensor<1x14x14x256xf32> loc(#loc787) + %682 = tensor.empty() : tensor<1x14x256x14xf32> loc(#loc788) + %683 = "ttir.transpose"(%681, %682) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x14x256xf32>, tensor<1x14x256x14xf32>) -> tensor<1x14x256x14xf32> loc(#loc788) + %684 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc789) + %685 = "ttir.transpose"(%683, %684) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x256x14xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc789) + %686 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc790) + %687 = "ttir.multiply"(%685, %arg83, %686) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<256x1x1xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc790) + %688 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc791) + %689 = "ttir.add"(%687, %arg84, %688) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<256x1x1xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc791) + %690 = tensor.empty() : tensor<1x256x14x14xf32> loc(#loc792) + %691 = "ttir.relu"(%689, %690) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<1x256x14x14xf32>) -> tensor<1x256x14x14xf32> loc(#loc792) + %692 = tensor.empty() : tensor<1x14x256x14xf32> loc(#loc793) + %693 = "ttir.transpose"(%691, %692) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x256x14x14xf32>, tensor<1x14x256x14xf32>) -> tensor<1x14x256x14xf32> loc(#loc793) + %694 = tensor.empty() : tensor<1x14x14x256xf32> loc(#loc794) + %695 = "ttir.transpose"(%693, %694) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x256x14xf32>, tensor<1x14x14x256xf32>) -> tensor<1x14x14x256xf32> loc(#loc794) + %696 = tensor.empty() : tensor<1x14x14x1024xf32> loc(#loc795) + %697 = "ttir.conv2d"(%695, %arg149, %696) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x14x14x256xf32>, tensor<1024x256x1x1xf32>, tensor<1x14x14x1024xf32>) -> tensor<1x14x14x1024xf32> loc(#loc795) + %698 = tensor.empty() : tensor<1x14x1024x14xf32> loc(#loc796) + %699 = "ttir.transpose"(%697, %698) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x14x1024xf32>, tensor<1x14x1024x14xf32>) -> tensor<1x14x1024x14xf32> loc(#loc796) + %700 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc797) + %701 = "ttir.transpose"(%699, %700) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x1024x14xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc797) + %702 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc798) + %703 = "ttir.multiply"(%701, %arg85, %702) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1024x1x1xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc798) + %704 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc799) + %705 = "ttir.add"(%703, %arg86, %704) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1024x1x1xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc799) + %706 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc800) + %707 = "ttir.add"(%705, %659, %706) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1x1024x14x14xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc800) + %708 = tensor.empty() : tensor<1x1024x14x14xf32> loc(#loc801) + %709 = "ttir.relu"(%707, %708) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1x1024x14x14xf32>) -> tensor<1x1024x14x14xf32> loc(#loc801) + %710 = tensor.empty() : tensor<1x14x1024x14xf32> loc(#loc802) + %711 = "ttir.transpose"(%709, %710) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1x14x1024x14xf32>) -> tensor<1x14x1024x14xf32> loc(#loc802) + %712 = tensor.empty() : tensor<1x14x14x1024xf32> loc(#loc803) + %713 = "ttir.transpose"(%711, %712) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x1024x14xf32>, tensor<1x14x14x1024xf32>) -> tensor<1x14x14x1024xf32> loc(#loc803) + %714 = tensor.empty() : tensor<1x14x14x512xf32> loc(#loc804) + %715 = "ttir.conv2d"(%713, %arg150, %714) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x14x14x1024xf32>, tensor<512x1024x1x1xf32>, tensor<1x14x14x512xf32>) -> tensor<1x14x14x512xf32> loc(#loc804) + %716 = tensor.empty() : tensor<1x14x512x14xf32> loc(#loc805) + %717 = "ttir.transpose"(%715, %716) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x14x512xf32>, tensor<1x14x512x14xf32>) -> tensor<1x14x512x14xf32> loc(#loc805) + %718 = tensor.empty() : tensor<1x512x14x14xf32> loc(#loc806) + %719 = "ttir.transpose"(%717, %718) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x512x14xf32>, tensor<1x512x14x14xf32>) -> tensor<1x512x14x14xf32> loc(#loc806) + %720 = tensor.empty() : tensor<1x512x14x14xf32> loc(#loc807) + %721 = "ttir.multiply"(%719, %arg87, %720) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x14x14xf32>, tensor<512x1x1xf32>, tensor<1x512x14x14xf32>) -> tensor<1x512x14x14xf32> loc(#loc807) + %722 = tensor.empty() : tensor<1x512x14x14xf32> loc(#loc808) + %723 = "ttir.add"(%721, %arg88, %722) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x14x14xf32>, tensor<512x1x1xf32>, tensor<1x512x14x14xf32>) -> tensor<1x512x14x14xf32> loc(#loc808) + %724 = tensor.empty() : tensor<1x512x14x14xf32> loc(#loc809) + %725 = "ttir.relu"(%723, %724) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x512x14x14xf32>, tensor<1x512x14x14xf32>) -> tensor<1x512x14x14xf32> loc(#loc809) + %726 = tensor.empty() : tensor<1x14x512x14xf32> loc(#loc810) + %727 = "ttir.transpose"(%725, %726) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x512x14x14xf32>, tensor<1x14x512x14xf32>) -> tensor<1x14x512x14xf32> loc(#loc810) + %728 = tensor.empty() : tensor<1x14x14x512xf32> loc(#loc811) + %729 = "ttir.transpose"(%727, %728) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x512x14xf32>, tensor<1x14x14x512xf32>) -> tensor<1x14x14x512xf32> loc(#loc811) + %730 = tensor.empty() : tensor<1x7x7x512xf32> loc(#loc812) + %731 = "ttir.conv2d"(%729, %arg151, %730) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 1 : si32, padding_left = 1 : si32, padding_right = 1 : si32, padding_top = 1 : si32, stride_height = 2 : si32, stride_width = 2 : si32}> {channel_last = 1 : si32} : (tensor<1x14x14x512xf32>, tensor<512x512x3x3xf32>, tensor<1x7x7x512xf32>) -> tensor<1x7x7x512xf32> loc(#loc812) + %732 = tensor.empty() : tensor<1x7x512x7xf32> loc(#loc813) + %733 = "ttir.transpose"(%731, %732) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x7x7x512xf32>, tensor<1x7x512x7xf32>) -> tensor<1x7x512x7xf32> loc(#loc813) + %734 = tensor.empty() : tensor<1x512x7x7xf32> loc(#loc814) + %735 = "ttir.transpose"(%733, %734) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x7x512x7xf32>, tensor<1x512x7x7xf32>) -> tensor<1x512x7x7xf32> loc(#loc814) + %736 = tensor.empty() : tensor<1x512x7x7xf32> loc(#loc815) + %737 = "ttir.multiply"(%735, %arg89, %736) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x7x7xf32>, tensor<512x1x1xf32>, tensor<1x512x7x7xf32>) -> tensor<1x512x7x7xf32> loc(#loc815) + %738 = tensor.empty() : tensor<1x512x7x7xf32> loc(#loc816) + %739 = "ttir.add"(%737, %arg90, %738) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x7x7xf32>, tensor<512x1x1xf32>, tensor<1x512x7x7xf32>) -> tensor<1x512x7x7xf32> loc(#loc816) + %740 = tensor.empty() : tensor<1x512x7x7xf32> loc(#loc817) + %741 = "ttir.relu"(%739, %740) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x512x7x7xf32>, tensor<1x512x7x7xf32>) -> tensor<1x512x7x7xf32> loc(#loc817) + %742 = tensor.empty() : tensor<1x7x512x7xf32> loc(#loc818) + %743 = "ttir.transpose"(%741, %742) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x512x7x7xf32>, tensor<1x7x512x7xf32>) -> tensor<1x7x512x7xf32> loc(#loc818) + %744 = tensor.empty() : tensor<1x7x7x512xf32> loc(#loc819) + %745 = "ttir.transpose"(%743, %744) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x7x512x7xf32>, tensor<1x7x7x512xf32>) -> tensor<1x7x7x512xf32> loc(#loc819) + %746 = tensor.empty() : tensor<1x7x7x2048xf32> loc(#loc820) + %747 = "ttir.conv2d"(%745, %arg152, %746) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x7x7x512xf32>, tensor<2048x512x1x1xf32>, tensor<1x7x7x2048xf32>) -> tensor<1x7x7x2048xf32> loc(#loc820) + %748 = tensor.empty() : tensor<1x7x2048x7xf32> loc(#loc821) + %749 = "ttir.transpose"(%747, %748) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x7x7x2048xf32>, tensor<1x7x2048x7xf32>) -> tensor<1x7x2048x7xf32> loc(#loc821) + %750 = tensor.empty() : tensor<1x2048x7x7xf32> loc(#loc822) + %751 = "ttir.transpose"(%749, %750) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x7x2048x7xf32>, tensor<1x2048x7x7xf32>) -> tensor<1x2048x7x7xf32> loc(#loc822) + %752 = tensor.empty() : tensor<1x2048x7x7xf32> loc(#loc823) + %753 = "ttir.multiply"(%751, %arg91, %752) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x2048x7x7xf32>, tensor<2048x1x1xf32>, tensor<1x2048x7x7xf32>) -> tensor<1x2048x7x7xf32> loc(#loc823) + %754 = tensor.empty() : tensor<1x2048x7x7xf32> loc(#loc824) + %755 = "ttir.add"(%753, %arg92, %754) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x2048x7x7xf32>, tensor<2048x1x1xf32>, tensor<1x2048x7x7xf32>) -> tensor<1x2048x7x7xf32> loc(#loc824) + %756 = tensor.empty() : tensor<1x14x1024x14xf32> loc(#loc825) + %757 = "ttir.transpose"(%709, %756) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x1024x14x14xf32>, tensor<1x14x1024x14xf32>) -> tensor<1x14x1024x14xf32> loc(#loc825) + %758 = tensor.empty() : tensor<1x14x14x1024xf32> loc(#loc826) + %759 = "ttir.transpose"(%757, %758) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x14x1024x14xf32>, tensor<1x14x14x1024xf32>) -> tensor<1x14x14x1024xf32> loc(#loc826) + %760 = tensor.empty() : tensor<1x7x7x2048xf32> loc(#loc827) + %761 = "ttir.conv2d"(%759, %arg153, %760) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 2 : si32, stride_width = 2 : si32}> {channel_last = 1 : si32} : (tensor<1x14x14x1024xf32>, tensor<2048x1024x1x1xf32>, tensor<1x7x7x2048xf32>) -> tensor<1x7x7x2048xf32> loc(#loc827) + %762 = tensor.empty() : tensor<1x7x2048x7xf32> loc(#loc828) + %763 = "ttir.transpose"(%761, %762) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x7x7x2048xf32>, tensor<1x7x2048x7xf32>) -> tensor<1x7x2048x7xf32> loc(#loc828) + %764 = tensor.empty() : tensor<1x2048x7x7xf32> loc(#loc829) + %765 = "ttir.transpose"(%763, %764) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x7x2048x7xf32>, tensor<1x2048x7x7xf32>) -> tensor<1x2048x7x7xf32> loc(#loc829) + %766 = tensor.empty() : tensor<1x2048x7x7xf32> loc(#loc830) + %767 = "ttir.multiply"(%765, %arg93, %766) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x2048x7x7xf32>, tensor<2048x1x1xf32>, tensor<1x2048x7x7xf32>) -> tensor<1x2048x7x7xf32> loc(#loc830) + %768 = tensor.empty() : tensor<1x2048x7x7xf32> loc(#loc831) + %769 = "ttir.add"(%767, %arg94, %768) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x2048x7x7xf32>, tensor<2048x1x1xf32>, tensor<1x2048x7x7xf32>) -> tensor<1x2048x7x7xf32> loc(#loc831) + %770 = tensor.empty() : tensor<1x2048x7x7xf32> loc(#loc832) + %771 = "ttir.add"(%755, %769, %770) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x2048x7x7xf32>, tensor<1x2048x7x7xf32>, tensor<1x2048x7x7xf32>) -> tensor<1x2048x7x7xf32> loc(#loc832) + %772 = tensor.empty() : tensor<1x2048x7x7xf32> loc(#loc833) + %773 = "ttir.relu"(%771, %772) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x2048x7x7xf32>, tensor<1x2048x7x7xf32>) -> tensor<1x2048x7x7xf32> loc(#loc833) + %774 = tensor.empty() : tensor<1x7x2048x7xf32> loc(#loc834) + %775 = "ttir.transpose"(%773, %774) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x2048x7x7xf32>, tensor<1x7x2048x7xf32>) -> tensor<1x7x2048x7xf32> loc(#loc834) + %776 = tensor.empty() : tensor<1x7x7x2048xf32> loc(#loc835) + %777 = "ttir.transpose"(%775, %776) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x7x2048x7xf32>, tensor<1x7x7x2048xf32>) -> tensor<1x7x7x2048xf32> loc(#loc835) + %778 = tensor.empty() : tensor<1x7x7x512xf32> loc(#loc836) + %779 = "ttir.conv2d"(%777, %arg154, %778) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x7x7x2048xf32>, tensor<512x2048x1x1xf32>, tensor<1x7x7x512xf32>) -> tensor<1x7x7x512xf32> loc(#loc836) + %780 = tensor.empty() : tensor<1x7x512x7xf32> loc(#loc837) + %781 = "ttir.transpose"(%779, %780) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x7x7x512xf32>, tensor<1x7x512x7xf32>) -> tensor<1x7x512x7xf32> loc(#loc837) + %782 = tensor.empty() : tensor<1x512x7x7xf32> loc(#loc838) + %783 = "ttir.transpose"(%781, %782) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x7x512x7xf32>, tensor<1x512x7x7xf32>) -> tensor<1x512x7x7xf32> loc(#loc838) + %784 = tensor.empty() : tensor<1x512x7x7xf32> loc(#loc839) + %785 = "ttir.multiply"(%783, %arg95, %784) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x7x7xf32>, tensor<512x1x1xf32>, tensor<1x512x7x7xf32>) -> tensor<1x512x7x7xf32> loc(#loc839) + %786 = tensor.empty() : tensor<1x512x7x7xf32> loc(#loc840) + %787 = "ttir.add"(%785, %arg96, %786) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x7x7xf32>, tensor<512x1x1xf32>, tensor<1x512x7x7xf32>) -> tensor<1x512x7x7xf32> loc(#loc840) + %788 = tensor.empty() : tensor<1x512x7x7xf32> loc(#loc841) + %789 = "ttir.relu"(%787, %788) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x512x7x7xf32>, tensor<1x512x7x7xf32>) -> tensor<1x512x7x7xf32> loc(#loc841) + %790 = tensor.empty() : tensor<1x7x512x7xf32> loc(#loc842) + %791 = "ttir.transpose"(%789, %790) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x512x7x7xf32>, tensor<1x7x512x7xf32>) -> tensor<1x7x512x7xf32> loc(#loc842) + %792 = tensor.empty() : tensor<1x7x7x512xf32> loc(#loc843) + %793 = "ttir.transpose"(%791, %792) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x7x512x7xf32>, tensor<1x7x7x512xf32>) -> tensor<1x7x7x512xf32> loc(#loc843) + %794 = tensor.empty() : tensor<1x7x7x512xf32> loc(#loc844) + %795 = "ttir.conv2d"(%793, %arg155, %794) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 1 : si32, padding_left = 1 : si32, padding_right = 1 : si32, padding_top = 1 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x7x7x512xf32>, tensor<512x512x3x3xf32>, tensor<1x7x7x512xf32>) -> tensor<1x7x7x512xf32> loc(#loc844) + %796 = tensor.empty() : tensor<1x7x512x7xf32> loc(#loc845) + %797 = "ttir.transpose"(%795, %796) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x7x7x512xf32>, tensor<1x7x512x7xf32>) -> tensor<1x7x512x7xf32> loc(#loc845) + %798 = tensor.empty() : tensor<1x512x7x7xf32> loc(#loc846) + %799 = "ttir.transpose"(%797, %798) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x7x512x7xf32>, tensor<1x512x7x7xf32>) -> tensor<1x512x7x7xf32> loc(#loc846) + %800 = tensor.empty() : tensor<1x512x7x7xf32> loc(#loc847) + %801 = "ttir.multiply"(%799, %arg97, %800) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x7x7xf32>, tensor<512x1x1xf32>, tensor<1x512x7x7xf32>) -> tensor<1x512x7x7xf32> loc(#loc847) + %802 = tensor.empty() : tensor<1x512x7x7xf32> loc(#loc848) + %803 = "ttir.add"(%801, %arg98, %802) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x7x7xf32>, tensor<512x1x1xf32>, tensor<1x512x7x7xf32>) -> tensor<1x512x7x7xf32> loc(#loc848) + %804 = tensor.empty() : tensor<1x512x7x7xf32> loc(#loc849) + %805 = "ttir.relu"(%803, %804) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x512x7x7xf32>, tensor<1x512x7x7xf32>) -> tensor<1x512x7x7xf32> loc(#loc849) + %806 = tensor.empty() : tensor<1x7x512x7xf32> loc(#loc850) + %807 = "ttir.transpose"(%805, %806) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x512x7x7xf32>, tensor<1x7x512x7xf32>) -> tensor<1x7x512x7xf32> loc(#loc850) + %808 = tensor.empty() : tensor<1x7x7x512xf32> loc(#loc851) + %809 = "ttir.transpose"(%807, %808) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x7x512x7xf32>, tensor<1x7x7x512xf32>) -> tensor<1x7x7x512xf32> loc(#loc851) + %810 = tensor.empty() : tensor<1x7x7x2048xf32> loc(#loc852) + %811 = "ttir.conv2d"(%809, %arg156, %810) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x7x7x512xf32>, tensor<2048x512x1x1xf32>, tensor<1x7x7x2048xf32>) -> tensor<1x7x7x2048xf32> loc(#loc852) + %812 = tensor.empty() : tensor<1x7x2048x7xf32> loc(#loc853) + %813 = "ttir.transpose"(%811, %812) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x7x7x2048xf32>, tensor<1x7x2048x7xf32>) -> tensor<1x7x2048x7xf32> loc(#loc853) + %814 = tensor.empty() : tensor<1x2048x7x7xf32> loc(#loc854) + %815 = "ttir.transpose"(%813, %814) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x7x2048x7xf32>, tensor<1x2048x7x7xf32>) -> tensor<1x2048x7x7xf32> loc(#loc854) + %816 = tensor.empty() : tensor<1x2048x7x7xf32> loc(#loc855) + %817 = "ttir.multiply"(%815, %arg99, %816) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x2048x7x7xf32>, tensor<2048x1x1xf32>, tensor<1x2048x7x7xf32>) -> tensor<1x2048x7x7xf32> loc(#loc855) + %818 = tensor.empty() : tensor<1x2048x7x7xf32> loc(#loc856) + %819 = "ttir.add"(%817, %arg100, %818) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x2048x7x7xf32>, tensor<2048x1x1xf32>, tensor<1x2048x7x7xf32>) -> tensor<1x2048x7x7xf32> loc(#loc856) + %820 = tensor.empty() : tensor<1x2048x7x7xf32> loc(#loc857) + %821 = "ttir.add"(%819, %773, %820) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x2048x7x7xf32>, tensor<1x2048x7x7xf32>, tensor<1x2048x7x7xf32>) -> tensor<1x2048x7x7xf32> loc(#loc857) + %822 = tensor.empty() : tensor<1x2048x7x7xf32> loc(#loc858) + %823 = "ttir.relu"(%821, %822) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x2048x7x7xf32>, tensor<1x2048x7x7xf32>) -> tensor<1x2048x7x7xf32> loc(#loc858) + %824 = tensor.empty() : tensor<1x7x2048x7xf32> loc(#loc859) + %825 = "ttir.transpose"(%823, %824) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x2048x7x7xf32>, tensor<1x7x2048x7xf32>) -> tensor<1x7x2048x7xf32> loc(#loc859) + %826 = tensor.empty() : tensor<1x7x7x2048xf32> loc(#loc860) + %827 = "ttir.transpose"(%825, %826) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x7x2048x7xf32>, tensor<1x7x7x2048xf32>) -> tensor<1x7x7x2048xf32> loc(#loc860) + %828 = tensor.empty() : tensor<1x7x7x512xf32> loc(#loc861) + %829 = "ttir.conv2d"(%827, %arg157, %828) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x7x7x2048xf32>, tensor<512x2048x1x1xf32>, tensor<1x7x7x512xf32>) -> tensor<1x7x7x512xf32> loc(#loc861) + %830 = tensor.empty() : tensor<1x7x512x7xf32> loc(#loc862) + %831 = "ttir.transpose"(%829, %830) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x7x7x512xf32>, tensor<1x7x512x7xf32>) -> tensor<1x7x512x7xf32> loc(#loc862) + %832 = tensor.empty() : tensor<1x512x7x7xf32> loc(#loc863) + %833 = "ttir.transpose"(%831, %832) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x7x512x7xf32>, tensor<1x512x7x7xf32>) -> tensor<1x512x7x7xf32> loc(#loc863) + %834 = tensor.empty() : tensor<1x512x7x7xf32> loc(#loc864) + %835 = "ttir.multiply"(%833, %arg101, %834) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x7x7xf32>, tensor<512x1x1xf32>, tensor<1x512x7x7xf32>) -> tensor<1x512x7x7xf32> loc(#loc864) + %836 = tensor.empty() : tensor<1x512x7x7xf32> loc(#loc865) + %837 = "ttir.add"(%835, %arg102, %836) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x7x7xf32>, tensor<512x1x1xf32>, tensor<1x512x7x7xf32>) -> tensor<1x512x7x7xf32> loc(#loc865) + %838 = tensor.empty() : tensor<1x512x7x7xf32> loc(#loc866) + %839 = "ttir.relu"(%837, %838) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x512x7x7xf32>, tensor<1x512x7x7xf32>) -> tensor<1x512x7x7xf32> loc(#loc866) + %840 = tensor.empty() : tensor<1x7x512x7xf32> loc(#loc867) + %841 = "ttir.transpose"(%839, %840) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x512x7x7xf32>, tensor<1x7x512x7xf32>) -> tensor<1x7x512x7xf32> loc(#loc867) + %842 = tensor.empty() : tensor<1x7x7x512xf32> loc(#loc868) + %843 = "ttir.transpose"(%841, %842) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x7x512x7xf32>, tensor<1x7x7x512xf32>) -> tensor<1x7x7x512xf32> loc(#loc868) + %844 = tensor.empty() : tensor<1x7x7x512xf32> loc(#loc869) + %845 = "ttir.conv2d"(%843, %arg158, %844) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 1 : si32, padding_left = 1 : si32, padding_right = 1 : si32, padding_top = 1 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x7x7x512xf32>, tensor<512x512x3x3xf32>, tensor<1x7x7x512xf32>) -> tensor<1x7x7x512xf32> loc(#loc869) + %846 = tensor.empty() : tensor<1x7x512x7xf32> loc(#loc870) + %847 = "ttir.transpose"(%845, %846) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x7x7x512xf32>, tensor<1x7x512x7xf32>) -> tensor<1x7x512x7xf32> loc(#loc870) + %848 = tensor.empty() : tensor<1x512x7x7xf32> loc(#loc871) + %849 = "ttir.transpose"(%847, %848) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x7x512x7xf32>, tensor<1x512x7x7xf32>) -> tensor<1x512x7x7xf32> loc(#loc871) + %850 = tensor.empty() : tensor<1x512x7x7xf32> loc(#loc872) + %851 = "ttir.multiply"(%849, %arg103, %850) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x7x7xf32>, tensor<512x1x1xf32>, tensor<1x512x7x7xf32>) -> tensor<1x512x7x7xf32> loc(#loc872) + %852 = tensor.empty() : tensor<1x512x7x7xf32> loc(#loc873) + %853 = "ttir.add"(%851, %arg104, %852) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x512x7x7xf32>, tensor<512x1x1xf32>, tensor<1x512x7x7xf32>) -> tensor<1x512x7x7xf32> loc(#loc873) + %854 = tensor.empty() : tensor<1x512x7x7xf32> loc(#loc874) + %855 = "ttir.relu"(%853, %854) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x512x7x7xf32>, tensor<1x512x7x7xf32>) -> tensor<1x512x7x7xf32> loc(#loc874) + %856 = tensor.empty() : tensor<1x7x512x7xf32> loc(#loc875) + %857 = "ttir.transpose"(%855, %856) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x512x7x7xf32>, tensor<1x7x512x7xf32>) -> tensor<1x7x512x7xf32> loc(#loc875) + %858 = tensor.empty() : tensor<1x7x7x512xf32> loc(#loc876) + %859 = "ttir.transpose"(%857, %858) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x7x512x7xf32>, tensor<1x7x7x512xf32>) -> tensor<1x7x7x512xf32> loc(#loc876) + %860 = tensor.empty() : tensor<1x7x7x2048xf32> loc(#loc877) + %861 = "ttir.conv2d"(%859, %arg159, %860) <{dilation_height = 1 : si32, dilation_width = 1 : si32, groups = 1 : si32, operand_constraints = [#any_device, #any_device, #any_device], padding_bottom = 0 : si32, padding_left = 0 : si32, padding_right = 0 : si32, padding_top = 0 : si32, stride_height = 1 : si32, stride_width = 1 : si32}> {channel_last = 1 : si32} : (tensor<1x7x7x512xf32>, tensor<2048x512x1x1xf32>, tensor<1x7x7x2048xf32>) -> tensor<1x7x7x2048xf32> loc(#loc877) + %862 = tensor.empty() : tensor<1x7x2048x7xf32> loc(#loc878) + %863 = "ttir.transpose"(%861, %862) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x7x7x2048xf32>, tensor<1x7x2048x7xf32>) -> tensor<1x7x2048x7xf32> loc(#loc878) + %864 = tensor.empty() : tensor<1x2048x7x7xf32> loc(#loc879) + %865 = "ttir.transpose"(%863, %864) <{dim0 = -3 : si32, dim1 = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x7x2048x7xf32>, tensor<1x2048x7x7xf32>) -> tensor<1x2048x7x7xf32> loc(#loc879) + %866 = tensor.empty() : tensor<1x2048x7x7xf32> loc(#loc880) + %867 = "ttir.multiply"(%865, %arg105, %866) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x2048x7x7xf32>, tensor<2048x1x1xf32>, tensor<1x2048x7x7xf32>) -> tensor<1x2048x7x7xf32> loc(#loc880) + %868 = tensor.empty() : tensor<1x2048x7x7xf32> loc(#loc881) + %869 = "ttir.add"(%867, %arg106, %868) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x2048x7x7xf32>, tensor<2048x1x1xf32>, tensor<1x2048x7x7xf32>) -> tensor<1x2048x7x7xf32> loc(#loc881) + %870 = tensor.empty() : tensor<1x2048x7x7xf32> loc(#loc882) + %871 = "ttir.add"(%869, %823, %870) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x2048x7x7xf32>, tensor<1x2048x7x7xf32>, tensor<1x2048x7x7xf32>) -> tensor<1x2048x7x7xf32> loc(#loc882) + %872 = tensor.empty() : tensor<1x2048x7x7xf32> loc(#loc883) + %873 = "ttir.relu"(%871, %872) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device]}> : (tensor<1x2048x7x7xf32>, tensor<1x2048x7x7xf32>) -> tensor<1x2048x7x7xf32> loc(#loc883) + %874 = tensor.empty() : tensor<1x1x2048x49xf32> loc(#loc884) + %875 = "ttir.reshape"(%873, %874) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 1 : i32, 2048 : i32, 49 : i32]}> : (tensor<1x2048x7x7xf32>, tensor<1x1x2048x49xf32>) -> tensor<1x1x2048x49xf32> loc(#loc884) + %876 = tensor.empty() : tensor<1x1x49x2048xf32> loc(#loc885) + %877 = "ttir.transpose"(%875, %876) <{dim0 = -2 : si32, dim1 = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x1x2048x49xf32>, tensor<1x1x49x2048xf32>) -> tensor<1x1x49x2048xf32> loc(#loc885) + %878 = tensor.empty() : tensor<1x1x1x2048xf32> loc(#loc886) + %879 = "ttir.mean"(%877, %878) <{keep_dim = true, operand_constraints = [#any_device, #any_device]}> {dim = -2 : si32} : (tensor<1x1x49x2048xf32>, tensor<1x1x1x2048xf32>) -> tensor<1x1x1x2048xf32> loc(#loc886) + %880 = tensor.empty() : tensor<1x2048x1x1xf32> loc(#loc887) + %881 = "ttir.reshape"(%879, %880) <{operand_constraints = [#any_device, #any_device], shape = [1 : i32, 2048 : i32, 1 : i32, 1 : i32]}> : (tensor<1x1x1x2048xf32>, tensor<1x2048x1x1xf32>) -> tensor<1x2048x1x1xf32> loc(#loc887) + %882 = tensor.empty() : tensor<1x2048x1xf32> loc(#loc888) + %883 = "ttir.squeeze"(%881, %882) <{dim = -2 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x2048x1x1xf32>, tensor<1x2048x1xf32>) -> tensor<1x2048x1xf32> loc(#loc888) + %884 = tensor.empty() : tensor<1x2048xf32> loc(#loc889) + %885 = "ttir.squeeze"(%883, %884) <{dim = -1 : si32, operand_constraints = [#any_device, #any_device]}> : (tensor<1x2048x1xf32>, tensor<1x2048xf32>) -> tensor<1x2048xf32> loc(#loc889) + %886 = tensor.empty() : tensor<1x1000xf32> loc(#loc890) + %887 = "ttir.matmul"(%885, %arg160, %886) <{operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x2048xf32>, tensor<2048x1000xf32>, tensor<1x1000xf32>) -> tensor<1x1000xf32> loc(#loc890) + %888 = tensor.empty() : tensor<1x1000xf32> loc(#loc891) + %889 = "ttir.add"(%887, %arg161, %888) <{operandSegmentSizes = array, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<1x1000xf32>, tensor<1000xf32>, tensor<1x1000xf32>) -> tensor<1x1000xf32> loc(#loc891) + return %889 : tensor<1x1000xf32> loc(#loc446) + } loc(#loc) +} loc(#loc) +#loc1 = loc("forward":4294967295:2951) +#loc2 = loc("forward":4294967295:2952) +#loc3 = loc("forward":4294967295:2954) +#loc4 = loc("forward":4294967295:2955) +#loc5 = loc("forward":4294967295:2956) +#loc6 = loc("forward":4294967295:2958) +#loc7 = loc("forward":4294967295:2960) +#loc8 = loc("forward":4294967295:2961) +#loc9 = loc("forward":4294967295:2962) +#loc10 = loc("forward":4294967295:2963) +#loc11 = loc("forward":4294967295:2964) +#loc12 = loc("forward":4294967295:2966) +#loc13 = loc("forward":4294967295:2967) +#loc14 = loc("forward":4294967295:2968) +#loc15 = loc("forward":4294967295:2970) +#loc16 = loc("forward":4294967295:2972) +#loc17 = loc("forward":4294967295:2973) +#loc18 = loc("forward":4294967295:2974) +#loc19 = loc("forward":4294967295:2975) +#loc20 = loc("forward":4294967295:2977) +#loc21 = loc("forward":4294967295:2978) +#loc22 = loc("forward":4294967295:2979) +#loc23 = loc("forward":4294967295:2981) +#loc24 = loc("forward":4294967295:2983) +#loc25 = loc("forward":4294967295:2984) +#loc26 = loc("forward":4294967295:2985) +#loc27 = loc("forward":4294967295:2986) +#loc28 = loc("forward":4294967295:2988) +#loc29 = loc("forward":4294967295:2989) +#loc30 = loc("forward":4294967295:2990) +#loc31 = loc("forward":4294967295:2992) +#loc32 = loc("forward":4294967295:2994) +#loc33 = loc("forward":4294967295:2995) +#loc34 = loc("forward":4294967295:2996) +#loc35 = loc("forward":4294967295:2998) +#loc36 = loc("forward":4294967295:2999) +#loc37 = loc("forward":4294967295:3000) +#loc38 = loc("forward":4294967295:3002) +#loc39 = loc("forward":4294967295:3004) +#loc40 = loc("forward":4294967295:3005) +#loc41 = loc("forward":4294967295:3006) +#loc42 = loc("forward":4294967295:3007) +#loc43 = loc("forward":4294967295:3008) +#loc44 = loc("forward":4294967295:3010) +#loc45 = loc("forward":4294967295:3011) +#loc46 = loc("forward":4294967295:3012) +#loc47 = loc("forward":4294967295:3014) +#loc48 = loc("forward":4294967295:3016) +#loc49 = loc("forward":4294967295:3017) +#loc50 = loc("forward":4294967295:3018) +#loc51 = loc("forward":4294967295:3019) +#loc52 = loc("forward":4294967295:3021) +#loc53 = loc("forward":4294967295:3022) +#loc54 = loc("forward":4294967295:3023) +#loc55 = loc("forward":4294967295:3025) +#loc56 = loc("forward":4294967295:3027) +#loc57 = loc("forward":4294967295:3028) +#loc58 = loc("forward":4294967295:3029) +#loc59 = loc("forward":4294967295:3030) +#loc60 = loc("forward":4294967295:3032) +#loc61 = loc("forward":4294967295:3033) +#loc62 = loc("forward":4294967295:3034) +#loc63 = loc("forward":4294967295:3036) +#loc64 = loc("forward":4294967295:3038) +#loc65 = loc("forward":4294967295:3039) +#loc66 = loc("forward":4294967295:3040) +#loc67 = loc("forward":4294967295:3041) +#loc68 = loc("forward":4294967295:3042) +#loc69 = loc("forward":4294967295:3044) +#loc70 = loc("forward":4294967295:3045) +#loc71 = loc("forward":4294967295:3046) +#loc72 = loc("forward":4294967295:3048) +#loc73 = loc("forward":4294967295:3050) +#loc74 = loc("forward":4294967295:3051) +#loc75 = loc("forward":4294967295:3052) +#loc76 = loc("forward":4294967295:3053) +#loc77 = loc("forward":4294967295:3055) +#loc78 = loc("forward":4294967295:3056) +#loc79 = loc("forward":4294967295:3057) +#loc80 = loc("forward":4294967295:3059) +#loc81 = loc("forward":4294967295:3061) +#loc82 = loc("forward":4294967295:3062) +#loc83 = loc("forward":4294967295:3063) +#loc84 = loc("forward":4294967295:3064) +#loc85 = loc("forward":4294967295:3066) +#loc86 = loc("forward":4294967295:3067) +#loc87 = loc("forward":4294967295:3068) +#loc88 = loc("forward":4294967295:3070) +#loc89 = loc("forward":4294967295:3072) +#loc90 = loc("forward":4294967295:3073) +#loc91 = loc("forward":4294967295:3074) +#loc92 = loc("forward":4294967295:3075) +#loc93 = loc("forward":4294967295:3076) +#loc94 = loc("forward":4294967295:3078) +#loc95 = loc("forward":4294967295:3079) +#loc96 = loc("forward":4294967295:3080) +#loc97 = loc("forward":4294967295:3082) +#loc98 = loc("forward":4294967295:3084) +#loc99 = loc("forward":4294967295:3085) +#loc100 = loc("forward":4294967295:3086) +#loc101 = loc("forward":4294967295:3087) +#loc102 = loc("forward":4294967295:3089) +#loc103 = loc("forward":4294967295:3090) +#loc104 = loc("forward":4294967295:3091) +#loc105 = loc("forward":4294967295:3093) +#loc106 = loc("forward":4294967295:3095) +#loc107 = loc("forward":4294967295:3096) +#loc108 = loc("forward":4294967295:3097) +#loc109 = loc("forward":4294967295:3098) +#loc110 = loc("forward":4294967295:3100) +#loc111 = loc("forward":4294967295:3101) +#loc112 = loc("forward":4294967295:3102) +#loc113 = loc("forward":4294967295:3104) +#loc114 = loc("forward":4294967295:3106) +#loc115 = loc("forward":4294967295:3107) +#loc116 = loc("forward":4294967295:3108) +#loc117 = loc("forward":4294967295:3110) +#loc118 = loc("forward":4294967295:3111) +#loc119 = loc("forward":4294967295:3112) +#loc120 = loc("forward":4294967295:3114) +#loc121 = loc("forward":4294967295:3116) +#loc122 = loc("forward":4294967295:3117) +#loc123 = loc("forward":4294967295:3118) +#loc124 = loc("forward":4294967295:3119) +#loc125 = loc("forward":4294967295:3120) +#loc126 = loc("forward":4294967295:3122) +#loc127 = loc("forward":4294967295:3123) +#loc128 = loc("forward":4294967295:3124) +#loc129 = loc("forward":4294967295:3126) +#loc130 = loc("forward":4294967295:3128) +#loc131 = loc("forward":4294967295:3129) +#loc132 = loc("forward":4294967295:3130) +#loc133 = loc("forward":4294967295:3131) +#loc134 = loc("forward":4294967295:3133) +#loc135 = loc("forward":4294967295:3134) +#loc136 = loc("forward":4294967295:3135) +#loc137 = loc("forward":4294967295:3137) +#loc138 = loc("forward":4294967295:3139) +#loc139 = loc("forward":4294967295:3140) +#loc140 = loc("forward":4294967295:3141) +#loc141 = loc("forward":4294967295:3142) +#loc142 = loc("forward":4294967295:3144) +#loc143 = loc("forward":4294967295:3145) +#loc144 = loc("forward":4294967295:3146) +#loc145 = loc("forward":4294967295:3148) +#loc146 = loc("forward":4294967295:3150) +#loc147 = loc("forward":4294967295:3151) +#loc148 = loc("forward":4294967295:3152) +#loc149 = loc("forward":4294967295:3153) +#loc150 = loc("forward":4294967295:3154) +#loc151 = loc("forward":4294967295:3156) +#loc152 = loc("forward":4294967295:3157) +#loc153 = loc("forward":4294967295:3158) +#loc154 = loc("forward":4294967295:3160) +#loc155 = loc("forward":4294967295:3162) +#loc156 = loc("forward":4294967295:3163) +#loc157 = loc("forward":4294967295:3164) +#loc158 = loc("forward":4294967295:3165) +#loc159 = loc("forward":4294967295:3167) +#loc160 = loc("forward":4294967295:3168) +#loc161 = loc("forward":4294967295:3169) +#loc162 = loc("forward":4294967295:3171) +#loc163 = loc("forward":4294967295:3173) +#loc164 = loc("forward":4294967295:3174) +#loc165 = loc("forward":4294967295:3175) +#loc166 = loc("forward":4294967295:3176) +#loc167 = loc("forward":4294967295:3178) +#loc168 = loc("forward":4294967295:3179) +#loc169 = loc("forward":4294967295:3180) +#loc170 = loc("forward":4294967295:3182) +#loc171 = loc("forward":4294967295:3184) +#loc172 = loc("forward":4294967295:3185) +#loc173 = loc("forward":4294967295:3186) +#loc174 = loc("forward":4294967295:3187) +#loc175 = loc("forward":4294967295:3188) +#loc176 = loc("forward":4294967295:3190) +#loc177 = loc("forward":4294967295:3191) +#loc178 = loc("forward":4294967295:3192) +#loc179 = loc("forward":4294967295:3194) +#loc180 = loc("forward":4294967295:3196) +#loc181 = loc("forward":4294967295:3197) +#loc182 = loc("forward":4294967295:3198) +#loc183 = loc("forward":4294967295:3199) +#loc184 = loc("forward":4294967295:3201) +#loc185 = loc("forward":4294967295:3202) +#loc186 = loc("forward":4294967295:3203) +#loc187 = loc("forward":4294967295:3205) +#loc188 = loc("forward":4294967295:3207) +#loc189 = loc("forward":4294967295:3208) +#loc190 = loc("forward":4294967295:3209) +#loc191 = loc("forward":4294967295:3210) +#loc192 = loc("forward":4294967295:3212) +#loc193 = loc("forward":4294967295:3213) +#loc194 = loc("forward":4294967295:3214) +#loc195 = loc("forward":4294967295:3216) +#loc196 = loc("forward":4294967295:3218) +#loc197 = loc("forward":4294967295:3219) +#loc198 = loc("forward":4294967295:3220) +#loc199 = loc("forward":4294967295:3221) +#loc200 = loc("forward":4294967295:3222) +#loc201 = loc("forward":4294967295:3224) +#loc202 = loc("forward":4294967295:3225) +#loc203 = loc("forward":4294967295:3226) +#loc204 = loc("forward":4294967295:3228) +#loc205 = loc("forward":4294967295:3230) +#loc206 = loc("forward":4294967295:3231) +#loc207 = loc("forward":4294967295:3232) +#loc208 = loc("forward":4294967295:3233) +#loc209 = loc("forward":4294967295:3235) +#loc210 = loc("forward":4294967295:3236) +#loc211 = loc("forward":4294967295:3237) +#loc212 = loc("forward":4294967295:3239) +#loc213 = loc("forward":4294967295:3241) +#loc214 = loc("forward":4294967295:3242) +#loc215 = loc("forward":4294967295:3243) +#loc216 = loc("forward":4294967295:3244) +#loc217 = loc("forward":4294967295:3246) +#loc218 = loc("forward":4294967295:3247) +#loc219 = loc("forward":4294967295:3248) +#loc220 = loc("forward":4294967295:3250) +#loc221 = loc("forward":4294967295:3252) +#loc222 = loc("forward":4294967295:3253) +#loc223 = loc("forward":4294967295:3254) +#loc224 = loc("forward":4294967295:3256) +#loc225 = loc("forward":4294967295:3257) +#loc226 = loc("forward":4294967295:3258) +#loc227 = loc("forward":4294967295:3260) +#loc228 = loc("forward":4294967295:3262) +#loc229 = loc("forward":4294967295:3263) +#loc230 = loc("forward":4294967295:3264) +#loc231 = loc("forward":4294967295:3265) +#loc232 = loc("forward":4294967295:3266) +#loc233 = loc("forward":4294967295:3268) +#loc234 = loc("forward":4294967295:3269) +#loc235 = loc("forward":4294967295:3270) +#loc236 = loc("forward":4294967295:3272) +#loc237 = loc("forward":4294967295:3274) +#loc238 = loc("forward":4294967295:3275) +#loc239 = loc("forward":4294967295:3276) +#loc240 = loc("forward":4294967295:3277) +#loc241 = loc("forward":4294967295:3279) +#loc242 = loc("forward":4294967295:3280) +#loc243 = loc("forward":4294967295:3281) +#loc244 = loc("forward":4294967295:3283) +#loc245 = loc("forward":4294967295:3285) +#loc246 = loc("forward":4294967295:3286) +#loc247 = loc("forward":4294967295:3287) +#loc248 = loc("forward":4294967295:3288) +#loc249 = loc("forward":4294967295:3290) +#loc250 = loc("forward":4294967295:3291) +#loc251 = loc("forward":4294967295:3292) +#loc252 = loc("forward":4294967295:3294) +#loc253 = loc("forward":4294967295:3296) +#loc254 = loc("forward":4294967295:3297) +#loc255 = loc("forward":4294967295:3298) +#loc256 = loc("forward":4294967295:3299) +#loc257 = loc("forward":4294967295:3300) +#loc258 = loc("forward":4294967295:3302) +#loc259 = loc("forward":4294967295:3303) +#loc260 = loc("forward":4294967295:3304) +#loc261 = loc("forward":4294967295:3306) +#loc262 = loc("forward":4294967295:3308) +#loc263 = loc("forward":4294967295:3309) +#loc264 = loc("forward":4294967295:3310) +#loc265 = loc("forward":4294967295:3311) +#loc266 = loc("forward":4294967295:3313) +#loc267 = loc("forward":4294967295:3314) +#loc268 = loc("forward":4294967295:3315) +#loc269 = loc("forward":4294967295:3317) +#loc270 = loc("forward":4294967295:3319) +#loc271 = loc("forward":4294967295:3320) +#loc272 = loc("forward":4294967295:3321) +#loc273 = loc("forward":4294967295:3322) +#loc274 = loc("forward":4294967295:3324) +#loc275 = loc("forward":4294967295:3325) +#loc276 = loc("forward":4294967295:3326) +#loc277 = loc("forward":4294967295:3328) +#loc278 = loc("forward":4294967295:3330) +#loc279 = loc("forward":4294967295:3331) +#loc280 = loc("forward":4294967295:3332) +#loc281 = loc("forward":4294967295:3333) +#loc282 = loc("forward":4294967295:3334) +#loc283 = loc("forward":4294967295:3336) +#loc284 = loc("forward":4294967295:3337) +#loc285 = loc("forward":4294967295:3338) +#loc286 = loc("forward":4294967295:3340) +#loc287 = loc("forward":4294967295:3342) +#loc288 = loc("forward":4294967295:3343) +#loc289 = loc("forward":4294967295:3344) +#loc290 = loc("forward":4294967295:3345) +#loc291 = loc("forward":4294967295:3347) +#loc292 = loc("forward":4294967295:3348) +#loc293 = loc("forward":4294967295:3349) +#loc294 = loc("forward":4294967295:3351) +#loc295 = loc("forward":4294967295:3353) +#loc296 = loc("forward":4294967295:3354) +#loc297 = loc("forward":4294967295:3355) +#loc298 = loc("forward":4294967295:3356) +#loc299 = loc("forward":4294967295:3358) +#loc300 = loc("forward":4294967295:3359) +#loc301 = loc("forward":4294967295:3360) +#loc302 = loc("forward":4294967295:3362) +#loc303 = loc("forward":4294967295:3364) +#loc304 = loc("forward":4294967295:3365) +#loc305 = loc("forward":4294967295:3366) +#loc306 = loc("forward":4294967295:3367) +#loc307 = loc("forward":4294967295:3368) +#loc308 = loc("forward":4294967295:3370) +#loc309 = loc("forward":4294967295:3371) +#loc310 = loc("forward":4294967295:3372) +#loc311 = loc("forward":4294967295:3374) +#loc312 = loc("forward":4294967295:3376) +#loc313 = loc("forward":4294967295:3377) +#loc314 = loc("forward":4294967295:3378) +#loc315 = loc("forward":4294967295:3379) +#loc316 = loc("forward":4294967295:3381) +#loc317 = loc("forward":4294967295:3382) +#loc318 = loc("forward":4294967295:3383) +#loc319 = loc("forward":4294967295:3385) +#loc320 = loc("forward":4294967295:3387) +#loc321 = loc("forward":4294967295:3388) +#loc322 = loc("forward":4294967295:3389) +#loc323 = loc("forward":4294967295:3390) +#loc324 = loc("forward":4294967295:3392) +#loc325 = loc("forward":4294967295:3393) +#loc326 = loc("forward":4294967295:3394) +#loc327 = loc("forward":4294967295:3396) +#loc328 = loc("forward":4294967295:3398) +#loc329 = loc("forward":4294967295:3399) +#loc330 = loc("forward":4294967295:3400) +#loc331 = loc("forward":4294967295:3401) +#loc332 = loc("forward":4294967295:3402) +#loc333 = loc("forward":4294967295:3404) +#loc334 = loc("forward":4294967295:3405) +#loc335 = loc("forward":4294967295:3406) +#loc336 = loc("forward":4294967295:3408) +#loc337 = loc("forward":4294967295:3410) +#loc338 = loc("forward":4294967295:3411) +#loc339 = loc("forward":4294967295:3412) +#loc340 = loc("forward":4294967295:3413) +#loc341 = loc("forward":4294967295:3415) +#loc342 = loc("forward":4294967295:3416) +#loc343 = loc("forward":4294967295:3417) +#loc344 = loc("forward":4294967295:3419) +#loc345 = loc("forward":4294967295:3421) +#loc346 = loc("forward":4294967295:3422) +#loc347 = loc("forward":4294967295:3423) +#loc348 = loc("forward":4294967295:3424) +#loc349 = loc("forward":4294967295:3426) +#loc350 = loc("forward":4294967295:3427) +#loc351 = loc("forward":4294967295:3428) +#loc352 = loc("forward":4294967295:3430) +#loc353 = loc("forward":4294967295:3432) +#loc354 = loc("forward":4294967295:3433) +#loc355 = loc("forward":4294967295:3434) +#loc356 = loc("forward":4294967295:3435) +#loc357 = loc("forward":4294967295:3436) +#loc358 = loc("forward":4294967295:3438) +#loc359 = loc("forward":4294967295:3439) +#loc360 = loc("forward":4294967295:3440) +#loc361 = loc("forward":4294967295:3442) +#loc362 = loc("forward":4294967295:3444) +#loc363 = loc("forward":4294967295:3445) +#loc364 = loc("forward":4294967295:3446) +#loc365 = loc("forward":4294967295:3447) +#loc366 = loc("forward":4294967295:3449) +#loc367 = loc("forward":4294967295:3450) +#loc368 = loc("forward":4294967295:3451) +#loc369 = loc("forward":4294967295:3453) +#loc370 = loc("forward":4294967295:3455) +#loc371 = loc("forward":4294967295:3456) +#loc372 = loc("forward":4294967295:3457) +#loc373 = loc("forward":4294967295:3458) +#loc374 = loc("forward":4294967295:3460) +#loc375 = loc("forward":4294967295:3461) +#loc376 = loc("forward":4294967295:3462) +#loc377 = loc("forward":4294967295:3464) +#loc378 = loc("forward":4294967295:3466) +#loc379 = loc("forward":4294967295:3467) +#loc380 = loc("forward":4294967295:3468) +#loc381 = loc("forward":4294967295:3470) +#loc382 = loc("forward":4294967295:3471) +#loc383 = loc("forward":4294967295:3472) +#loc384 = loc("forward":4294967295:3474) +#loc385 = loc("forward":4294967295:3476) +#loc386 = loc("forward":4294967295:3477) +#loc387 = loc("forward":4294967295:3478) +#loc388 = loc("forward":4294967295:3479) +#loc389 = loc("forward":4294967295:3480) +#loc390 = loc("forward":4294967295:3482) +#loc391 = loc("forward":4294967295:3483) +#loc392 = loc("forward":4294967295:3484) +#loc393 = loc("forward":4294967295:3486) +#loc394 = loc("forward":4294967295:3488) +#loc395 = loc("forward":4294967295:3489) +#loc396 = loc("forward":4294967295:3490) +#loc397 = loc("forward":4294967295:3491) +#loc398 = loc("forward":4294967295:3493) +#loc399 = loc("forward":4294967295:3494) +#loc400 = loc("forward":4294967295:3495) +#loc401 = loc("forward":4294967295:3497) +#loc402 = loc("forward":4294967295:3499) +#loc403 = loc("forward":4294967295:3500) +#loc404 = loc("forward":4294967295:3501) +#loc405 = loc("forward":4294967295:3502) +#loc406 = loc("forward":4294967295:3504) +#loc407 = loc("forward":4294967295:3505) +#loc408 = loc("forward":4294967295:3506) +#loc409 = loc("forward":4294967295:3508) +#loc410 = loc("forward":4294967295:3510) +#loc411 = loc("forward":4294967295:3511) +#loc412 = loc("forward":4294967295:3512) +#loc413 = loc("forward":4294967295:3513) +#loc414 = loc("forward":4294967295:3514) +#loc415 = loc("forward":4294967295:3516) +#loc416 = loc("forward":4294967295:3517) +#loc417 = loc("forward":4294967295:3518) +#loc418 = loc("forward":4294967295:3520) +#loc419 = loc("forward":4294967295:3522) +#loc420 = loc("forward":4294967295:3523) +#loc421 = loc("forward":4294967295:3524) +#loc422 = loc("forward":4294967295:3525) +#loc423 = loc("forward":4294967295:3527) +#loc424 = loc("forward":4294967295:3528) +#loc425 = loc("forward":4294967295:3529) +#loc426 = loc("forward":4294967295:3531) +#loc427 = loc("forward":4294967295:3533) +#loc428 = loc("forward":4294967295:3534) +#loc429 = loc("forward":4294967295:3535) +#loc430 = loc("forward":4294967295:3536) +#loc431 = loc("forward":4294967295:3538) +#loc432 = loc("forward":4294967295:3539) +#loc433 = loc("forward":4294967295:3540) +#loc434 = loc("forward":4294967295:3542) +#loc435 = loc("forward":4294967295:3544) +#loc436 = loc("forward":4294967295:3545) +#loc437 = loc("forward":4294967295:3546) +#loc438 = loc("forward":4294967295:3547) +#loc439 = loc("forward":4294967295:3548) +#loc440 = loc("forward":4294967295:3549) +#loc441 = loc("forward":4294967295:3550) +#loc442 = loc("forward":4294967295:3551) +#loc443 = loc("forward":4294967295:3552) +#loc444 = loc("forward":4294967295:3554) +#loc445 = loc("forward":4294967295:3556) +#loc446 = loc(unknown) +#loc447 = loc("conv2d_0.dc.transpose.0"(#loc1)) +#loc448 = loc("conv2d_0.dc.transpose.1"(#loc2)) +#loc449 = loc("conv2d_0.dc.conv2d.2"(#loc3)) +#loc450 = loc("conv2d_0.dc.transpose.3"(#loc4)) +#loc451 = loc("conv2d_0.dc.transpose.4"(#loc5)) +#loc452 = loc("multiply_8"(#loc6)) +#loc453 = loc("add_14"(#loc7)) +#loc454 = loc("relu_15"(#loc8)) +#loc455 = loc("max_pool2d_16"(#loc9)) +#loc456 = loc("conv2d_17.dc.transpose.0"(#loc10)) +#loc457 = loc("conv2d_17.dc.transpose.1"(#loc11)) +#loc458 = loc("conv2d_17.dc.conv2d.2"(#loc12)) +#loc459 = loc("conv2d_17.dc.transpose.3"(#loc13)) +#loc460 = loc("conv2d_17.dc.transpose.4"(#loc14)) +#loc461 = loc("multiply_25"(#loc15)) +#loc462 = loc("add_31"(#loc16)) +#loc463 = loc("relu_32"(#loc17)) +#loc464 = loc("conv2d_33.dc.transpose.0"(#loc18)) +#loc465 = loc("conv2d_33.dc.transpose.1"(#loc19)) +#loc466 = loc("conv2d_33.dc.conv2d.2"(#loc20)) +#loc467 = loc("conv2d_33.dc.transpose.3"(#loc21)) +#loc468 = loc("conv2d_33.dc.transpose.4"(#loc22)) +#loc469 = loc("multiply_41"(#loc23)) +#loc470 = loc("add_47"(#loc24)) +#loc471 = loc("relu_48"(#loc25)) +#loc472 = loc("conv2d_49.dc.transpose.0"(#loc26)) +#loc473 = loc("conv2d_49.dc.transpose.1"(#loc27)) +#loc474 = loc("conv2d_49.dc.conv2d.2"(#loc28)) +#loc475 = loc("conv2d_49.dc.transpose.3"(#loc29)) +#loc476 = loc("conv2d_49.dc.transpose.4"(#loc30)) +#loc477 = loc("multiply_57"(#loc31)) +#loc478 = loc("add_63"(#loc32)) +#loc479 = loc("conv2d_64.dc.transpose.0"(#loc33)) +#loc480 = loc("conv2d_64.dc.transpose.1"(#loc34)) +#loc481 = loc("conv2d_64.dc.conv2d.2"(#loc35)) +#loc482 = loc("conv2d_64.dc.transpose.3"(#loc36)) +#loc483 = loc("conv2d_64.dc.transpose.4"(#loc37)) +#loc484 = loc("multiply_72"(#loc38)) +#loc485 = loc("add_78"(#loc39)) +#loc486 = loc("add_79"(#loc40)) +#loc487 = loc("relu_80"(#loc41)) +#loc488 = loc("conv2d_81.dc.transpose.0"(#loc42)) +#loc489 = loc("conv2d_81.dc.transpose.1"(#loc43)) +#loc490 = loc("conv2d_81.dc.conv2d.2"(#loc44)) +#loc491 = loc("conv2d_81.dc.transpose.3"(#loc45)) +#loc492 = loc("conv2d_81.dc.transpose.4"(#loc46)) +#loc493 = loc("multiply_89"(#loc47)) +#loc494 = loc("add_95"(#loc48)) +#loc495 = loc("relu_96"(#loc49)) +#loc496 = loc("conv2d_97.dc.transpose.0"(#loc50)) +#loc497 = loc("conv2d_97.dc.transpose.1"(#loc51)) +#loc498 = loc("conv2d_97.dc.conv2d.2"(#loc52)) +#loc499 = loc("conv2d_97.dc.transpose.3"(#loc53)) +#loc500 = loc("conv2d_97.dc.transpose.4"(#loc54)) +#loc501 = loc("multiply_105"(#loc55)) +#loc502 = loc("add_111"(#loc56)) +#loc503 = loc("relu_112"(#loc57)) +#loc504 = loc("conv2d_113.dc.transpose.0"(#loc58)) +#loc505 = loc("conv2d_113.dc.transpose.1"(#loc59)) +#loc506 = loc("conv2d_113.dc.conv2d.2"(#loc60)) +#loc507 = loc("conv2d_113.dc.transpose.3"(#loc61)) +#loc508 = loc("conv2d_113.dc.transpose.4"(#loc62)) +#loc509 = loc("multiply_121"(#loc63)) +#loc510 = loc("add_127"(#loc64)) +#loc511 = loc("add_128"(#loc65)) +#loc512 = loc("relu_129"(#loc66)) +#loc513 = loc("conv2d_130.dc.transpose.0"(#loc67)) +#loc514 = loc("conv2d_130.dc.transpose.1"(#loc68)) +#loc515 = loc("conv2d_130.dc.conv2d.2"(#loc69)) +#loc516 = loc("conv2d_130.dc.transpose.3"(#loc70)) +#loc517 = loc("conv2d_130.dc.transpose.4"(#loc71)) +#loc518 = loc("multiply_138"(#loc72)) +#loc519 = loc("add_144"(#loc73)) +#loc520 = loc("relu_145"(#loc74)) +#loc521 = loc("conv2d_146.dc.transpose.0"(#loc75)) +#loc522 = loc("conv2d_146.dc.transpose.1"(#loc76)) +#loc523 = loc("conv2d_146.dc.conv2d.2"(#loc77)) +#loc524 = loc("conv2d_146.dc.transpose.3"(#loc78)) +#loc525 = loc("conv2d_146.dc.transpose.4"(#loc79)) +#loc526 = loc("multiply_154"(#loc80)) +#loc527 = loc("add_160"(#loc81)) +#loc528 = loc("relu_161"(#loc82)) +#loc529 = loc("conv2d_162.dc.transpose.0"(#loc83)) +#loc530 = loc("conv2d_162.dc.transpose.1"(#loc84)) +#loc531 = loc("conv2d_162.dc.conv2d.2"(#loc85)) +#loc532 = loc("conv2d_162.dc.transpose.3"(#loc86)) +#loc533 = loc("conv2d_162.dc.transpose.4"(#loc87)) +#loc534 = loc("multiply_170"(#loc88)) +#loc535 = loc("add_176"(#loc89)) +#loc536 = loc("add_177"(#loc90)) +#loc537 = loc("relu_178"(#loc91)) +#loc538 = loc("conv2d_179.dc.transpose.0"(#loc92)) +#loc539 = loc("conv2d_179.dc.transpose.1"(#loc93)) +#loc540 = loc("conv2d_179.dc.conv2d.2"(#loc94)) +#loc541 = loc("conv2d_179.dc.transpose.3"(#loc95)) +#loc542 = loc("conv2d_179.dc.transpose.4"(#loc96)) +#loc543 = loc("multiply_187"(#loc97)) +#loc544 = loc("add_193"(#loc98)) +#loc545 = loc("relu_194"(#loc99)) +#loc546 = loc("conv2d_195.dc.transpose.0"(#loc100)) +#loc547 = loc("conv2d_195.dc.transpose.1"(#loc101)) +#loc548 = loc("conv2d_195.dc.conv2d.2"(#loc102)) +#loc549 = loc("conv2d_195.dc.transpose.3"(#loc103)) +#loc550 = loc("conv2d_195.dc.transpose.4"(#loc104)) +#loc551 = loc("multiply_203"(#loc105)) +#loc552 = loc("add_209"(#loc106)) +#loc553 = loc("relu_210"(#loc107)) +#loc554 = loc("conv2d_211.dc.transpose.0"(#loc108)) +#loc555 = loc("conv2d_211.dc.transpose.1"(#loc109)) +#loc556 = loc("conv2d_211.dc.conv2d.2"(#loc110)) +#loc557 = loc("conv2d_211.dc.transpose.3"(#loc111)) +#loc558 = loc("conv2d_211.dc.transpose.4"(#loc112)) +#loc559 = loc("multiply_219"(#loc113)) +#loc560 = loc("add_225"(#loc114)) +#loc561 = loc("conv2d_226.dc.transpose.0"(#loc115)) +#loc562 = loc("conv2d_226.dc.transpose.1"(#loc116)) +#loc563 = loc("conv2d_226.dc.conv2d.2"(#loc117)) +#loc564 = loc("conv2d_226.dc.transpose.3"(#loc118)) +#loc565 = loc("conv2d_226.dc.transpose.4"(#loc119)) +#loc566 = loc("multiply_234"(#loc120)) +#loc567 = loc("add_240"(#loc121)) +#loc568 = loc("add_241"(#loc122)) +#loc569 = loc("relu_242"(#loc123)) +#loc570 = loc("conv2d_243.dc.transpose.0"(#loc124)) +#loc571 = loc("conv2d_243.dc.transpose.1"(#loc125)) +#loc572 = loc("conv2d_243.dc.conv2d.2"(#loc126)) +#loc573 = loc("conv2d_243.dc.transpose.3"(#loc127)) +#loc574 = loc("conv2d_243.dc.transpose.4"(#loc128)) +#loc575 = loc("multiply_251"(#loc129)) +#loc576 = loc("add_257"(#loc130)) +#loc577 = loc("relu_258"(#loc131)) +#loc578 = loc("conv2d_259.dc.transpose.0"(#loc132)) +#loc579 = loc("conv2d_259.dc.transpose.1"(#loc133)) +#loc580 = loc("conv2d_259.dc.conv2d.2"(#loc134)) +#loc581 = loc("conv2d_259.dc.transpose.3"(#loc135)) +#loc582 = loc("conv2d_259.dc.transpose.4"(#loc136)) +#loc583 = loc("multiply_267"(#loc137)) +#loc584 = loc("add_273"(#loc138)) +#loc585 = loc("relu_274"(#loc139)) +#loc586 = loc("conv2d_275.dc.transpose.0"(#loc140)) +#loc587 = loc("conv2d_275.dc.transpose.1"(#loc141)) +#loc588 = loc("conv2d_275.dc.conv2d.2"(#loc142)) +#loc589 = loc("conv2d_275.dc.transpose.3"(#loc143)) +#loc590 = loc("conv2d_275.dc.transpose.4"(#loc144)) +#loc591 = loc("multiply_283"(#loc145)) +#loc592 = loc("add_289"(#loc146)) +#loc593 = loc("add_290"(#loc147)) +#loc594 = loc("relu_291"(#loc148)) +#loc595 = loc("conv2d_292.dc.transpose.0"(#loc149)) +#loc596 = loc("conv2d_292.dc.transpose.1"(#loc150)) +#loc597 = loc("conv2d_292.dc.conv2d.2"(#loc151)) +#loc598 = loc("conv2d_292.dc.transpose.3"(#loc152)) +#loc599 = loc("conv2d_292.dc.transpose.4"(#loc153)) +#loc600 = loc("multiply_300"(#loc154)) +#loc601 = loc("add_306"(#loc155)) +#loc602 = loc("relu_307"(#loc156)) +#loc603 = loc("conv2d_308.dc.transpose.0"(#loc157)) +#loc604 = loc("conv2d_308.dc.transpose.1"(#loc158)) +#loc605 = loc("conv2d_308.dc.conv2d.2"(#loc159)) +#loc606 = loc("conv2d_308.dc.transpose.3"(#loc160)) +#loc607 = loc("conv2d_308.dc.transpose.4"(#loc161)) +#loc608 = loc("multiply_316"(#loc162)) +#loc609 = loc("add_322"(#loc163)) +#loc610 = loc("relu_323"(#loc164)) +#loc611 = loc("conv2d_324.dc.transpose.0"(#loc165)) +#loc612 = loc("conv2d_324.dc.transpose.1"(#loc166)) +#loc613 = loc("conv2d_324.dc.conv2d.2"(#loc167)) +#loc614 = loc("conv2d_324.dc.transpose.3"(#loc168)) +#loc615 = loc("conv2d_324.dc.transpose.4"(#loc169)) +#loc616 = loc("multiply_332"(#loc170)) +#loc617 = loc("add_338"(#loc171)) +#loc618 = loc("add_339"(#loc172)) +#loc619 = loc("relu_340"(#loc173)) +#loc620 = loc("conv2d_341.dc.transpose.0"(#loc174)) +#loc621 = loc("conv2d_341.dc.transpose.1"(#loc175)) +#loc622 = loc("conv2d_341.dc.conv2d.2"(#loc176)) +#loc623 = loc("conv2d_341.dc.transpose.3"(#loc177)) +#loc624 = loc("conv2d_341.dc.transpose.4"(#loc178)) +#loc625 = loc("multiply_349"(#loc179)) +#loc626 = loc("add_355"(#loc180)) +#loc627 = loc("relu_356"(#loc181)) +#loc628 = loc("conv2d_357.dc.transpose.0"(#loc182)) +#loc629 = loc("conv2d_357.dc.transpose.1"(#loc183)) +#loc630 = loc("conv2d_357.dc.conv2d.2"(#loc184)) +#loc631 = loc("conv2d_357.dc.transpose.3"(#loc185)) +#loc632 = loc("conv2d_357.dc.transpose.4"(#loc186)) +#loc633 = loc("multiply_365"(#loc187)) +#loc634 = loc("add_371"(#loc188)) +#loc635 = loc("relu_372"(#loc189)) +#loc636 = loc("conv2d_373.dc.transpose.0"(#loc190)) +#loc637 = loc("conv2d_373.dc.transpose.1"(#loc191)) +#loc638 = loc("conv2d_373.dc.conv2d.2"(#loc192)) +#loc639 = loc("conv2d_373.dc.transpose.3"(#loc193)) +#loc640 = loc("conv2d_373.dc.transpose.4"(#loc194)) +#loc641 = loc("multiply_381"(#loc195)) +#loc642 = loc("add_387"(#loc196)) +#loc643 = loc("add_388"(#loc197)) +#loc644 = loc("relu_389"(#loc198)) +#loc645 = loc("conv2d_390.dc.transpose.0"(#loc199)) +#loc646 = loc("conv2d_390.dc.transpose.1"(#loc200)) +#loc647 = loc("conv2d_390.dc.conv2d.2"(#loc201)) +#loc648 = loc("conv2d_390.dc.transpose.3"(#loc202)) +#loc649 = loc("conv2d_390.dc.transpose.4"(#loc203)) +#loc650 = loc("multiply_398"(#loc204)) +#loc651 = loc("add_404"(#loc205)) +#loc652 = loc("relu_405"(#loc206)) +#loc653 = loc("conv2d_406.dc.transpose.0"(#loc207)) +#loc654 = loc("conv2d_406.dc.transpose.1"(#loc208)) +#loc655 = loc("conv2d_406.dc.conv2d.2"(#loc209)) +#loc656 = loc("conv2d_406.dc.transpose.3"(#loc210)) +#loc657 = loc("conv2d_406.dc.transpose.4"(#loc211)) +#loc658 = loc("multiply_414"(#loc212)) +#loc659 = loc("add_420"(#loc213)) +#loc660 = loc("relu_421"(#loc214)) +#loc661 = loc("conv2d_422.dc.transpose.0"(#loc215)) +#loc662 = loc("conv2d_422.dc.transpose.1"(#loc216)) +#loc663 = loc("conv2d_422.dc.conv2d.2"(#loc217)) +#loc664 = loc("conv2d_422.dc.transpose.3"(#loc218)) +#loc665 = loc("conv2d_422.dc.transpose.4"(#loc219)) +#loc666 = loc("multiply_430"(#loc220)) +#loc667 = loc("add_436"(#loc221)) +#loc668 = loc("conv2d_437.dc.transpose.0"(#loc222)) +#loc669 = loc("conv2d_437.dc.transpose.1"(#loc223)) +#loc670 = loc("conv2d_437.dc.conv2d.2"(#loc224)) +#loc671 = loc("conv2d_437.dc.transpose.3"(#loc225)) +#loc672 = loc("conv2d_437.dc.transpose.4"(#loc226)) +#loc673 = loc("multiply_445"(#loc227)) +#loc674 = loc("add_451"(#loc228)) +#loc675 = loc("add_452"(#loc229)) +#loc676 = loc("relu_453"(#loc230)) +#loc677 = loc("conv2d_454.dc.transpose.0"(#loc231)) +#loc678 = loc("conv2d_454.dc.transpose.1"(#loc232)) +#loc679 = loc("conv2d_454.dc.conv2d.2"(#loc233)) +#loc680 = loc("conv2d_454.dc.transpose.3"(#loc234)) +#loc681 = loc("conv2d_454.dc.transpose.4"(#loc235)) +#loc682 = loc("multiply_462"(#loc236)) +#loc683 = loc("add_468"(#loc237)) +#loc684 = loc("relu_469"(#loc238)) +#loc685 = loc("conv2d_470.dc.transpose.0"(#loc239)) +#loc686 = loc("conv2d_470.dc.transpose.1"(#loc240)) +#loc687 = loc("conv2d_470.dc.conv2d.2"(#loc241)) +#loc688 = loc("conv2d_470.dc.transpose.3"(#loc242)) +#loc689 = loc("conv2d_470.dc.transpose.4"(#loc243)) +#loc690 = loc("multiply_478"(#loc244)) +#loc691 = loc("add_484"(#loc245)) +#loc692 = loc("relu_485"(#loc246)) +#loc693 = loc("conv2d_486.dc.transpose.0"(#loc247)) +#loc694 = loc("conv2d_486.dc.transpose.1"(#loc248)) +#loc695 = loc("conv2d_486.dc.conv2d.2"(#loc249)) +#loc696 = loc("conv2d_486.dc.transpose.3"(#loc250)) +#loc697 = loc("conv2d_486.dc.transpose.4"(#loc251)) +#loc698 = loc("multiply_494"(#loc252)) +#loc699 = loc("add_500"(#loc253)) +#loc700 = loc("add_501"(#loc254)) +#loc701 = loc("relu_502"(#loc255)) +#loc702 = loc("conv2d_503.dc.transpose.0"(#loc256)) +#loc703 = loc("conv2d_503.dc.transpose.1"(#loc257)) +#loc704 = loc("conv2d_503.dc.conv2d.2"(#loc258)) +#loc705 = loc("conv2d_503.dc.transpose.3"(#loc259)) +#loc706 = loc("conv2d_503.dc.transpose.4"(#loc260)) +#loc707 = loc("multiply_511"(#loc261)) +#loc708 = loc("add_517"(#loc262)) +#loc709 = loc("relu_518"(#loc263)) +#loc710 = loc("conv2d_519.dc.transpose.0"(#loc264)) +#loc711 = loc("conv2d_519.dc.transpose.1"(#loc265)) +#loc712 = loc("conv2d_519.dc.conv2d.2"(#loc266)) +#loc713 = loc("conv2d_519.dc.transpose.3"(#loc267)) +#loc714 = loc("conv2d_519.dc.transpose.4"(#loc268)) +#loc715 = loc("multiply_527"(#loc269)) +#loc716 = loc("add_533"(#loc270)) +#loc717 = loc("relu_534"(#loc271)) +#loc718 = loc("conv2d_535.dc.transpose.0"(#loc272)) +#loc719 = loc("conv2d_535.dc.transpose.1"(#loc273)) +#loc720 = loc("conv2d_535.dc.conv2d.2"(#loc274)) +#loc721 = loc("conv2d_535.dc.transpose.3"(#loc275)) +#loc722 = loc("conv2d_535.dc.transpose.4"(#loc276)) +#loc723 = loc("multiply_543"(#loc277)) +#loc724 = loc("add_549"(#loc278)) +#loc725 = loc("add_550"(#loc279)) +#loc726 = loc("relu_551"(#loc280)) +#loc727 = loc("conv2d_552.dc.transpose.0"(#loc281)) +#loc728 = loc("conv2d_552.dc.transpose.1"(#loc282)) +#loc729 = loc("conv2d_552.dc.conv2d.2"(#loc283)) +#loc730 = loc("conv2d_552.dc.transpose.3"(#loc284)) +#loc731 = loc("conv2d_552.dc.transpose.4"(#loc285)) +#loc732 = loc("multiply_560"(#loc286)) +#loc733 = loc("add_566"(#loc287)) +#loc734 = loc("relu_567"(#loc288)) +#loc735 = loc("conv2d_568.dc.transpose.0"(#loc289)) +#loc736 = loc("conv2d_568.dc.transpose.1"(#loc290)) +#loc737 = loc("conv2d_568.dc.conv2d.2"(#loc291)) +#loc738 = loc("conv2d_568.dc.transpose.3"(#loc292)) +#loc739 = loc("conv2d_568.dc.transpose.4"(#loc293)) +#loc740 = loc("multiply_576"(#loc294)) +#loc741 = loc("add_582"(#loc295)) +#loc742 = loc("relu_583"(#loc296)) +#loc743 = loc("conv2d_584.dc.transpose.0"(#loc297)) +#loc744 = loc("conv2d_584.dc.transpose.1"(#loc298)) +#loc745 = loc("conv2d_584.dc.conv2d.2"(#loc299)) +#loc746 = loc("conv2d_584.dc.transpose.3"(#loc300)) +#loc747 = loc("conv2d_584.dc.transpose.4"(#loc301)) +#loc748 = loc("multiply_592"(#loc302)) +#loc749 = loc("add_598"(#loc303)) +#loc750 = loc("add_599"(#loc304)) +#loc751 = loc("relu_600"(#loc305)) +#loc752 = loc("conv2d_601.dc.transpose.0"(#loc306)) +#loc753 = loc("conv2d_601.dc.transpose.1"(#loc307)) +#loc754 = loc("conv2d_601.dc.conv2d.2"(#loc308)) +#loc755 = loc("conv2d_601.dc.transpose.3"(#loc309)) +#loc756 = loc("conv2d_601.dc.transpose.4"(#loc310)) +#loc757 = loc("multiply_609"(#loc311)) +#loc758 = loc("add_615"(#loc312)) +#loc759 = loc("relu_616"(#loc313)) +#loc760 = loc("conv2d_617.dc.transpose.0"(#loc314)) +#loc761 = loc("conv2d_617.dc.transpose.1"(#loc315)) +#loc762 = loc("conv2d_617.dc.conv2d.2"(#loc316)) +#loc763 = loc("conv2d_617.dc.transpose.3"(#loc317)) +#loc764 = loc("conv2d_617.dc.transpose.4"(#loc318)) +#loc765 = loc("multiply_625"(#loc319)) +#loc766 = loc("add_631"(#loc320)) +#loc767 = loc("relu_632"(#loc321)) +#loc768 = loc("conv2d_633.dc.transpose.0"(#loc322)) +#loc769 = loc("conv2d_633.dc.transpose.1"(#loc323)) +#loc770 = loc("conv2d_633.dc.conv2d.2"(#loc324)) +#loc771 = loc("conv2d_633.dc.transpose.3"(#loc325)) +#loc772 = loc("conv2d_633.dc.transpose.4"(#loc326)) +#loc773 = loc("multiply_641"(#loc327)) +#loc774 = loc("add_647"(#loc328)) +#loc775 = loc("add_648"(#loc329)) +#loc776 = loc("relu_649"(#loc330)) +#loc777 = loc("conv2d_650.dc.transpose.0"(#loc331)) +#loc778 = loc("conv2d_650.dc.transpose.1"(#loc332)) +#loc779 = loc("conv2d_650.dc.conv2d.2"(#loc333)) +#loc780 = loc("conv2d_650.dc.transpose.3"(#loc334)) +#loc781 = loc("conv2d_650.dc.transpose.4"(#loc335)) +#loc782 = loc("multiply_658"(#loc336)) +#loc783 = loc("add_664"(#loc337)) +#loc784 = loc("relu_665"(#loc338)) +#loc785 = loc("conv2d_666.dc.transpose.0"(#loc339)) +#loc786 = loc("conv2d_666.dc.transpose.1"(#loc340)) +#loc787 = loc("conv2d_666.dc.conv2d.2"(#loc341)) +#loc788 = loc("conv2d_666.dc.transpose.3"(#loc342)) +#loc789 = loc("conv2d_666.dc.transpose.4"(#loc343)) +#loc790 = loc("multiply_674"(#loc344)) +#loc791 = loc("add_680"(#loc345)) +#loc792 = loc("relu_681"(#loc346)) +#loc793 = loc("conv2d_682.dc.transpose.0"(#loc347)) +#loc794 = loc("conv2d_682.dc.transpose.1"(#loc348)) +#loc795 = loc("conv2d_682.dc.conv2d.2"(#loc349)) +#loc796 = loc("conv2d_682.dc.transpose.3"(#loc350)) +#loc797 = loc("conv2d_682.dc.transpose.4"(#loc351)) +#loc798 = loc("multiply_690"(#loc352)) +#loc799 = loc("add_696"(#loc353)) +#loc800 = loc("add_697"(#loc354)) +#loc801 = loc("relu_698"(#loc355)) +#loc802 = loc("conv2d_699.dc.transpose.0"(#loc356)) +#loc803 = loc("conv2d_699.dc.transpose.1"(#loc357)) +#loc804 = loc("conv2d_699.dc.conv2d.2"(#loc358)) +#loc805 = loc("conv2d_699.dc.transpose.3"(#loc359)) +#loc806 = loc("conv2d_699.dc.transpose.4"(#loc360)) +#loc807 = loc("multiply_707"(#loc361)) +#loc808 = loc("add_713"(#loc362)) +#loc809 = loc("relu_714"(#loc363)) +#loc810 = loc("conv2d_715.dc.transpose.0"(#loc364)) +#loc811 = loc("conv2d_715.dc.transpose.1"(#loc365)) +#loc812 = loc("conv2d_715.dc.conv2d.2"(#loc366)) +#loc813 = loc("conv2d_715.dc.transpose.3"(#loc367)) +#loc814 = loc("conv2d_715.dc.transpose.4"(#loc368)) +#loc815 = loc("multiply_723"(#loc369)) +#loc816 = loc("add_729"(#loc370)) +#loc817 = loc("relu_730"(#loc371)) +#loc818 = loc("conv2d_731.dc.transpose.0"(#loc372)) +#loc819 = loc("conv2d_731.dc.transpose.1"(#loc373)) +#loc820 = loc("conv2d_731.dc.conv2d.2"(#loc374)) +#loc821 = loc("conv2d_731.dc.transpose.3"(#loc375)) +#loc822 = loc("conv2d_731.dc.transpose.4"(#loc376)) +#loc823 = loc("multiply_739"(#loc377)) +#loc824 = loc("add_745"(#loc378)) +#loc825 = loc("conv2d_746.dc.transpose.0"(#loc379)) +#loc826 = loc("conv2d_746.dc.transpose.1"(#loc380)) +#loc827 = loc("conv2d_746.dc.conv2d.2"(#loc381)) +#loc828 = loc("conv2d_746.dc.transpose.3"(#loc382)) +#loc829 = loc("conv2d_746.dc.transpose.4"(#loc383)) +#loc830 = loc("multiply_754"(#loc384)) +#loc831 = loc("add_760"(#loc385)) +#loc832 = loc("add_761"(#loc386)) +#loc833 = loc("relu_762"(#loc387)) +#loc834 = loc("conv2d_763.dc.transpose.0"(#loc388)) +#loc835 = loc("conv2d_763.dc.transpose.1"(#loc389)) +#loc836 = loc("conv2d_763.dc.conv2d.2"(#loc390)) +#loc837 = loc("conv2d_763.dc.transpose.3"(#loc391)) +#loc838 = loc("conv2d_763.dc.transpose.4"(#loc392)) +#loc839 = loc("multiply_771"(#loc393)) +#loc840 = loc("add_777"(#loc394)) +#loc841 = loc("relu_778"(#loc395)) +#loc842 = loc("conv2d_779.dc.transpose.0"(#loc396)) +#loc843 = loc("conv2d_779.dc.transpose.1"(#loc397)) +#loc844 = loc("conv2d_779.dc.conv2d.2"(#loc398)) +#loc845 = loc("conv2d_779.dc.transpose.3"(#loc399)) +#loc846 = loc("conv2d_779.dc.transpose.4"(#loc400)) +#loc847 = loc("multiply_787"(#loc401)) +#loc848 = loc("add_793"(#loc402)) +#loc849 = loc("relu_794"(#loc403)) +#loc850 = loc("conv2d_795.dc.transpose.0"(#loc404)) +#loc851 = loc("conv2d_795.dc.transpose.1"(#loc405)) +#loc852 = loc("conv2d_795.dc.conv2d.2"(#loc406)) +#loc853 = loc("conv2d_795.dc.transpose.3"(#loc407)) +#loc854 = loc("conv2d_795.dc.transpose.4"(#loc408)) +#loc855 = loc("multiply_803"(#loc409)) +#loc856 = loc("add_809"(#loc410)) +#loc857 = loc("add_810"(#loc411)) +#loc858 = loc("relu_811"(#loc412)) +#loc859 = loc("conv2d_812.dc.transpose.0"(#loc413)) +#loc860 = loc("conv2d_812.dc.transpose.1"(#loc414)) +#loc861 = loc("conv2d_812.dc.conv2d.2"(#loc415)) +#loc862 = loc("conv2d_812.dc.transpose.3"(#loc416)) +#loc863 = loc("conv2d_812.dc.transpose.4"(#loc417)) +#loc864 = loc("multiply_820"(#loc418)) +#loc865 = loc("add_826"(#loc419)) +#loc866 = loc("relu_827"(#loc420)) +#loc867 = loc("conv2d_828.dc.transpose.0"(#loc421)) +#loc868 = loc("conv2d_828.dc.transpose.1"(#loc422)) +#loc869 = loc("conv2d_828.dc.conv2d.2"(#loc423)) +#loc870 = loc("conv2d_828.dc.transpose.3"(#loc424)) +#loc871 = loc("conv2d_828.dc.transpose.4"(#loc425)) +#loc872 = loc("multiply_836"(#loc426)) +#loc873 = loc("add_842"(#loc427)) +#loc874 = loc("relu_843"(#loc428)) +#loc875 = loc("conv2d_844.dc.transpose.0"(#loc429)) +#loc876 = loc("conv2d_844.dc.transpose.1"(#loc430)) +#loc877 = loc("conv2d_844.dc.conv2d.2"(#loc431)) +#loc878 = loc("conv2d_844.dc.transpose.3"(#loc432)) +#loc879 = loc("conv2d_844.dc.transpose.4"(#loc433)) +#loc880 = loc("multiply_852"(#loc434)) +#loc881 = loc("add_858"(#loc435)) +#loc882 = loc("add_859"(#loc436)) +#loc883 = loc("relu_860"(#loc437)) +#loc884 = loc("avg_pool2d_861.dc.reshape.0"(#loc438)) +#loc885 = loc("avg_pool2d_861.dc.transpose.1.dc.transpose.0"(#loc439)) +#loc886 = loc("avg_pool2d_861.dc.reduce_avg.2"(#loc440)) +#loc887 = loc("avg_pool2d_861.dc.reshape.4"(#loc441)) +#loc888 = loc("squeeze_863"(#loc442)) +#loc889 = loc("squeeze_864"(#loc443)) +#loc890 = loc("matmul_866"(#loc444)) +#loc891 = loc("add_867"(#loc445)) diff --git a/tools/explorer/test/models/resnet_ttnn.mlir b/tools/explorer/test/models/resnet_ttnn.mlir new file mode 100644 index 000000000..1b1343aa8 --- /dev/null +++ b/tools/explorer/test/models/resnet_ttnn.mlir @@ -0,0 +1,2102 @@ + +#device = #tt.device (0, d0, d1)>, l1Map = (d0, d1)[s0, s1] -> (0, d0 floordiv s0, d1 floordiv s1, (d0 mod s0) * s1 + d1 mod s1), dramMap = (d0, d1)[s0, s1] -> (0, 0, ((((d0 floordiv s0) * 8 + d1 floordiv s1) * (s1 * s0) + (d0 mod s0) * s1 + d1 mod s1) floordiv 8192) mod 12, (((d0 floordiv s0) * 8 + d1 floordiv s1) * (s1 * s0) + (d0 mod s0) * s1 + d1 mod s1) floordiv 98304 + (((d0 floordiv s0) * 8 + d1 floordiv s1) * (s1 * s0) + (d0 mod s0) * s1 + d1 mod s1) mod 8192), meshShape = , chipIds = [0]> +#dram = #tt.memory_space +#loc = loc("ResNet":0:0) +#system = #tt.memory_space +#system_desc = #tt.system_desc<[{arch = , grid = 8x8, l1_size = 1499136, num_dram_channels = 12, dram_channel_size = 1073741824, noc_l1_address_align_bytes = 16, pcie_address_align_bytes = 32, noc_dram_address_align_bytes = 32, l1_unreserved_base = 1024, erisc_l1_unreserved_base = 1024, dram_unreserved_base = 1024, dram_unreserved_end = 1073741824, physical_cores = {worker = [ 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7, 1x0, 1x1, 1x2, 1x3, 1x4, 1x5, 1x6, 1x7, 2x0, 2x1, 2x2, 2x3, 2x4, 2x5, 2x6, 2x7, 3x0, 3x1, 3x2, 3x3, 3x4, 3x5, 3x6, 3x7, 4x0, 4x1, 4x2, 4x3, 4x4, 4x5, 4x6, 4x7, 5x0, 5x1, 5x2, 5x3, 5x4, 5x5, 5x6, 5x7, 6x0, 6x1, 6x2, 6x3, 6x4, 6x5, 6x6, 6x7, 7x0, 7x1, 7x2, 7x3, 7x4, 7x5, 7x6, 7x7] dram = [ 8x0, 9x0, 10x0, 8x1, 9x1, 10x1, 8x2, 9x2, 10x2, 8x3, 9x3, 10x3]}, supported_data_types = [, , , , , , , , , , , ], supported_tile_sizes = [ 4x16, 16x16, 32x16, 4x32, 16x32, 32x32]}], [0], [3 : i32], [ 0x0x0x0]> +#layout = #tt.layout<(d0, d1, d2, d3) -> (d0 * 672 + d1 * 224 + d2, d3), undef, <1x1>, memref<672x224xf32, #system>> +#layout1 = #tt.layout<(d0, d1, d2) -> (d0 + d1, d2), undef, <1x1>, memref<64x1xf32, #system>> +#layout2 = #tt.layout<(d0, d1, d2) -> (d0 + d1, d2), undef, <1x1>, memref<256x1xf32, #system>> +#layout3 = #tt.layout<(d0, d1, d2) -> (d0 + d1, d2), undef, <1x1>, memref<128x1xf32, #system>> +#layout4 = #tt.layout<(d0, d1, d2) -> (d0 + d1, d2), undef, <1x1>, memref<512x1xf32, #system>> +#layout5 = #tt.layout<(d0, d1, d2) -> (d0 + d1, d2), undef, <1x1>, memref<1024x1xf32, #system>> +#layout6 = #tt.layout<(d0, d1, d2) -> (d0 + d1, d2), undef, <1x1>, memref<2048x1xf32, #system>> +#layout7 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 21 + d1 * 7 + d2, d3), undef, <1x1>, memref<1344x7xf32, #system>> +#layout8 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 64 + d1 + d2, d3), undef, <1x1>, memref<4096x1xf32, #system>> +#layout9 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 192 + d1 * 3 + d2, d3), undef, <1x1>, memref<12288x3xf32, #system>> +#layout10 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 64 + d1 + d2, d3), undef, <1x1>, memref<16384x1xf32, #system>> +#layout11 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 256 + d1 + d2, d3), undef, <1x1>, memref<16384x1xf32, #system>> +#layout12 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 256 + d1 + d2, d3), undef, <1x1>, memref<32768x1xf32, #system>> +#layout13 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 384 + d1 * 3 + d2, d3), undef, <1x1>, memref<49152x3xf32, #system>> +#layout14 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 128 + d1 + d2, d3), undef, <1x1>, memref<65536x1xf32, #system>> +#layout15 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 256 + d1 + d2, d3), undef, <1x1>, memref<131072x1xf32, #system>> +#layout16 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 512 + d1 + d2, d3), undef, <1x1>, memref<65536x1xf32, #system>> +#layout17 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 512 + d1 + d2, d3), undef, <1x1>, memref<131072x1xf32, #system>> +#layout18 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 768 + d1 * 3 + d2, d3), undef, <1x1>, memref<196608x3xf32, #system>> +#layout19 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 256 + d1 + d2, d3), undef, <1x1>, memref<262144x1xf32, #system>> +#layout20 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 512 + d1 + d2, d3), undef, <1x1>, memref<524288x1xf32, #system>> +#layout21 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 1024 + d1 + d2, d3), undef, <1x1>, memref<262144x1xf32, #system>> +#layout22 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 1024 + d1 + d2, d3), undef, <1x1>, memref<524288x1xf32, #system>> +#layout23 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 1536 + d1 * 3 + d2, d3), undef, <1x1>, memref<786432x3xf32, #system>> +#layout24 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 512 + d1 + d2, d3), undef, <1x1>, memref<1048576x1xf32, #system>> +#layout25 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 1024 + d1 + d2, d3), undef, <1x1>, memref<2097152x1xf32, #system>> +#layout26 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 2048 + d1 + d2, d3), undef, <1x1>, memref<1048576x1xf32, #system>> +#layout27 = #tt.layout<(d0, d1) -> (d0, d1), undef, <1x1>, memref<2048x1000xf32, #system>> +#layout28 = #tt.layout<(d0) -> (0, d0), undef, <1x1>, memref<1x1000xf32, #system>> +#layout29 = #tt.layout<(d0, d1) -> (d0, d1), undef, <1x1>, memref<1x1000xf32, #system>> +#layout30 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 672 + d1 * 224 + d2, d3), undef, <1x1>, memref<672x224xf32, #dram>, interleaved> +#layout31 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 672 + d1 * 3 + d2, d3), undef, <1x1>, memref<672x224xf32, #dram>, interleaved> +#layout32 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 50176 + d1 * 224 + d2, d3), undef, <1x1>, memref<50176x3xf32, #dram>, interleaved> +#layout33 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 12544 + d1 * 112 + d2, d3), undef, <1x1>, memref<12544x64xf32, #dram>, interleaved> +#layout34 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 7168 + d1 * 64 + d2, d3), undef, <1x1>, memref<7168x112xf32, #dram>, interleaved> +#layout35 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 7168 + d1 * 112 + d2, d3), undef, <1x1>, memref<7168x112xf32, #dram>, interleaved> +#layout36 = #tt.layout<(d0, d1, d2) -> (d0 + d1, d2), undef, <1x1>, memref<64x1xf32, #dram>, interleaved> +#layout37 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 7168 + d1 * 7168 + d2, d3), undef, <1x1>, memref<7168x112xf32, #dram>, interleaved> +#layout38 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 3584 + d1 * 3584 + d2, d3), undef, <1x1>, memref<3584x56xf32, #dram>, interleaved> +#layout39 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 3584 + d1 * 56 + d2, d3), undef, <1x1>, memref<3584x56xf32, #dram>, interleaved> +#layout40 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 3584 + d1 * 64 + d2, d3), undef, <1x1>, memref<3584x56xf32, #dram>, interleaved> +#layout41 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 3136 + d1 * 56 + d2, d3), undef, <1x1>, memref<3136x64xf32, #dram>, interleaved> +#layout42 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 3136 + d1 * 56 + d2, d3), undef, <1x1>, memref<3136x256xf32, #dram>, interleaved> +#layout43 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 14336 + d1 * 256 + d2, d3), undef, <1x1>, memref<14336x56xf32, #dram>, interleaved> +#layout44 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 14336 + d1 * 56 + d2, d3), undef, <1x1>, memref<14336x56xf32, #dram>, interleaved> +#layout45 = #tt.layout<(d0, d1, d2) -> (d0 + d1, d2), undef, <1x1>, memref<256x1xf32, #dram>, interleaved> +#layout46 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 3136 + d1 * 56 + d2, d3), undef, <1x1>, memref<3136x128xf32, #dram>, interleaved> +#layout47 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 7168 + d1 * 128 + d2, d3), undef, <1x1>, memref<7168x56xf32, #dram>, interleaved> +#layout48 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 7168 + d1 * 56 + d2, d3), undef, <1x1>, memref<7168x56xf32, #dram>, interleaved> +#layout49 = #tt.layout<(d0, d1, d2) -> (d0 + d1, d2), undef, <1x1>, memref<128x1xf32, #dram>, interleaved> +#layout50 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 784 + d1 * 28 + d2, d3), undef, <1x1>, memref<784x128xf32, #dram>, interleaved> +#layout51 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 3584 + d1 * 128 + d2, d3), undef, <1x1>, memref<3584x28xf32, #dram>, interleaved> +#layout52 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 3584 + d1 * 28 + d2, d3), undef, <1x1>, memref<3584x28xf32, #dram>, interleaved> +#layout53 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 784 + d1 * 28 + d2, d3), undef, <1x1>, memref<784x512xf32, #dram>, interleaved> +#layout54 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 14336 + d1 * 512 + d2, d3), undef, <1x1>, memref<14336x28xf32, #dram>, interleaved> +#layout55 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 14336 + d1 * 28 + d2, d3), undef, <1x1>, memref<14336x28xf32, #dram>, interleaved> +#layout56 = #tt.layout<(d0, d1, d2) -> (d0 + d1, d2), undef, <1x1>, memref<512x1xf32, #dram>, interleaved> +#layout57 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 784 + d1 * 28 + d2, d3), undef, <1x1>, memref<784x256xf32, #dram>, interleaved> +#layout58 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 7168 + d1 * 256 + d2, d3), undef, <1x1>, memref<7168x28xf32, #dram>, interleaved> +#layout59 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 7168 + d1 * 28 + d2, d3), undef, <1x1>, memref<7168x28xf32, #dram>, interleaved> +#layout60 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 196 + d1 * 14 + d2, d3), undef, <1x1>, memref<196x256xf32, #dram>, interleaved> +#layout61 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 3584 + d1 * 256 + d2, d3), undef, <1x1>, memref<3584x14xf32, #dram>, interleaved> +#layout62 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 3584 + d1 * 14 + d2, d3), undef, <1x1>, memref<3584x14xf32, #dram>, interleaved> +#layout63 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 196 + d1 * 14 + d2, d3), undef, <1x1>, memref<196x1024xf32, #dram>, interleaved> +#layout64 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 14336 + d1 * 1024 + d2, d3), undef, <1x1>, memref<14336x14xf32, #dram>, interleaved> +#layout65 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 14336 + d1 * 14 + d2, d3), undef, <1x1>, memref<14336x14xf32, #dram>, interleaved> +#layout66 = #tt.layout<(d0, d1, d2) -> (d0 + d1, d2), undef, <1x1>, memref<1024x1xf32, #dram>, interleaved> +#layout67 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 196 + d1 * 14 + d2, d3), undef, <1x1>, memref<196x512xf32, #dram>, interleaved> +#layout68 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 7168 + d1 * 512 + d2, d3), undef, <1x1>, memref<7168x14xf32, #dram>, interleaved> +#layout69 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 7168 + d1 * 14 + d2, d3), undef, <1x1>, memref<7168x14xf32, #dram>, interleaved> +#layout70 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 49 + d1 * 7 + d2, d3), undef, <1x1>, memref<49x512xf32, #dram>, interleaved> +#layout71 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 3584 + d1 * 512 + d2, d3), undef, <1x1>, memref<3584x7xf32, #dram>, interleaved> +#layout72 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 3584 + d1 * 7 + d2, d3), undef, <1x1>, memref<3584x7xf32, #dram>, interleaved> +#layout73 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 49 + d1 * 7 + d2, d3), undef, <1x1>, memref<49x2048xf32, #dram>, interleaved> +#layout74 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 14336 + d1 * 2048 + d2, d3), undef, <1x1>, memref<14336x7xf32, #dram>, interleaved> +#layout75 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 14336 + d1 * 7 + d2, d3), undef, <1x1>, memref<14336x7xf32, #dram>, interleaved> +#layout76 = #tt.layout<(d0, d1, d2) -> (d0 + d1, d2), undef, <1x1>, memref<2048x1xf32, #dram>, interleaved> +#layout77 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 2048 + d1 * 2048 + d2, d3), undef, <1x1>, memref<2048x49xf32, #dram>, interleaved> +#layout78 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 49 + d1 * 49 + d2, d3), undef, <1x1>, memref<49x2048xf32, #dram>, interleaved> +#layout79 = #tt.layout<(d0, d1, d2, d3) -> (d0 + d1 + d2, d3), undef, <1x1>, memref<1x2048xf32, #dram>, interleaved> +#layout80 = #tt.layout<(d0, d1, d2, d3) -> (d0 * 2048 + d1 + d2, d3), undef, <1x1>, memref<2048x1xf32, #dram>, interleaved> +#layout81 = #tt.layout<(d0, d1, d2) -> (d0 * 2048 + d1, d2), undef, <1x1>, memref<2048x1xf32, #dram>, interleaved> +#layout82 = #tt.layout<(d0, d1) -> (d0, d1), undef, <1x1>, memref<1x2048xf32, #dram>, interleaved> +#layout83 = #tt.layout<(d0, d1) -> (d0, d1), undef, <1x1>, memref<2048x1000xf32, #dram>, interleaved> +#layout84 = #tt.layout<(d0, d1) -> (d0, d1), undef, <1x1>, memref<1x1000xf32, #dram>, interleaved> +#layout85 = #tt.layout<(d0) -> (0, d0), undef, <1x1>, memref<1x1000xf32, #dram>, interleaved> +module @ResNet attributes {tt.device = #device, tt.system_desc = #system_desc} { + func.func @forward(%arg0: tensor<1x3x224x224xf32, #layout> {ttir.name = "input_1"} loc("ResNet":0:0), %arg1: tensor<64x1x1xf32, #layout1> {ttir.name = "input_1_add_1"} loc("ResNet":0:0), %arg2: tensor<64x1x1xf32, #layout1> {ttir.name = "input_1_add_1_fork_clone1229"} loc("ResNet":0:0), %arg3: tensor<64x1x1xf32, #layout1> {ttir.name = "input_1_add_18"} loc("ResNet":0:0), %arg4: tensor<64x1x1xf32, #layout1> {ttir.name = "input_1_add_18_fork_clone1271"} loc("ResNet":0:0), %arg5: tensor<64x1x1xf32, #layout1> {ttir.name = "input_1_add_34"} loc("ResNet":0:0), %arg6: tensor<64x1x1xf32, #layout1> {ttir.name = "input_1_add_34_fork_clone1204"} loc("ResNet":0:0), %arg7: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_50"} loc("ResNet":0:0), %arg8: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_50_fork_clone1108"} loc("ResNet":0:0), %arg9: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_65"} loc("ResNet":0:0), %arg10: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_65_fork_clone1112"} loc("ResNet":0:0), %arg11: tensor<64x1x1xf32, #layout1> {ttir.name = "input_1_add_82"} loc("ResNet":0:0), %arg12: tensor<64x1x1xf32, #layout1> {ttir.name = "input_1_add_82_fork_clone1238"} loc("ResNet":0:0), %arg13: tensor<64x1x1xf32, #layout1> {ttir.name = "input_1_add_98"} loc("ResNet":0:0), %arg14: tensor<64x1x1xf32, #layout1> {ttir.name = "input_1_add_98_fork_clone1152"} loc("ResNet":0:0), %arg15: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_114"} loc("ResNet":0:0), %arg16: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_114_fork_clone1051"} loc("ResNet":0:0), %arg17: tensor<64x1x1xf32, #layout1> {ttir.name = "input_1_add_131"} loc("ResNet":0:0), %arg18: tensor<64x1x1xf32, #layout1> {ttir.name = "input_1_add_131_fork_clone1192"} loc("ResNet":0:0), %arg19: tensor<64x1x1xf32, #layout1> {ttir.name = "input_1_add_147"} loc("ResNet":0:0), %arg20: tensor<64x1x1xf32, #layout1> {ttir.name = "input_1_add_147_fork_clone1096"} loc("ResNet":0:0), %arg21: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_163"} loc("ResNet":0:0), %arg22: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_163_fork_clone992"} loc("ResNet":0:0), %arg23: tensor<128x1x1xf32, #layout3> {ttir.name = "input_1_add_180"} loc("ResNet":0:0), %arg24: tensor<128x1x1xf32, #layout3> {ttir.name = "input_1_add_180_fork_clone1065"} loc("ResNet":0:0), %arg25: tensor<128x1x1xf32, #layout3> {ttir.name = "input_1_add_196"} loc("ResNet":0:0), %arg26: tensor<128x1x1xf32, #layout3> {ttir.name = "input_1_add_196_fork_clone962"} loc("ResNet":0:0), %arg27: tensor<512x1x1xf32, #layout4> {ttir.name = "input_1_add_212"} loc("ResNet":0:0), %arg28: tensor<512x1x1xf32, #layout4> {ttir.name = "input_1_add_212_fork_clone853"} loc("ResNet":0:0), %arg29: tensor<512x1x1xf32, #layout4> {ttir.name = "input_1_add_227"} loc("ResNet":0:0), %arg30: tensor<512x1x1xf32, #layout4> {ttir.name = "input_1_add_227_fork_clone857"} loc("ResNet":0:0), %arg31: tensor<128x1x1xf32, #layout3> {ttir.name = "input_1_add_244"} loc("ResNet":0:0), %arg32: tensor<128x1x1xf32, #layout3> {ttir.name = "input_1_add_244_fork_clone1007"} loc("ResNet":0:0), %arg33: tensor<128x1x1xf32, #layout3> {ttir.name = "input_1_add_260"} loc("ResNet":0:0), %arg34: tensor<128x1x1xf32, #layout3> {ttir.name = "input_1_add_260_fork_clone901"} loc("ResNet":0:0), %arg35: tensor<512x1x1xf32, #layout4> {ttir.name = "input_1_add_276"} loc("ResNet":0:0), %arg36: tensor<512x1x1xf32, #layout4> {ttir.name = "input_1_add_276_fork_clone791"} loc("ResNet":0:0), %arg37: tensor<128x1x1xf32, #layout3> {ttir.name = "input_1_add_293"} loc("ResNet":0:0), %arg38: tensor<128x1x1xf32, #layout3> {ttir.name = "input_1_add_293_fork_clone950"} loc("ResNet":0:0), %arg39: tensor<128x1x1xf32, #layout3> {ttir.name = "input_1_add_309"} loc("ResNet":0:0), %arg40: tensor<128x1x1xf32, #layout3> {ttir.name = "input_1_add_309_fork_clone841"} loc("ResNet":0:0), %arg41: tensor<512x1x1xf32, #layout4> {ttir.name = "input_1_add_325"} loc("ResNet":0:0), %arg42: tensor<512x1x1xf32, #layout4> {ttir.name = "input_1_add_325_fork_clone735"} loc("ResNet":0:0), %arg43: tensor<128x1x1xf32, #layout3> {ttir.name = "input_1_add_342"} loc("ResNet":0:0), %arg44: tensor<128x1x1xf32, #layout3> {ttir.name = "input_1_add_342_fork_clone889"} loc("ResNet":0:0), %arg45: tensor<128x1x1xf32, #layout3> {ttir.name = "input_1_add_358"} loc("ResNet":0:0), %arg46: tensor<128x1x1xf32, #layout3> {ttir.name = "input_1_add_358_fork_clone779"} loc("ResNet":0:0), %arg47: tensor<512x1x1xf32, #layout4> {ttir.name = "input_1_add_374"} loc("ResNet":0:0), %arg48: tensor<512x1x1xf32, #layout4> {ttir.name = "input_1_add_374_fork_clone677"} loc("ResNet":0:0), %arg49: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_391"} loc("ResNet":0:0), %arg50: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_391_fork_clone748"} loc("ResNet":0:0), %arg51: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_407"} loc("ResNet":0:0), %arg52: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_407_fork_clone645"} loc("ResNet":0:0), %arg53: tensor<1024x1x1xf32, #layout5> {ttir.name = "input_1_add_423"} loc("ResNet":0:0), %arg54: tensor<1024x1x1xf32, #layout5> {ttir.name = "input_1_add_423_fork_clone524"} loc("ResNet":0:0), %arg55: tensor<1024x1x1xf32, #layout5> {ttir.name = "input_1_add_438"} loc("ResNet":0:0), %arg56: tensor<1024x1x1xf32, #layout5> {ttir.name = "input_1_add_438_fork_clone528"} loc("ResNet":0:0), %arg57: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_455"} loc("ResNet":0:0), %arg58: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_455_fork_clone692"} loc("ResNet":0:0), %arg59: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_471"} loc("ResNet":0:0), %arg60: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_471_fork_clone580"} loc("ResNet":0:0), %arg61: tensor<1024x1x1xf32, #layout5> {ttir.name = "input_1_add_487"} loc("ResNet":0:0), %arg62: tensor<1024x1x1xf32, #layout5> {ttir.name = "input_1_add_487_fork_clone453"} loc("ResNet":0:0), %arg63: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_504"} loc("ResNet":0:0), %arg64: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_504_fork_clone633"} loc("ResNet":0:0), %arg65: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_520"} loc("ResNet":0:0), %arg66: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_520_fork_clone512"} loc("ResNet":0:0), %arg67: tensor<1024x1x1xf32, #layout5> {ttir.name = "input_1_add_536"} loc("ResNet":0:0), %arg68: tensor<1024x1x1xf32, #layout5> {ttir.name = "input_1_add_536_fork_clone389"} loc("ResNet":0:0), %arg69: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_553"} loc("ResNet":0:0), %arg70: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_553_fork_clone568"} loc("ResNet":0:0), %arg71: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_569"} loc("ResNet":0:0), %arg72: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_569_fork_clone441"} loc("ResNet":0:0), %arg73: tensor<1024x1x1xf32, #layout5> {ttir.name = "input_1_add_585"} loc("ResNet":0:0), %arg74: tensor<1024x1x1xf32, #layout5> {ttir.name = "input_1_add_585_fork_clone329"} loc("ResNet":0:0), %arg75: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_602"} loc("ResNet":0:0), %arg76: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_602_fork_clone500"} loc("ResNet":0:0), %arg77: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_618"} loc("ResNet":0:0), %arg78: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_618_fork_clone377"} loc("ResNet":0:0), %arg79: tensor<1024x1x1xf32, #layout5> {ttir.name = "input_1_add_634"} loc("ResNet":0:0), %arg80: tensor<1024x1x1xf32, #layout5> {ttir.name = "input_1_add_634_fork_clone274"} loc("ResNet":0:0), %arg81: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_651"} loc("ResNet":0:0), %arg82: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_651_fork_clone429"} loc("ResNet":0:0), %arg83: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_667"} loc("ResNet":0:0), %arg84: tensor<256x1x1xf32, #layout2> {ttir.name = "input_1_add_667_fork_clone317"} loc("ResNet":0:0), %arg85: tensor<1024x1x1xf32, #layout5> {ttir.name = "input_1_add_683"} loc("ResNet":0:0), %arg86: tensor<1024x1x1xf32, #layout5> {ttir.name = "input_1_add_683_fork_clone219"} loc("ResNet":0:0), %arg87: tensor<512x1x1xf32, #layout4> {ttir.name = "input_1_add_700"} loc("ResNet":0:0), %arg88: tensor<512x1x1xf32, #layout4> {ttir.name = "input_1_add_700_fork_clone287"} loc("ResNet":0:0), %arg89: tensor<512x1x1xf32, #layout4> {ttir.name = "input_1_add_716"} loc("ResNet":0:0), %arg90: tensor<512x1x1xf32, #layout4> {ttir.name = "input_1_add_716_fork_clone190"} loc("ResNet":0:0), %arg91: tensor<2048x1x1xf32, #layout6> {ttir.name = "input_1_add_732"} loc("ResNet":0:0), %arg92: tensor<2048x1x1xf32, #layout6> {ttir.name = "input_1_add_732_fork_clone101"} loc("ResNet":0:0), %arg93: tensor<2048x1x1xf32, #layout6> {ttir.name = "input_1_add_747"} loc("ResNet":0:0), %arg94: tensor<2048x1x1xf32, #layout6> {ttir.name = "input_1_add_747_fork_clone105"} loc("ResNet":0:0), %arg95: tensor<512x1x1xf32, #layout4> {ttir.name = "input_1_add_764"} loc("ResNet":0:0), %arg96: tensor<512x1x1xf32, #layout4> {ttir.name = "input_1_add_764_fork_clone233"} loc("ResNet":0:0), %arg97: tensor<512x1x1xf32, #layout4> {ttir.name = "input_1_add_780"} loc("ResNet":0:0), %arg98: tensor<512x1x1xf32, #layout4> {ttir.name = "input_1_add_780_fork_clone138"} loc("ResNet":0:0), %arg99: tensor<2048x1x1xf32, #layout6> {ttir.name = "input_1_add_796"} loc("ResNet":0:0), %arg100: tensor<2048x1x1xf32, #layout6> {ttir.name = "input_1_add_796_fork_clone61"} loc("ResNet":0:0), %arg101: tensor<512x1x1xf32, #layout4> {ttir.name = "input_1_add_813"} loc("ResNet":0:0), %arg102: tensor<512x1x1xf32, #layout4> {ttir.name = "input_1_add_813_fork_clone178"} loc("ResNet":0:0), %arg103: tensor<512x1x1xf32, #layout4> {ttir.name = "input_1_add_829"} loc("ResNet":0:0), %arg104: tensor<512x1x1xf32, #layout4> {ttir.name = "input_1_add_829_fork_clone89"} loc("ResNet":0:0), %arg105: tensor<2048x1x1xf32, #layout6> {ttir.name = "input_1_add_845"} loc("ResNet":0:0), %arg106: tensor<2048x1x1xf32, #layout6> {ttir.name = "input_1_add_845_fork_clone32"} loc("ResNet":0:0), %arg107: tensor<64x3x7x7xf32, #layout7> {ttir.name = "conv1.weight"} loc("ResNet":0:0), %arg108: tensor<64x64x1x1xf32, #layout8> {ttir.name = "layer1.0.conv1.weight"} loc("ResNet":0:0), %arg109: tensor<64x64x3x3xf32, #layout9> {ttir.name = "layer1.0.conv2.weight"} loc("ResNet":0:0), %arg110: tensor<256x64x1x1xf32, #layout10> {ttir.name = "layer1.0.conv3.weight"} loc("ResNet":0:0), %arg111: tensor<256x64x1x1xf32, #layout10> {ttir.name = "layer1.0.downsample.0.weight"} loc("ResNet":0:0), %arg112: tensor<64x256x1x1xf32, #layout11> {ttir.name = "layer1.1.conv1.weight"} loc("ResNet":0:0), %arg113: tensor<64x64x3x3xf32, #layout9> {ttir.name = "layer1.1.conv2.weight"} loc("ResNet":0:0), %arg114: tensor<256x64x1x1xf32, #layout10> {ttir.name = "layer1.1.conv3.weight"} loc("ResNet":0:0), %arg115: tensor<64x256x1x1xf32, #layout11> {ttir.name = "layer1.2.conv1.weight"} loc("ResNet":0:0), %arg116: tensor<64x64x3x3xf32, #layout9> {ttir.name = "layer1.2.conv2.weight"} loc("ResNet":0:0), %arg117: tensor<256x64x1x1xf32, #layout10> {ttir.name = "layer1.2.conv3.weight"} loc("ResNet":0:0), %arg118: tensor<128x256x1x1xf32, #layout12> {ttir.name = "layer2.0.conv1.weight"} loc("ResNet":0:0), %arg119: tensor<128x128x3x3xf32, #layout13> {ttir.name = "layer2.0.conv2.weight"} loc("ResNet":0:0), %arg120: tensor<512x128x1x1xf32, #layout14> {ttir.name = "layer2.0.conv3.weight"} loc("ResNet":0:0), %arg121: tensor<512x256x1x1xf32, #layout15> {ttir.name = "layer2.0.downsample.0.weight"} loc("ResNet":0:0), %arg122: tensor<128x512x1x1xf32, #layout16> {ttir.name = "layer2.1.conv1.weight"} loc("ResNet":0:0), %arg123: tensor<128x128x3x3xf32, #layout13> {ttir.name = "layer2.1.conv2.weight"} loc("ResNet":0:0), %arg124: tensor<512x128x1x1xf32, #layout14> {ttir.name = "layer2.1.conv3.weight"} loc("ResNet":0:0), %arg125: tensor<128x512x1x1xf32, #layout16> {ttir.name = "layer2.2.conv1.weight"} loc("ResNet":0:0), %arg126: tensor<128x128x3x3xf32, #layout13> {ttir.name = "layer2.2.conv2.weight"} loc("ResNet":0:0), %arg127: tensor<512x128x1x1xf32, #layout14> {ttir.name = "layer2.2.conv3.weight"} loc("ResNet":0:0), %arg128: tensor<128x512x1x1xf32, #layout16> {ttir.name = "layer2.3.conv1.weight"} loc("ResNet":0:0), %arg129: tensor<128x128x3x3xf32, #layout13> {ttir.name = "layer2.3.conv2.weight"} loc("ResNet":0:0), %arg130: tensor<512x128x1x1xf32, #layout14> {ttir.name = "layer2.3.conv3.weight"} loc("ResNet":0:0), %arg131: tensor<256x512x1x1xf32, #layout17> {ttir.name = "layer3.0.conv1.weight"} loc("ResNet":0:0), %arg132: tensor<256x256x3x3xf32, #layout18> {ttir.name = "layer3.0.conv2.weight"} loc("ResNet":0:0), %arg133: tensor<1024x256x1x1xf32, #layout19> {ttir.name = "layer3.0.conv3.weight"} loc("ResNet":0:0), %arg134: tensor<1024x512x1x1xf32, #layout20> {ttir.name = "layer3.0.downsample.0.weight"} loc("ResNet":0:0), %arg135: tensor<256x1024x1x1xf32, #layout21> {ttir.name = "layer3.1.conv1.weight"} loc("ResNet":0:0), %arg136: tensor<256x256x3x3xf32, #layout18> {ttir.name = "layer3.1.conv2.weight"} loc("ResNet":0:0), %arg137: tensor<1024x256x1x1xf32, #layout19> {ttir.name = "layer3.1.conv3.weight"} loc("ResNet":0:0), %arg138: tensor<256x1024x1x1xf32, #layout21> {ttir.name = "layer3.2.conv1.weight"} loc("ResNet":0:0), %arg139: tensor<256x256x3x3xf32, #layout18> {ttir.name = "layer3.2.conv2.weight"} loc("ResNet":0:0), %arg140: tensor<1024x256x1x1xf32, #layout19> {ttir.name = "layer3.2.conv3.weight"} loc("ResNet":0:0), %arg141: tensor<256x1024x1x1xf32, #layout21> {ttir.name = "layer3.3.conv1.weight"} loc("ResNet":0:0), %arg142: tensor<256x256x3x3xf32, #layout18> {ttir.name = "layer3.3.conv2.weight"} loc("ResNet":0:0), %arg143: tensor<1024x256x1x1xf32, #layout19> {ttir.name = "layer3.3.conv3.weight"} loc("ResNet":0:0), %arg144: tensor<256x1024x1x1xf32, #layout21> {ttir.name = "layer3.4.conv1.weight"} loc("ResNet":0:0), %arg145: tensor<256x256x3x3xf32, #layout18> {ttir.name = "layer3.4.conv2.weight"} loc("ResNet":0:0), %arg146: tensor<1024x256x1x1xf32, #layout19> {ttir.name = "layer3.4.conv3.weight"} loc("ResNet":0:0), %arg147: tensor<256x1024x1x1xf32, #layout21> {ttir.name = "layer3.5.conv1.weight"} loc("ResNet":0:0), %arg148: tensor<256x256x3x3xf32, #layout18> {ttir.name = "layer3.5.conv2.weight"} loc("ResNet":0:0), %arg149: tensor<1024x256x1x1xf32, #layout19> {ttir.name = "layer3.5.conv3.weight"} loc("ResNet":0:0), %arg150: tensor<512x1024x1x1xf32, #layout22> {ttir.name = "layer4.0.conv1.weight"} loc("ResNet":0:0), %arg151: tensor<512x512x3x3xf32, #layout23> {ttir.name = "layer4.0.conv2.weight"} loc("ResNet":0:0), %arg152: tensor<2048x512x1x1xf32, #layout24> {ttir.name = "layer4.0.conv3.weight"} loc("ResNet":0:0), %arg153: tensor<2048x1024x1x1xf32, #layout25> {ttir.name = "layer4.0.downsample.0.weight"} loc("ResNet":0:0), %arg154: tensor<512x2048x1x1xf32, #layout26> {ttir.name = "layer4.1.conv1.weight"} loc("ResNet":0:0), %arg155: tensor<512x512x3x3xf32, #layout23> {ttir.name = "layer4.1.conv2.weight"} loc("ResNet":0:0), %arg156: tensor<2048x512x1x1xf32, #layout24> {ttir.name = "layer4.1.conv3.weight"} loc("ResNet":0:0), %arg157: tensor<512x2048x1x1xf32, #layout26> {ttir.name = "layer4.2.conv1.weight"} loc("ResNet":0:0), %arg158: tensor<512x512x3x3xf32, #layout23> {ttir.name = "layer4.2.conv2.weight"} loc("ResNet":0:0), %arg159: tensor<2048x512x1x1xf32, #layout24> {ttir.name = "layer4.2.conv3.weight"} loc("ResNet":0:0), %arg160: tensor<2048x1000xf32, #layout27> {ttir.name = "fc.weight"} loc("ResNet":0:0), %arg161: tensor<1000xf32, #layout28> {ttir.name = "fc.bias"} loc("ResNet":0:0)) -> (tensor<1x1000xf32, #layout29> {ttir.name = "ResNet.output_add_867"}) { + %0 = "ttnn.get_device"() <{mesh_shape = #ttnn}> : () -> !tt.device<#device> loc(#loc447) + %1 = "ttnn.to_layout"(%arg0, %0) <{layout = #ttnn.layout}> : (tensor<1x3x224x224xf32, #layout>, !tt.device<#device>) -> tensor<1x3x224x224xf32, #layout30> loc(#loc447) + %2 = "ttnn.to_device"(%1, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<1x3x224x224xf32, #layout30>, !tt.device<#device>) -> tensor<1x3x224x224xf32, #layout30> loc(#loc447) + %3 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x224x3x224>}> : (!tt.device<#device>) -> tensor<1x224x3x224xf32, #layout31> loc(#loc447) + %4 = "ttnn.transpose"(%2, %3) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x3x224x224xf32, #layout30>, tensor<1x224x3x224xf32, #layout31>) -> tensor<1x224x3x224xf32, #layout31> loc(#loc447) + %5 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x224x224x3>}> : (!tt.device<#device>) -> tensor<1x224x224x3xf32, #layout32> loc(#loc448) + %6 = "ttnn.transpose"(%4, %5) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x224x3x224xf32, #layout31>, tensor<1x224x224x3xf32, #layout32>) -> tensor<1x224x224x3xf32, #layout32> loc(#loc448) + %7 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x112x112x64>}> : (!tt.device<#device>) -> tensor<1x112x112x64xf32, #layout33> loc(#loc449) + %8 = "ttnn.conv2d"(%6, %arg107, %7, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 3 : i32, input_height = 224 : i32, input_width = 224 : i32, kernel_height = 7 : i32, kernel_width = 7 : i32, out_channels = 64 : i32, padding_height = 3 : i32, padding_width = 3 : i32, stride_height = 2 : i32, stride_width = 2 : i32}> : (tensor<1x224x224x3xf32, #layout32>, tensor<64x3x7x7xf32, #layout7>, tensor<1x112x112x64xf32, #layout33>, !tt.device<#device>) -> tensor<1x112x112x64xf32, #layout33> loc(#loc449) + %9 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x112x64x112>}> : (!tt.device<#device>) -> tensor<1x112x64x112xf32, #layout34> loc(#loc450) + %10 = "ttnn.transpose"(%8, %9) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x112x112x64xf32, #layout33>, tensor<1x112x64x112xf32, #layout34>) -> tensor<1x112x64x112xf32, #layout34> loc(#loc450) + %11 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x112x112>}> : (!tt.device<#device>) -> tensor<1x64x112x112xf32, #layout35> loc(#loc451) + %12 = "ttnn.transpose"(%10, %11) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x112x64x112xf32, #layout34>, tensor<1x64x112x112xf32, #layout35>) -> tensor<1x64x112x112xf32, #layout35> loc(#loc451) + %13 = "ttnn.to_layout"(%arg1, %0) <{layout = #ttnn.layout}> : (tensor<64x1x1xf32, #layout1>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc452) + %14 = "ttnn.to_device"(%13, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<64x1x1xf32, #layout36>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc452) + %15 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x112x112>}> : (!tt.device<#device>) -> tensor<1x64x112x112xf32, #layout35> loc(#loc452) + %16 = "ttnn.multiply"(%12, %14, %15) <{operandSegmentSizes = array}> : (tensor<1x64x112x112xf32, #layout35>, tensor<64x1x1xf32, #layout36>, tensor<1x64x112x112xf32, #layout35>) -> tensor<1x64x112x112xf32, #layout35> loc(#loc452) + %17 = "ttnn.to_layout"(%arg2, %0) <{layout = #ttnn.layout}> : (tensor<64x1x1xf32, #layout1>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc453) + %18 = "ttnn.to_device"(%17, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<64x1x1xf32, #layout36>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc453) + %19 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x112x112>}> : (!tt.device<#device>) -> tensor<1x64x112x112xf32, #layout35> loc(#loc453) + %20 = "ttnn.add"(%16, %18, %19) <{operandSegmentSizes = array}> : (tensor<1x64x112x112xf32, #layout35>, tensor<64x1x1xf32, #layout36>, tensor<1x64x112x112xf32, #layout35>) -> tensor<1x64x112x112xf32, #layout35> loc(#loc453) + %21 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x112x112>}> : (!tt.device<#device>) -> tensor<1x64x112x112xf32, #layout35> loc(#loc454) + %22 = "ttnn.relu"(%20, %21) <{operandSegmentSizes = array}> : (tensor<1x64x112x112xf32, #layout35>, tensor<1x64x112x112xf32, #layout35>) -> tensor<1x64x112x112xf32, #layout35> loc(#loc454) + %23 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1x7168x112>}> : (!tt.device<#device>) -> tensor<1x1x7168x112xf32, #layout37> loc(#loc455) + %24 = "ttnn.reshape"(%22, %23) <{shape = [1 : i32, 1 : i32, 7168 : i32, 112 : i32]}> : (tensor<1x64x112x112xf32, #layout35>, tensor<1x1x7168x112xf32, #layout37>) -> tensor<1x1x7168x112xf32, #layout37> loc(#loc455) + %25 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1x3584x56>}> : (!tt.device<#device>) -> tensor<1x1x3584x56xf32, #layout38> loc(#loc455) + %26 = "ttnn.max_pool2d"(%24, %25, %0) <{batch_size = 1 : si32, ceil_mode = false, channels = 112 : si32, dilation_height = 1 : si32, dilation_width = 1 : si32, input_height = 64 : si32, input_width = 112 : si32, kernel_height = 3 : si32, kernel_width = 3 : si32, padding_height = 1 : si32, padding_width = 1 : si32, stride_height = 2 : si32, stride_width = 2 : si32}> : (tensor<1x1x7168x112xf32, #layout37>, tensor<1x1x3584x56xf32, #layout38>, !tt.device<#device>) -> tensor<1x1x3584x56xf32, #layout38> loc(#loc455) + %27 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x56x56>}> : (!tt.device<#device>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc455) + %28 = "ttnn.reshape"(%26, %27) <{shape = [1 : i32, 64 : i32, 56 : i32, 56 : i32]}> : (tensor<1x1x3584x56xf32, #layout38>, tensor<1x64x56x56xf32, #layout39>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc455) + %29 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x64x56>}> : (!tt.device<#device>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc456) + %30 = "ttnn.transpose"(%28, %29) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x64x56x56xf32, #layout39>, tensor<1x56x64x56xf32, #layout40>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc456) + %31 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x56x64>}> : (!tt.device<#device>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc457) + %32 = "ttnn.transpose"(%30, %31) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x56x64x56xf32, #layout40>, tensor<1x56x56x64xf32, #layout41>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc457) + %33 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x56x64>}> : (!tt.device<#device>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc458) + %34 = "ttnn.conv2d"(%32, %arg108, %33, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 64 : i32, input_height = 56 : i32, input_width = 56 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 64 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x56x56x64xf32, #layout41>, tensor<64x64x1x1xf32, #layout8>, tensor<1x56x56x64xf32, #layout41>, !tt.device<#device>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc458) + %35 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x64x56>}> : (!tt.device<#device>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc459) + %36 = "ttnn.transpose"(%34, %35) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x56x56x64xf32, #layout41>, tensor<1x56x64x56xf32, #layout40>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc459) + %37 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x56x56>}> : (!tt.device<#device>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc460) + %38 = "ttnn.transpose"(%36, %37) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x56x64x56xf32, #layout40>, tensor<1x64x56x56xf32, #layout39>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc460) + %39 = "ttnn.to_layout"(%arg3, %0) <{layout = #ttnn.layout}> : (tensor<64x1x1xf32, #layout1>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc461) + %40 = "ttnn.to_device"(%39, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<64x1x1xf32, #layout36>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc461) + %41 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x56x56>}> : (!tt.device<#device>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc461) + %42 = "ttnn.multiply"(%38, %40, %41) <{operandSegmentSizes = array}> : (tensor<1x64x56x56xf32, #layout39>, tensor<64x1x1xf32, #layout36>, tensor<1x64x56x56xf32, #layout39>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc461) + %43 = "ttnn.to_layout"(%arg4, %0) <{layout = #ttnn.layout}> : (tensor<64x1x1xf32, #layout1>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc462) + %44 = "ttnn.to_device"(%43, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<64x1x1xf32, #layout36>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc462) + %45 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x56x56>}> : (!tt.device<#device>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc462) + %46 = "ttnn.add"(%42, %44, %45) <{operandSegmentSizes = array}> : (tensor<1x64x56x56xf32, #layout39>, tensor<64x1x1xf32, #layout36>, tensor<1x64x56x56xf32, #layout39>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc462) + %47 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x56x56>}> : (!tt.device<#device>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc463) + %48 = "ttnn.relu"(%46, %47) <{operandSegmentSizes = array}> : (tensor<1x64x56x56xf32, #layout39>, tensor<1x64x56x56xf32, #layout39>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc463) + %49 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x64x56>}> : (!tt.device<#device>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc464) + %50 = "ttnn.transpose"(%48, %49) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x64x56x56xf32, #layout39>, tensor<1x56x64x56xf32, #layout40>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc464) + %51 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x56x64>}> : (!tt.device<#device>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc465) + %52 = "ttnn.transpose"(%50, %51) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x56x64x56xf32, #layout40>, tensor<1x56x56x64xf32, #layout41>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc465) + %53 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x56x64>}> : (!tt.device<#device>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc466) + %54 = "ttnn.conv2d"(%52, %arg109, %53, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 64 : i32, input_height = 56 : i32, input_width = 56 : i32, kernel_height = 3 : i32, kernel_width = 3 : i32, out_channels = 64 : i32, padding_height = 1 : i32, padding_width = 1 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x56x56x64xf32, #layout41>, tensor<64x64x3x3xf32, #layout9>, tensor<1x56x56x64xf32, #layout41>, !tt.device<#device>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc466) + %55 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x64x56>}> : (!tt.device<#device>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc467) + %56 = "ttnn.transpose"(%54, %55) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x56x56x64xf32, #layout41>, tensor<1x56x64x56xf32, #layout40>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc467) + %57 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x56x56>}> : (!tt.device<#device>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc468) + %58 = "ttnn.transpose"(%56, %57) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x56x64x56xf32, #layout40>, tensor<1x64x56x56xf32, #layout39>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc468) + %59 = "ttnn.to_layout"(%arg5, %0) <{layout = #ttnn.layout}> : (tensor<64x1x1xf32, #layout1>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc469) + %60 = "ttnn.to_device"(%59, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<64x1x1xf32, #layout36>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc469) + %61 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x56x56>}> : (!tt.device<#device>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc469) + %62 = "ttnn.multiply"(%58, %60, %61) <{operandSegmentSizes = array}> : (tensor<1x64x56x56xf32, #layout39>, tensor<64x1x1xf32, #layout36>, tensor<1x64x56x56xf32, #layout39>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc469) + %63 = "ttnn.to_layout"(%arg6, %0) <{layout = #ttnn.layout}> : (tensor<64x1x1xf32, #layout1>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc470) + %64 = "ttnn.to_device"(%63, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<64x1x1xf32, #layout36>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc470) + %65 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x56x56>}> : (!tt.device<#device>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc470) + %66 = "ttnn.add"(%62, %64, %65) <{operandSegmentSizes = array}> : (tensor<1x64x56x56xf32, #layout39>, tensor<64x1x1xf32, #layout36>, tensor<1x64x56x56xf32, #layout39>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc470) + %67 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x56x56>}> : (!tt.device<#device>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc471) + %68 = "ttnn.relu"(%66, %67) <{operandSegmentSizes = array}> : (tensor<1x64x56x56xf32, #layout39>, tensor<1x64x56x56xf32, #layout39>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc471) + %69 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x64x56>}> : (!tt.device<#device>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc472) + %70 = "ttnn.transpose"(%68, %69) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x64x56x56xf32, #layout39>, tensor<1x56x64x56xf32, #layout40>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc472) + %71 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x56x64>}> : (!tt.device<#device>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc473) + %72 = "ttnn.transpose"(%70, %71) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x56x64x56xf32, #layout40>, tensor<1x56x56x64xf32, #layout41>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc473) + %73 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x56x256>}> : (!tt.device<#device>) -> tensor<1x56x56x256xf32, #layout42> loc(#loc474) + %74 = "ttnn.conv2d"(%72, %arg110, %73, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 64 : i32, input_height = 56 : i32, input_width = 56 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 256 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x56x56x64xf32, #layout41>, tensor<256x64x1x1xf32, #layout10>, tensor<1x56x56x256xf32, #layout42>, !tt.device<#device>) -> tensor<1x56x56x256xf32, #layout42> loc(#loc474) + %75 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x256x56>}> : (!tt.device<#device>) -> tensor<1x56x256x56xf32, #layout43> loc(#loc475) + %76 = "ttnn.transpose"(%74, %75) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x56x56x256xf32, #layout42>, tensor<1x56x256x56xf32, #layout43>) -> tensor<1x56x256x56xf32, #layout43> loc(#loc475) + %77 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x56x56>}> : (!tt.device<#device>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc476) + %78 = "ttnn.transpose"(%76, %77) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x56x256x56xf32, #layout43>, tensor<1x256x56x56xf32, #layout44>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc476) + %79 = "ttnn.to_layout"(%arg7, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc477) + %80 = "ttnn.to_device"(%79, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc477) + %81 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x56x56>}> : (!tt.device<#device>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc477) + %82 = "ttnn.multiply"(%78, %80, %81) <{operandSegmentSizes = array}> : (tensor<1x256x56x56xf32, #layout44>, tensor<256x1x1xf32, #layout45>, tensor<1x256x56x56xf32, #layout44>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc477) + %83 = "ttnn.to_layout"(%arg8, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc478) + %84 = "ttnn.to_device"(%83, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc478) + %85 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x56x56>}> : (!tt.device<#device>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc478) + %86 = "ttnn.add"(%82, %84, %85) <{operandSegmentSizes = array}> : (tensor<1x256x56x56xf32, #layout44>, tensor<256x1x1xf32, #layout45>, tensor<1x256x56x56xf32, #layout44>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc478) + %87 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x64x56>}> : (!tt.device<#device>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc479) + %88 = "ttnn.transpose"(%28, %87) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x64x56x56xf32, #layout39>, tensor<1x56x64x56xf32, #layout40>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc479) + %89 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x56x64>}> : (!tt.device<#device>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc480) + %90 = "ttnn.transpose"(%88, %89) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x56x64x56xf32, #layout40>, tensor<1x56x56x64xf32, #layout41>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc480) + %91 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x56x256>}> : (!tt.device<#device>) -> tensor<1x56x56x256xf32, #layout42> loc(#loc481) + %92 = "ttnn.conv2d"(%90, %arg111, %91, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 64 : i32, input_height = 56 : i32, input_width = 56 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 256 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x56x56x64xf32, #layout41>, tensor<256x64x1x1xf32, #layout10>, tensor<1x56x56x256xf32, #layout42>, !tt.device<#device>) -> tensor<1x56x56x256xf32, #layout42> loc(#loc481) + %93 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x256x56>}> : (!tt.device<#device>) -> tensor<1x56x256x56xf32, #layout43> loc(#loc482) + %94 = "ttnn.transpose"(%92, %93) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x56x56x256xf32, #layout42>, tensor<1x56x256x56xf32, #layout43>) -> tensor<1x56x256x56xf32, #layout43> loc(#loc482) + %95 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x56x56>}> : (!tt.device<#device>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc483) + %96 = "ttnn.transpose"(%94, %95) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x56x256x56xf32, #layout43>, tensor<1x256x56x56xf32, #layout44>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc483) + %97 = "ttnn.to_layout"(%arg9, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc484) + %98 = "ttnn.to_device"(%97, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc484) + %99 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x56x56>}> : (!tt.device<#device>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc484) + %100 = "ttnn.multiply"(%96, %98, %99) <{operandSegmentSizes = array}> : (tensor<1x256x56x56xf32, #layout44>, tensor<256x1x1xf32, #layout45>, tensor<1x256x56x56xf32, #layout44>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc484) + %101 = "ttnn.to_layout"(%arg10, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc485) + %102 = "ttnn.to_device"(%101, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc485) + %103 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x56x56>}> : (!tt.device<#device>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc485) + %104 = "ttnn.add"(%100, %102, %103) <{operandSegmentSizes = array}> : (tensor<1x256x56x56xf32, #layout44>, tensor<256x1x1xf32, #layout45>, tensor<1x256x56x56xf32, #layout44>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc485) + %105 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x56x56>}> : (!tt.device<#device>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc486) + %106 = "ttnn.add"(%86, %104, %105) <{operandSegmentSizes = array}> : (tensor<1x256x56x56xf32, #layout44>, tensor<1x256x56x56xf32, #layout44>, tensor<1x256x56x56xf32, #layout44>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc486) + %107 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x56x56>}> : (!tt.device<#device>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc487) + %108 = "ttnn.relu"(%106, %107) <{operandSegmentSizes = array}> : (tensor<1x256x56x56xf32, #layout44>, tensor<1x256x56x56xf32, #layout44>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc487) + %109 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x256x56>}> : (!tt.device<#device>) -> tensor<1x56x256x56xf32, #layout43> loc(#loc488) + %110 = "ttnn.transpose"(%108, %109) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x256x56x56xf32, #layout44>, tensor<1x56x256x56xf32, #layout43>) -> tensor<1x56x256x56xf32, #layout43> loc(#loc488) + %111 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x56x256>}> : (!tt.device<#device>) -> tensor<1x56x56x256xf32, #layout42> loc(#loc489) + %112 = "ttnn.transpose"(%110, %111) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x56x256x56xf32, #layout43>, tensor<1x56x56x256xf32, #layout42>) -> tensor<1x56x56x256xf32, #layout42> loc(#loc489) + %113 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x56x64>}> : (!tt.device<#device>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc490) + %114 = "ttnn.conv2d"(%112, %arg112, %113, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 256 : i32, input_height = 56 : i32, input_width = 56 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 64 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x56x56x256xf32, #layout42>, tensor<64x256x1x1xf32, #layout11>, tensor<1x56x56x64xf32, #layout41>, !tt.device<#device>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc490) + %115 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x64x56>}> : (!tt.device<#device>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc491) + %116 = "ttnn.transpose"(%114, %115) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x56x56x64xf32, #layout41>, tensor<1x56x64x56xf32, #layout40>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc491) + %117 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x56x56>}> : (!tt.device<#device>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc492) + %118 = "ttnn.transpose"(%116, %117) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x56x64x56xf32, #layout40>, tensor<1x64x56x56xf32, #layout39>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc492) + %119 = "ttnn.to_layout"(%arg11, %0) <{layout = #ttnn.layout}> : (tensor<64x1x1xf32, #layout1>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc493) + %120 = "ttnn.to_device"(%119, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<64x1x1xf32, #layout36>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc493) + %121 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x56x56>}> : (!tt.device<#device>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc493) + %122 = "ttnn.multiply"(%118, %120, %121) <{operandSegmentSizes = array}> : (tensor<1x64x56x56xf32, #layout39>, tensor<64x1x1xf32, #layout36>, tensor<1x64x56x56xf32, #layout39>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc493) + %123 = "ttnn.to_layout"(%arg12, %0) <{layout = #ttnn.layout}> : (tensor<64x1x1xf32, #layout1>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc494) + %124 = "ttnn.to_device"(%123, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<64x1x1xf32, #layout36>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc494) + %125 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x56x56>}> : (!tt.device<#device>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc494) + %126 = "ttnn.add"(%122, %124, %125) <{operandSegmentSizes = array}> : (tensor<1x64x56x56xf32, #layout39>, tensor<64x1x1xf32, #layout36>, tensor<1x64x56x56xf32, #layout39>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc494) + %127 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x56x56>}> : (!tt.device<#device>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc495) + %128 = "ttnn.relu"(%126, %127) <{operandSegmentSizes = array}> : (tensor<1x64x56x56xf32, #layout39>, tensor<1x64x56x56xf32, #layout39>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc495) + %129 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x64x56>}> : (!tt.device<#device>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc496) + %130 = "ttnn.transpose"(%128, %129) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x64x56x56xf32, #layout39>, tensor<1x56x64x56xf32, #layout40>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc496) + %131 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x56x64>}> : (!tt.device<#device>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc497) + %132 = "ttnn.transpose"(%130, %131) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x56x64x56xf32, #layout40>, tensor<1x56x56x64xf32, #layout41>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc497) + %133 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x56x64>}> : (!tt.device<#device>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc498) + %134 = "ttnn.conv2d"(%132, %arg113, %133, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 64 : i32, input_height = 56 : i32, input_width = 56 : i32, kernel_height = 3 : i32, kernel_width = 3 : i32, out_channels = 64 : i32, padding_height = 1 : i32, padding_width = 1 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x56x56x64xf32, #layout41>, tensor<64x64x3x3xf32, #layout9>, tensor<1x56x56x64xf32, #layout41>, !tt.device<#device>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc498) + %135 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x64x56>}> : (!tt.device<#device>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc499) + %136 = "ttnn.transpose"(%134, %135) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x56x56x64xf32, #layout41>, tensor<1x56x64x56xf32, #layout40>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc499) + %137 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x56x56>}> : (!tt.device<#device>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc500) + %138 = "ttnn.transpose"(%136, %137) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x56x64x56xf32, #layout40>, tensor<1x64x56x56xf32, #layout39>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc500) + %139 = "ttnn.to_layout"(%arg13, %0) <{layout = #ttnn.layout}> : (tensor<64x1x1xf32, #layout1>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc501) + %140 = "ttnn.to_device"(%139, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<64x1x1xf32, #layout36>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc501) + %141 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x56x56>}> : (!tt.device<#device>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc501) + %142 = "ttnn.multiply"(%138, %140, %141) <{operandSegmentSizes = array}> : (tensor<1x64x56x56xf32, #layout39>, tensor<64x1x1xf32, #layout36>, tensor<1x64x56x56xf32, #layout39>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc501) + %143 = "ttnn.to_layout"(%arg14, %0) <{layout = #ttnn.layout}> : (tensor<64x1x1xf32, #layout1>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc502) + %144 = "ttnn.to_device"(%143, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<64x1x1xf32, #layout36>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc502) + %145 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x56x56>}> : (!tt.device<#device>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc502) + %146 = "ttnn.add"(%142, %144, %145) <{operandSegmentSizes = array}> : (tensor<1x64x56x56xf32, #layout39>, tensor<64x1x1xf32, #layout36>, tensor<1x64x56x56xf32, #layout39>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc502) + %147 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x56x56>}> : (!tt.device<#device>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc503) + %148 = "ttnn.relu"(%146, %147) <{operandSegmentSizes = array}> : (tensor<1x64x56x56xf32, #layout39>, tensor<1x64x56x56xf32, #layout39>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc503) + %149 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x64x56>}> : (!tt.device<#device>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc504) + %150 = "ttnn.transpose"(%148, %149) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x64x56x56xf32, #layout39>, tensor<1x56x64x56xf32, #layout40>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc504) + %151 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x56x64>}> : (!tt.device<#device>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc505) + %152 = "ttnn.transpose"(%150, %151) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x56x64x56xf32, #layout40>, tensor<1x56x56x64xf32, #layout41>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc505) + %153 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x56x256>}> : (!tt.device<#device>) -> tensor<1x56x56x256xf32, #layout42> loc(#loc506) + %154 = "ttnn.conv2d"(%152, %arg114, %153, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 64 : i32, input_height = 56 : i32, input_width = 56 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 256 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x56x56x64xf32, #layout41>, tensor<256x64x1x1xf32, #layout10>, tensor<1x56x56x256xf32, #layout42>, !tt.device<#device>) -> tensor<1x56x56x256xf32, #layout42> loc(#loc506) + %155 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x256x56>}> : (!tt.device<#device>) -> tensor<1x56x256x56xf32, #layout43> loc(#loc507) + %156 = "ttnn.transpose"(%154, %155) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x56x56x256xf32, #layout42>, tensor<1x56x256x56xf32, #layout43>) -> tensor<1x56x256x56xf32, #layout43> loc(#loc507) + %157 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x56x56>}> : (!tt.device<#device>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc508) + %158 = "ttnn.transpose"(%156, %157) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x56x256x56xf32, #layout43>, tensor<1x256x56x56xf32, #layout44>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc508) + %159 = "ttnn.to_layout"(%arg15, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc509) + %160 = "ttnn.to_device"(%159, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc509) + %161 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x56x56>}> : (!tt.device<#device>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc509) + %162 = "ttnn.multiply"(%158, %160, %161) <{operandSegmentSizes = array}> : (tensor<1x256x56x56xf32, #layout44>, tensor<256x1x1xf32, #layout45>, tensor<1x256x56x56xf32, #layout44>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc509) + %163 = "ttnn.to_layout"(%arg16, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc510) + %164 = "ttnn.to_device"(%163, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc510) + %165 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x56x56>}> : (!tt.device<#device>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc510) + %166 = "ttnn.add"(%162, %164, %165) <{operandSegmentSizes = array}> : (tensor<1x256x56x56xf32, #layout44>, tensor<256x1x1xf32, #layout45>, tensor<1x256x56x56xf32, #layout44>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc510) + %167 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x56x56>}> : (!tt.device<#device>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc511) + %168 = "ttnn.add"(%166, %108, %167) <{operandSegmentSizes = array}> : (tensor<1x256x56x56xf32, #layout44>, tensor<1x256x56x56xf32, #layout44>, tensor<1x256x56x56xf32, #layout44>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc511) + %169 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x56x56>}> : (!tt.device<#device>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc512) + %170 = "ttnn.relu"(%168, %169) <{operandSegmentSizes = array}> : (tensor<1x256x56x56xf32, #layout44>, tensor<1x256x56x56xf32, #layout44>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc512) + %171 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x256x56>}> : (!tt.device<#device>) -> tensor<1x56x256x56xf32, #layout43> loc(#loc513) + %172 = "ttnn.transpose"(%170, %171) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x256x56x56xf32, #layout44>, tensor<1x56x256x56xf32, #layout43>) -> tensor<1x56x256x56xf32, #layout43> loc(#loc513) + %173 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x56x256>}> : (!tt.device<#device>) -> tensor<1x56x56x256xf32, #layout42> loc(#loc514) + %174 = "ttnn.transpose"(%172, %173) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x56x256x56xf32, #layout43>, tensor<1x56x56x256xf32, #layout42>) -> tensor<1x56x56x256xf32, #layout42> loc(#loc514) + %175 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x56x64>}> : (!tt.device<#device>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc515) + %176 = "ttnn.conv2d"(%174, %arg115, %175, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 256 : i32, input_height = 56 : i32, input_width = 56 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 64 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x56x56x256xf32, #layout42>, tensor<64x256x1x1xf32, #layout11>, tensor<1x56x56x64xf32, #layout41>, !tt.device<#device>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc515) + %177 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x64x56>}> : (!tt.device<#device>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc516) + %178 = "ttnn.transpose"(%176, %177) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x56x56x64xf32, #layout41>, tensor<1x56x64x56xf32, #layout40>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc516) + %179 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x56x56>}> : (!tt.device<#device>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc517) + %180 = "ttnn.transpose"(%178, %179) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x56x64x56xf32, #layout40>, tensor<1x64x56x56xf32, #layout39>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc517) + %181 = "ttnn.to_layout"(%arg17, %0) <{layout = #ttnn.layout}> : (tensor<64x1x1xf32, #layout1>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc518) + %182 = "ttnn.to_device"(%181, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<64x1x1xf32, #layout36>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc518) + %183 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x56x56>}> : (!tt.device<#device>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc518) + %184 = "ttnn.multiply"(%180, %182, %183) <{operandSegmentSizes = array}> : (tensor<1x64x56x56xf32, #layout39>, tensor<64x1x1xf32, #layout36>, tensor<1x64x56x56xf32, #layout39>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc518) + %185 = "ttnn.to_layout"(%arg18, %0) <{layout = #ttnn.layout}> : (tensor<64x1x1xf32, #layout1>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc519) + %186 = "ttnn.to_device"(%185, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<64x1x1xf32, #layout36>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc519) + %187 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x56x56>}> : (!tt.device<#device>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc519) + %188 = "ttnn.add"(%184, %186, %187) <{operandSegmentSizes = array}> : (tensor<1x64x56x56xf32, #layout39>, tensor<64x1x1xf32, #layout36>, tensor<1x64x56x56xf32, #layout39>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc519) + %189 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x56x56>}> : (!tt.device<#device>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc520) + %190 = "ttnn.relu"(%188, %189) <{operandSegmentSizes = array}> : (tensor<1x64x56x56xf32, #layout39>, tensor<1x64x56x56xf32, #layout39>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc520) + %191 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x64x56>}> : (!tt.device<#device>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc521) + %192 = "ttnn.transpose"(%190, %191) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x64x56x56xf32, #layout39>, tensor<1x56x64x56xf32, #layout40>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc521) + %193 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x56x64>}> : (!tt.device<#device>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc522) + %194 = "ttnn.transpose"(%192, %193) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x56x64x56xf32, #layout40>, tensor<1x56x56x64xf32, #layout41>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc522) + %195 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x56x64>}> : (!tt.device<#device>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc523) + %196 = "ttnn.conv2d"(%194, %arg116, %195, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 64 : i32, input_height = 56 : i32, input_width = 56 : i32, kernel_height = 3 : i32, kernel_width = 3 : i32, out_channels = 64 : i32, padding_height = 1 : i32, padding_width = 1 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x56x56x64xf32, #layout41>, tensor<64x64x3x3xf32, #layout9>, tensor<1x56x56x64xf32, #layout41>, !tt.device<#device>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc523) + %197 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x64x56>}> : (!tt.device<#device>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc524) + %198 = "ttnn.transpose"(%196, %197) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x56x56x64xf32, #layout41>, tensor<1x56x64x56xf32, #layout40>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc524) + %199 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x56x56>}> : (!tt.device<#device>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc525) + %200 = "ttnn.transpose"(%198, %199) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x56x64x56xf32, #layout40>, tensor<1x64x56x56xf32, #layout39>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc525) + %201 = "ttnn.to_layout"(%arg19, %0) <{layout = #ttnn.layout}> : (tensor<64x1x1xf32, #layout1>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc526) + %202 = "ttnn.to_device"(%201, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<64x1x1xf32, #layout36>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc526) + %203 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x56x56>}> : (!tt.device<#device>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc526) + %204 = "ttnn.multiply"(%200, %202, %203) <{operandSegmentSizes = array}> : (tensor<1x64x56x56xf32, #layout39>, tensor<64x1x1xf32, #layout36>, tensor<1x64x56x56xf32, #layout39>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc526) + %205 = "ttnn.to_layout"(%arg20, %0) <{layout = #ttnn.layout}> : (tensor<64x1x1xf32, #layout1>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc527) + %206 = "ttnn.to_device"(%205, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<64x1x1xf32, #layout36>, !tt.device<#device>) -> tensor<64x1x1xf32, #layout36> loc(#loc527) + %207 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x56x56>}> : (!tt.device<#device>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc527) + %208 = "ttnn.add"(%204, %206, %207) <{operandSegmentSizes = array}> : (tensor<1x64x56x56xf32, #layout39>, tensor<64x1x1xf32, #layout36>, tensor<1x64x56x56xf32, #layout39>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc527) + %209 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x64x56x56>}> : (!tt.device<#device>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc528) + %210 = "ttnn.relu"(%208, %209) <{operandSegmentSizes = array}> : (tensor<1x64x56x56xf32, #layout39>, tensor<1x64x56x56xf32, #layout39>) -> tensor<1x64x56x56xf32, #layout39> loc(#loc528) + %211 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x64x56>}> : (!tt.device<#device>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc529) + %212 = "ttnn.transpose"(%210, %211) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x64x56x56xf32, #layout39>, tensor<1x56x64x56xf32, #layout40>) -> tensor<1x56x64x56xf32, #layout40> loc(#loc529) + %213 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x56x64>}> : (!tt.device<#device>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc530) + %214 = "ttnn.transpose"(%212, %213) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x56x64x56xf32, #layout40>, tensor<1x56x56x64xf32, #layout41>) -> tensor<1x56x56x64xf32, #layout41> loc(#loc530) + %215 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x56x256>}> : (!tt.device<#device>) -> tensor<1x56x56x256xf32, #layout42> loc(#loc531) + %216 = "ttnn.conv2d"(%214, %arg117, %215, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 64 : i32, input_height = 56 : i32, input_width = 56 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 256 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x56x56x64xf32, #layout41>, tensor<256x64x1x1xf32, #layout10>, tensor<1x56x56x256xf32, #layout42>, !tt.device<#device>) -> tensor<1x56x56x256xf32, #layout42> loc(#loc531) + %217 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x256x56>}> : (!tt.device<#device>) -> tensor<1x56x256x56xf32, #layout43> loc(#loc532) + %218 = "ttnn.transpose"(%216, %217) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x56x56x256xf32, #layout42>, tensor<1x56x256x56xf32, #layout43>) -> tensor<1x56x256x56xf32, #layout43> loc(#loc532) + %219 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x56x56>}> : (!tt.device<#device>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc533) + %220 = "ttnn.transpose"(%218, %219) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x56x256x56xf32, #layout43>, tensor<1x256x56x56xf32, #layout44>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc533) + %221 = "ttnn.to_layout"(%arg21, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc534) + %222 = "ttnn.to_device"(%221, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc534) + %223 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x56x56>}> : (!tt.device<#device>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc534) + %224 = "ttnn.multiply"(%220, %222, %223) <{operandSegmentSizes = array}> : (tensor<1x256x56x56xf32, #layout44>, tensor<256x1x1xf32, #layout45>, tensor<1x256x56x56xf32, #layout44>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc534) + %225 = "ttnn.to_layout"(%arg22, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc535) + %226 = "ttnn.to_device"(%225, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc535) + %227 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x56x56>}> : (!tt.device<#device>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc535) + %228 = "ttnn.add"(%224, %226, %227) <{operandSegmentSizes = array}> : (tensor<1x256x56x56xf32, #layout44>, tensor<256x1x1xf32, #layout45>, tensor<1x256x56x56xf32, #layout44>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc535) + %229 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x56x56>}> : (!tt.device<#device>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc536) + %230 = "ttnn.add"(%228, %170, %229) <{operandSegmentSizes = array}> : (tensor<1x256x56x56xf32, #layout44>, tensor<1x256x56x56xf32, #layout44>, tensor<1x256x56x56xf32, #layout44>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc536) + %231 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x56x56>}> : (!tt.device<#device>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc537) + %232 = "ttnn.relu"(%230, %231) <{operandSegmentSizes = array}> : (tensor<1x256x56x56xf32, #layout44>, tensor<1x256x56x56xf32, #layout44>) -> tensor<1x256x56x56xf32, #layout44> loc(#loc537) + %233 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x256x56>}> : (!tt.device<#device>) -> tensor<1x56x256x56xf32, #layout43> loc(#loc538) + %234 = "ttnn.transpose"(%232, %233) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x256x56x56xf32, #layout44>, tensor<1x56x256x56xf32, #layout43>) -> tensor<1x56x256x56xf32, #layout43> loc(#loc538) + %235 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x56x256>}> : (!tt.device<#device>) -> tensor<1x56x56x256xf32, #layout42> loc(#loc539) + %236 = "ttnn.transpose"(%234, %235) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x56x256x56xf32, #layout43>, tensor<1x56x56x256xf32, #layout42>) -> tensor<1x56x56x256xf32, #layout42> loc(#loc539) + %237 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x56x128>}> : (!tt.device<#device>) -> tensor<1x56x56x128xf32, #layout46> loc(#loc540) + %238 = "ttnn.conv2d"(%236, %arg118, %237, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 256 : i32, input_height = 56 : i32, input_width = 56 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 128 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x56x56x256xf32, #layout42>, tensor<128x256x1x1xf32, #layout12>, tensor<1x56x56x128xf32, #layout46>, !tt.device<#device>) -> tensor<1x56x56x128xf32, #layout46> loc(#loc540) + %239 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x128x56>}> : (!tt.device<#device>) -> tensor<1x56x128x56xf32, #layout47> loc(#loc541) + %240 = "ttnn.transpose"(%238, %239) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x56x56x128xf32, #layout46>, tensor<1x56x128x56xf32, #layout47>) -> tensor<1x56x128x56xf32, #layout47> loc(#loc541) + %241 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x56x56>}> : (!tt.device<#device>) -> tensor<1x128x56x56xf32, #layout48> loc(#loc542) + %242 = "ttnn.transpose"(%240, %241) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x56x128x56xf32, #layout47>, tensor<1x128x56x56xf32, #layout48>) -> tensor<1x128x56x56xf32, #layout48> loc(#loc542) + %243 = "ttnn.to_layout"(%arg23, %0) <{layout = #ttnn.layout}> : (tensor<128x1x1xf32, #layout3>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc543) + %244 = "ttnn.to_device"(%243, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<128x1x1xf32, #layout49>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc543) + %245 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x56x56>}> : (!tt.device<#device>) -> tensor<1x128x56x56xf32, #layout48> loc(#loc543) + %246 = "ttnn.multiply"(%242, %244, %245) <{operandSegmentSizes = array}> : (tensor<1x128x56x56xf32, #layout48>, tensor<128x1x1xf32, #layout49>, tensor<1x128x56x56xf32, #layout48>) -> tensor<1x128x56x56xf32, #layout48> loc(#loc543) + %247 = "ttnn.to_layout"(%arg24, %0) <{layout = #ttnn.layout}> : (tensor<128x1x1xf32, #layout3>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc544) + %248 = "ttnn.to_device"(%247, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<128x1x1xf32, #layout49>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc544) + %249 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x56x56>}> : (!tt.device<#device>) -> tensor<1x128x56x56xf32, #layout48> loc(#loc544) + %250 = "ttnn.add"(%246, %248, %249) <{operandSegmentSizes = array}> : (tensor<1x128x56x56xf32, #layout48>, tensor<128x1x1xf32, #layout49>, tensor<1x128x56x56xf32, #layout48>) -> tensor<1x128x56x56xf32, #layout48> loc(#loc544) + %251 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x56x56>}> : (!tt.device<#device>) -> tensor<1x128x56x56xf32, #layout48> loc(#loc545) + %252 = "ttnn.relu"(%250, %251) <{operandSegmentSizes = array}> : (tensor<1x128x56x56xf32, #layout48>, tensor<1x128x56x56xf32, #layout48>) -> tensor<1x128x56x56xf32, #layout48> loc(#loc545) + %253 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x128x56>}> : (!tt.device<#device>) -> tensor<1x56x128x56xf32, #layout47> loc(#loc546) + %254 = "ttnn.transpose"(%252, %253) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x128x56x56xf32, #layout48>, tensor<1x56x128x56xf32, #layout47>) -> tensor<1x56x128x56xf32, #layout47> loc(#loc546) + %255 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x56x128>}> : (!tt.device<#device>) -> tensor<1x56x56x128xf32, #layout46> loc(#loc547) + %256 = "ttnn.transpose"(%254, %255) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x56x128x56xf32, #layout47>, tensor<1x56x56x128xf32, #layout46>) -> tensor<1x56x56x128xf32, #layout46> loc(#loc547) + %257 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x128>}> : (!tt.device<#device>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc548) + %258 = "ttnn.conv2d"(%256, %arg119, %257, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 128 : i32, input_height = 56 : i32, input_width = 56 : i32, kernel_height = 3 : i32, kernel_width = 3 : i32, out_channels = 128 : i32, padding_height = 1 : i32, padding_width = 1 : i32, stride_height = 2 : i32, stride_width = 2 : i32}> : (tensor<1x56x56x128xf32, #layout46>, tensor<128x128x3x3xf32, #layout13>, tensor<1x28x28x128xf32, #layout50>, !tt.device<#device>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc548) + %259 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x128x28>}> : (!tt.device<#device>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc549) + %260 = "ttnn.transpose"(%258, %259) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x28x128xf32, #layout50>, tensor<1x28x128x28xf32, #layout51>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc549) + %261 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc550) + %262 = "ttnn.transpose"(%260, %261) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x28x128x28xf32, #layout51>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc550) + %263 = "ttnn.to_layout"(%arg25, %0) <{layout = #ttnn.layout}> : (tensor<128x1x1xf32, #layout3>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc551) + %264 = "ttnn.to_device"(%263, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<128x1x1xf32, #layout49>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc551) + %265 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc551) + %266 = "ttnn.multiply"(%262, %264, %265) <{operandSegmentSizes = array}> : (tensor<1x128x28x28xf32, #layout52>, tensor<128x1x1xf32, #layout49>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc551) + %267 = "ttnn.to_layout"(%arg26, %0) <{layout = #ttnn.layout}> : (tensor<128x1x1xf32, #layout3>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc552) + %268 = "ttnn.to_device"(%267, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<128x1x1xf32, #layout49>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc552) + %269 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc552) + %270 = "ttnn.add"(%266, %268, %269) <{operandSegmentSizes = array}> : (tensor<1x128x28x28xf32, #layout52>, tensor<128x1x1xf32, #layout49>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc552) + %271 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc553) + %272 = "ttnn.relu"(%270, %271) <{operandSegmentSizes = array}> : (tensor<1x128x28x28xf32, #layout52>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc553) + %273 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x128x28>}> : (!tt.device<#device>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc554) + %274 = "ttnn.transpose"(%272, %273) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x128x28x28xf32, #layout52>, tensor<1x28x128x28xf32, #layout51>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc554) + %275 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x128>}> : (!tt.device<#device>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc555) + %276 = "ttnn.transpose"(%274, %275) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x128x28xf32, #layout51>, tensor<1x28x28x128xf32, #layout50>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc555) + %277 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x512>}> : (!tt.device<#device>) -> tensor<1x28x28x512xf32, #layout53> loc(#loc556) + %278 = "ttnn.conv2d"(%276, %arg120, %277, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 128 : i32, input_height = 28 : i32, input_width = 28 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 512 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x28x28x128xf32, #layout50>, tensor<512x128x1x1xf32, #layout14>, tensor<1x28x28x512xf32, #layout53>, !tt.device<#device>) -> tensor<1x28x28x512xf32, #layout53> loc(#loc556) + %279 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x512x28>}> : (!tt.device<#device>) -> tensor<1x28x512x28xf32, #layout54> loc(#loc557) + %280 = "ttnn.transpose"(%278, %279) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x28x512xf32, #layout53>, tensor<1x28x512x28xf32, #layout54>) -> tensor<1x28x512x28xf32, #layout54> loc(#loc557) + %281 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x28x28>}> : (!tt.device<#device>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc558) + %282 = "ttnn.transpose"(%280, %281) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x28x512x28xf32, #layout54>, tensor<1x512x28x28xf32, #layout55>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc558) + %283 = "ttnn.to_layout"(%arg27, %0) <{layout = #ttnn.layout}> : (tensor<512x1x1xf32, #layout4>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc559) + %284 = "ttnn.to_device"(%283, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<512x1x1xf32, #layout56>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc559) + %285 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x28x28>}> : (!tt.device<#device>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc559) + %286 = "ttnn.multiply"(%282, %284, %285) <{operandSegmentSizes = array}> : (tensor<1x512x28x28xf32, #layout55>, tensor<512x1x1xf32, #layout56>, tensor<1x512x28x28xf32, #layout55>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc559) + %287 = "ttnn.to_layout"(%arg28, %0) <{layout = #ttnn.layout}> : (tensor<512x1x1xf32, #layout4>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc560) + %288 = "ttnn.to_device"(%287, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<512x1x1xf32, #layout56>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc560) + %289 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x28x28>}> : (!tt.device<#device>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc560) + %290 = "ttnn.add"(%286, %288, %289) <{operandSegmentSizes = array}> : (tensor<1x512x28x28xf32, #layout55>, tensor<512x1x1xf32, #layout56>, tensor<1x512x28x28xf32, #layout55>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc560) + %291 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x256x56>}> : (!tt.device<#device>) -> tensor<1x56x256x56xf32, #layout43> loc(#loc561) + %292 = "ttnn.transpose"(%232, %291) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x256x56x56xf32, #layout44>, tensor<1x56x256x56xf32, #layout43>) -> tensor<1x56x256x56xf32, #layout43> loc(#loc561) + %293 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x56x56x256>}> : (!tt.device<#device>) -> tensor<1x56x56x256xf32, #layout42> loc(#loc562) + %294 = "ttnn.transpose"(%292, %293) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x56x256x56xf32, #layout43>, tensor<1x56x56x256xf32, #layout42>) -> tensor<1x56x56x256xf32, #layout42> loc(#loc562) + %295 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x512>}> : (!tt.device<#device>) -> tensor<1x28x28x512xf32, #layout53> loc(#loc563) + %296 = "ttnn.conv2d"(%294, %arg121, %295, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 256 : i32, input_height = 56 : i32, input_width = 56 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 512 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 2 : i32, stride_width = 2 : i32}> : (tensor<1x56x56x256xf32, #layout42>, tensor<512x256x1x1xf32, #layout15>, tensor<1x28x28x512xf32, #layout53>, !tt.device<#device>) -> tensor<1x28x28x512xf32, #layout53> loc(#loc563) + %297 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x512x28>}> : (!tt.device<#device>) -> tensor<1x28x512x28xf32, #layout54> loc(#loc564) + %298 = "ttnn.transpose"(%296, %297) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x28x512xf32, #layout53>, tensor<1x28x512x28xf32, #layout54>) -> tensor<1x28x512x28xf32, #layout54> loc(#loc564) + %299 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x28x28>}> : (!tt.device<#device>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc565) + %300 = "ttnn.transpose"(%298, %299) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x28x512x28xf32, #layout54>, tensor<1x512x28x28xf32, #layout55>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc565) + %301 = "ttnn.to_layout"(%arg29, %0) <{layout = #ttnn.layout}> : (tensor<512x1x1xf32, #layout4>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc566) + %302 = "ttnn.to_device"(%301, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<512x1x1xf32, #layout56>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc566) + %303 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x28x28>}> : (!tt.device<#device>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc566) + %304 = "ttnn.multiply"(%300, %302, %303) <{operandSegmentSizes = array}> : (tensor<1x512x28x28xf32, #layout55>, tensor<512x1x1xf32, #layout56>, tensor<1x512x28x28xf32, #layout55>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc566) + %305 = "ttnn.to_layout"(%arg30, %0) <{layout = #ttnn.layout}> : (tensor<512x1x1xf32, #layout4>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc567) + %306 = "ttnn.to_device"(%305, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<512x1x1xf32, #layout56>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc567) + %307 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x28x28>}> : (!tt.device<#device>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc567) + %308 = "ttnn.add"(%304, %306, %307) <{operandSegmentSizes = array}> : (tensor<1x512x28x28xf32, #layout55>, tensor<512x1x1xf32, #layout56>, tensor<1x512x28x28xf32, #layout55>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc567) + %309 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x28x28>}> : (!tt.device<#device>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc568) + %310 = "ttnn.add"(%290, %308, %309) <{operandSegmentSizes = array}> : (tensor<1x512x28x28xf32, #layout55>, tensor<1x512x28x28xf32, #layout55>, tensor<1x512x28x28xf32, #layout55>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc568) + %311 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x28x28>}> : (!tt.device<#device>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc569) + %312 = "ttnn.relu"(%310, %311) <{operandSegmentSizes = array}> : (tensor<1x512x28x28xf32, #layout55>, tensor<1x512x28x28xf32, #layout55>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc569) + %313 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x512x28>}> : (!tt.device<#device>) -> tensor<1x28x512x28xf32, #layout54> loc(#loc570) + %314 = "ttnn.transpose"(%312, %313) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x512x28x28xf32, #layout55>, tensor<1x28x512x28xf32, #layout54>) -> tensor<1x28x512x28xf32, #layout54> loc(#loc570) + %315 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x512>}> : (!tt.device<#device>) -> tensor<1x28x28x512xf32, #layout53> loc(#loc571) + %316 = "ttnn.transpose"(%314, %315) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x512x28xf32, #layout54>, tensor<1x28x28x512xf32, #layout53>) -> tensor<1x28x28x512xf32, #layout53> loc(#loc571) + %317 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x128>}> : (!tt.device<#device>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc572) + %318 = "ttnn.conv2d"(%316, %arg122, %317, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 512 : i32, input_height = 28 : i32, input_width = 28 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 128 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x28x28x512xf32, #layout53>, tensor<128x512x1x1xf32, #layout16>, tensor<1x28x28x128xf32, #layout50>, !tt.device<#device>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc572) + %319 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x128x28>}> : (!tt.device<#device>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc573) + %320 = "ttnn.transpose"(%318, %319) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x28x128xf32, #layout50>, tensor<1x28x128x28xf32, #layout51>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc573) + %321 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc574) + %322 = "ttnn.transpose"(%320, %321) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x28x128x28xf32, #layout51>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc574) + %323 = "ttnn.to_layout"(%arg31, %0) <{layout = #ttnn.layout}> : (tensor<128x1x1xf32, #layout3>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc575) + %324 = "ttnn.to_device"(%323, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<128x1x1xf32, #layout49>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc575) + %325 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc575) + %326 = "ttnn.multiply"(%322, %324, %325) <{operandSegmentSizes = array}> : (tensor<1x128x28x28xf32, #layout52>, tensor<128x1x1xf32, #layout49>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc575) + %327 = "ttnn.to_layout"(%arg32, %0) <{layout = #ttnn.layout}> : (tensor<128x1x1xf32, #layout3>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc576) + %328 = "ttnn.to_device"(%327, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<128x1x1xf32, #layout49>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc576) + %329 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc576) + %330 = "ttnn.add"(%326, %328, %329) <{operandSegmentSizes = array}> : (tensor<1x128x28x28xf32, #layout52>, tensor<128x1x1xf32, #layout49>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc576) + %331 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc577) + %332 = "ttnn.relu"(%330, %331) <{operandSegmentSizes = array}> : (tensor<1x128x28x28xf32, #layout52>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc577) + %333 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x128x28>}> : (!tt.device<#device>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc578) + %334 = "ttnn.transpose"(%332, %333) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x128x28x28xf32, #layout52>, tensor<1x28x128x28xf32, #layout51>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc578) + %335 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x128>}> : (!tt.device<#device>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc579) + %336 = "ttnn.transpose"(%334, %335) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x128x28xf32, #layout51>, tensor<1x28x28x128xf32, #layout50>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc579) + %337 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x128>}> : (!tt.device<#device>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc580) + %338 = "ttnn.conv2d"(%336, %arg123, %337, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 128 : i32, input_height = 28 : i32, input_width = 28 : i32, kernel_height = 3 : i32, kernel_width = 3 : i32, out_channels = 128 : i32, padding_height = 1 : i32, padding_width = 1 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x28x28x128xf32, #layout50>, tensor<128x128x3x3xf32, #layout13>, tensor<1x28x28x128xf32, #layout50>, !tt.device<#device>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc580) + %339 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x128x28>}> : (!tt.device<#device>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc581) + %340 = "ttnn.transpose"(%338, %339) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x28x128xf32, #layout50>, tensor<1x28x128x28xf32, #layout51>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc581) + %341 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc582) + %342 = "ttnn.transpose"(%340, %341) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x28x128x28xf32, #layout51>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc582) + %343 = "ttnn.to_layout"(%arg33, %0) <{layout = #ttnn.layout}> : (tensor<128x1x1xf32, #layout3>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc583) + %344 = "ttnn.to_device"(%343, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<128x1x1xf32, #layout49>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc583) + %345 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc583) + %346 = "ttnn.multiply"(%342, %344, %345) <{operandSegmentSizes = array}> : (tensor<1x128x28x28xf32, #layout52>, tensor<128x1x1xf32, #layout49>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc583) + %347 = "ttnn.to_layout"(%arg34, %0) <{layout = #ttnn.layout}> : (tensor<128x1x1xf32, #layout3>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc584) + %348 = "ttnn.to_device"(%347, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<128x1x1xf32, #layout49>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc584) + %349 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc584) + %350 = "ttnn.add"(%346, %348, %349) <{operandSegmentSizes = array}> : (tensor<1x128x28x28xf32, #layout52>, tensor<128x1x1xf32, #layout49>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc584) + %351 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc585) + %352 = "ttnn.relu"(%350, %351) <{operandSegmentSizes = array}> : (tensor<1x128x28x28xf32, #layout52>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc585) + %353 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x128x28>}> : (!tt.device<#device>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc586) + %354 = "ttnn.transpose"(%352, %353) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x128x28x28xf32, #layout52>, tensor<1x28x128x28xf32, #layout51>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc586) + %355 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x128>}> : (!tt.device<#device>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc587) + %356 = "ttnn.transpose"(%354, %355) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x128x28xf32, #layout51>, tensor<1x28x28x128xf32, #layout50>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc587) + %357 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x512>}> : (!tt.device<#device>) -> tensor<1x28x28x512xf32, #layout53> loc(#loc588) + %358 = "ttnn.conv2d"(%356, %arg124, %357, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 128 : i32, input_height = 28 : i32, input_width = 28 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 512 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x28x28x128xf32, #layout50>, tensor<512x128x1x1xf32, #layout14>, tensor<1x28x28x512xf32, #layout53>, !tt.device<#device>) -> tensor<1x28x28x512xf32, #layout53> loc(#loc588) + %359 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x512x28>}> : (!tt.device<#device>) -> tensor<1x28x512x28xf32, #layout54> loc(#loc589) + %360 = "ttnn.transpose"(%358, %359) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x28x512xf32, #layout53>, tensor<1x28x512x28xf32, #layout54>) -> tensor<1x28x512x28xf32, #layout54> loc(#loc589) + %361 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x28x28>}> : (!tt.device<#device>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc590) + %362 = "ttnn.transpose"(%360, %361) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x28x512x28xf32, #layout54>, tensor<1x512x28x28xf32, #layout55>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc590) + %363 = "ttnn.to_layout"(%arg35, %0) <{layout = #ttnn.layout}> : (tensor<512x1x1xf32, #layout4>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc591) + %364 = "ttnn.to_device"(%363, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<512x1x1xf32, #layout56>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc591) + %365 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x28x28>}> : (!tt.device<#device>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc591) + %366 = "ttnn.multiply"(%362, %364, %365) <{operandSegmentSizes = array}> : (tensor<1x512x28x28xf32, #layout55>, tensor<512x1x1xf32, #layout56>, tensor<1x512x28x28xf32, #layout55>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc591) + %367 = "ttnn.to_layout"(%arg36, %0) <{layout = #ttnn.layout}> : (tensor<512x1x1xf32, #layout4>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc592) + %368 = "ttnn.to_device"(%367, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<512x1x1xf32, #layout56>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc592) + %369 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x28x28>}> : (!tt.device<#device>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc592) + %370 = "ttnn.add"(%366, %368, %369) <{operandSegmentSizes = array}> : (tensor<1x512x28x28xf32, #layout55>, tensor<512x1x1xf32, #layout56>, tensor<1x512x28x28xf32, #layout55>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc592) + %371 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x28x28>}> : (!tt.device<#device>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc593) + %372 = "ttnn.add"(%370, %312, %371) <{operandSegmentSizes = array}> : (tensor<1x512x28x28xf32, #layout55>, tensor<1x512x28x28xf32, #layout55>, tensor<1x512x28x28xf32, #layout55>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc593) + %373 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x28x28>}> : (!tt.device<#device>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc594) + %374 = "ttnn.relu"(%372, %373) <{operandSegmentSizes = array}> : (tensor<1x512x28x28xf32, #layout55>, tensor<1x512x28x28xf32, #layout55>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc594) + %375 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x512x28>}> : (!tt.device<#device>) -> tensor<1x28x512x28xf32, #layout54> loc(#loc595) + %376 = "ttnn.transpose"(%374, %375) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x512x28x28xf32, #layout55>, tensor<1x28x512x28xf32, #layout54>) -> tensor<1x28x512x28xf32, #layout54> loc(#loc595) + %377 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x512>}> : (!tt.device<#device>) -> tensor<1x28x28x512xf32, #layout53> loc(#loc596) + %378 = "ttnn.transpose"(%376, %377) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x512x28xf32, #layout54>, tensor<1x28x28x512xf32, #layout53>) -> tensor<1x28x28x512xf32, #layout53> loc(#loc596) + %379 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x128>}> : (!tt.device<#device>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc597) + %380 = "ttnn.conv2d"(%378, %arg125, %379, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 512 : i32, input_height = 28 : i32, input_width = 28 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 128 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x28x28x512xf32, #layout53>, tensor<128x512x1x1xf32, #layout16>, tensor<1x28x28x128xf32, #layout50>, !tt.device<#device>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc597) + %381 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x128x28>}> : (!tt.device<#device>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc598) + %382 = "ttnn.transpose"(%380, %381) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x28x128xf32, #layout50>, tensor<1x28x128x28xf32, #layout51>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc598) + %383 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc599) + %384 = "ttnn.transpose"(%382, %383) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x28x128x28xf32, #layout51>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc599) + %385 = "ttnn.to_layout"(%arg37, %0) <{layout = #ttnn.layout}> : (tensor<128x1x1xf32, #layout3>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc600) + %386 = "ttnn.to_device"(%385, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<128x1x1xf32, #layout49>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc600) + %387 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc600) + %388 = "ttnn.multiply"(%384, %386, %387) <{operandSegmentSizes = array}> : (tensor<1x128x28x28xf32, #layout52>, tensor<128x1x1xf32, #layout49>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc600) + %389 = "ttnn.to_layout"(%arg38, %0) <{layout = #ttnn.layout}> : (tensor<128x1x1xf32, #layout3>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc601) + %390 = "ttnn.to_device"(%389, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<128x1x1xf32, #layout49>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc601) + %391 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc601) + %392 = "ttnn.add"(%388, %390, %391) <{operandSegmentSizes = array}> : (tensor<1x128x28x28xf32, #layout52>, tensor<128x1x1xf32, #layout49>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc601) + %393 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc602) + %394 = "ttnn.relu"(%392, %393) <{operandSegmentSizes = array}> : (tensor<1x128x28x28xf32, #layout52>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc602) + %395 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x128x28>}> : (!tt.device<#device>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc603) + %396 = "ttnn.transpose"(%394, %395) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x128x28x28xf32, #layout52>, tensor<1x28x128x28xf32, #layout51>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc603) + %397 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x128>}> : (!tt.device<#device>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc604) + %398 = "ttnn.transpose"(%396, %397) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x128x28xf32, #layout51>, tensor<1x28x28x128xf32, #layout50>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc604) + %399 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x128>}> : (!tt.device<#device>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc605) + %400 = "ttnn.conv2d"(%398, %arg126, %399, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 128 : i32, input_height = 28 : i32, input_width = 28 : i32, kernel_height = 3 : i32, kernel_width = 3 : i32, out_channels = 128 : i32, padding_height = 1 : i32, padding_width = 1 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x28x28x128xf32, #layout50>, tensor<128x128x3x3xf32, #layout13>, tensor<1x28x28x128xf32, #layout50>, !tt.device<#device>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc605) + %401 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x128x28>}> : (!tt.device<#device>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc606) + %402 = "ttnn.transpose"(%400, %401) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x28x128xf32, #layout50>, tensor<1x28x128x28xf32, #layout51>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc606) + %403 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc607) + %404 = "ttnn.transpose"(%402, %403) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x28x128x28xf32, #layout51>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc607) + %405 = "ttnn.to_layout"(%arg39, %0) <{layout = #ttnn.layout}> : (tensor<128x1x1xf32, #layout3>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc608) + %406 = "ttnn.to_device"(%405, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<128x1x1xf32, #layout49>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc608) + %407 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc608) + %408 = "ttnn.multiply"(%404, %406, %407) <{operandSegmentSizes = array}> : (tensor<1x128x28x28xf32, #layout52>, tensor<128x1x1xf32, #layout49>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc608) + %409 = "ttnn.to_layout"(%arg40, %0) <{layout = #ttnn.layout}> : (tensor<128x1x1xf32, #layout3>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc609) + %410 = "ttnn.to_device"(%409, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<128x1x1xf32, #layout49>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc609) + %411 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc609) + %412 = "ttnn.add"(%408, %410, %411) <{operandSegmentSizes = array}> : (tensor<1x128x28x28xf32, #layout52>, tensor<128x1x1xf32, #layout49>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc609) + %413 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc610) + %414 = "ttnn.relu"(%412, %413) <{operandSegmentSizes = array}> : (tensor<1x128x28x28xf32, #layout52>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc610) + %415 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x128x28>}> : (!tt.device<#device>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc611) + %416 = "ttnn.transpose"(%414, %415) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x128x28x28xf32, #layout52>, tensor<1x28x128x28xf32, #layout51>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc611) + %417 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x128>}> : (!tt.device<#device>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc612) + %418 = "ttnn.transpose"(%416, %417) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x128x28xf32, #layout51>, tensor<1x28x28x128xf32, #layout50>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc612) + %419 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x512>}> : (!tt.device<#device>) -> tensor<1x28x28x512xf32, #layout53> loc(#loc613) + %420 = "ttnn.conv2d"(%418, %arg127, %419, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 128 : i32, input_height = 28 : i32, input_width = 28 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 512 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x28x28x128xf32, #layout50>, tensor<512x128x1x1xf32, #layout14>, tensor<1x28x28x512xf32, #layout53>, !tt.device<#device>) -> tensor<1x28x28x512xf32, #layout53> loc(#loc613) + %421 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x512x28>}> : (!tt.device<#device>) -> tensor<1x28x512x28xf32, #layout54> loc(#loc614) + %422 = "ttnn.transpose"(%420, %421) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x28x512xf32, #layout53>, tensor<1x28x512x28xf32, #layout54>) -> tensor<1x28x512x28xf32, #layout54> loc(#loc614) + %423 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x28x28>}> : (!tt.device<#device>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc615) + %424 = "ttnn.transpose"(%422, %423) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x28x512x28xf32, #layout54>, tensor<1x512x28x28xf32, #layout55>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc615) + %425 = "ttnn.to_layout"(%arg41, %0) <{layout = #ttnn.layout}> : (tensor<512x1x1xf32, #layout4>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc616) + %426 = "ttnn.to_device"(%425, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<512x1x1xf32, #layout56>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc616) + %427 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x28x28>}> : (!tt.device<#device>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc616) + %428 = "ttnn.multiply"(%424, %426, %427) <{operandSegmentSizes = array}> : (tensor<1x512x28x28xf32, #layout55>, tensor<512x1x1xf32, #layout56>, tensor<1x512x28x28xf32, #layout55>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc616) + %429 = "ttnn.to_layout"(%arg42, %0) <{layout = #ttnn.layout}> : (tensor<512x1x1xf32, #layout4>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc617) + %430 = "ttnn.to_device"(%429, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<512x1x1xf32, #layout56>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc617) + %431 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x28x28>}> : (!tt.device<#device>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc617) + %432 = "ttnn.add"(%428, %430, %431) <{operandSegmentSizes = array}> : (tensor<1x512x28x28xf32, #layout55>, tensor<512x1x1xf32, #layout56>, tensor<1x512x28x28xf32, #layout55>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc617) + %433 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x28x28>}> : (!tt.device<#device>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc618) + %434 = "ttnn.add"(%432, %374, %433) <{operandSegmentSizes = array}> : (tensor<1x512x28x28xf32, #layout55>, tensor<1x512x28x28xf32, #layout55>, tensor<1x512x28x28xf32, #layout55>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc618) + %435 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x28x28>}> : (!tt.device<#device>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc619) + %436 = "ttnn.relu"(%434, %435) <{operandSegmentSizes = array}> : (tensor<1x512x28x28xf32, #layout55>, tensor<1x512x28x28xf32, #layout55>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc619) + %437 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x512x28>}> : (!tt.device<#device>) -> tensor<1x28x512x28xf32, #layout54> loc(#loc620) + %438 = "ttnn.transpose"(%436, %437) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x512x28x28xf32, #layout55>, tensor<1x28x512x28xf32, #layout54>) -> tensor<1x28x512x28xf32, #layout54> loc(#loc620) + %439 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x512>}> : (!tt.device<#device>) -> tensor<1x28x28x512xf32, #layout53> loc(#loc621) + %440 = "ttnn.transpose"(%438, %439) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x512x28xf32, #layout54>, tensor<1x28x28x512xf32, #layout53>) -> tensor<1x28x28x512xf32, #layout53> loc(#loc621) + %441 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x128>}> : (!tt.device<#device>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc622) + %442 = "ttnn.conv2d"(%440, %arg128, %441, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 512 : i32, input_height = 28 : i32, input_width = 28 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 128 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x28x28x512xf32, #layout53>, tensor<128x512x1x1xf32, #layout16>, tensor<1x28x28x128xf32, #layout50>, !tt.device<#device>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc622) + %443 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x128x28>}> : (!tt.device<#device>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc623) + %444 = "ttnn.transpose"(%442, %443) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x28x128xf32, #layout50>, tensor<1x28x128x28xf32, #layout51>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc623) + %445 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc624) + %446 = "ttnn.transpose"(%444, %445) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x28x128x28xf32, #layout51>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc624) + %447 = "ttnn.to_layout"(%arg43, %0) <{layout = #ttnn.layout}> : (tensor<128x1x1xf32, #layout3>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc625) + %448 = "ttnn.to_device"(%447, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<128x1x1xf32, #layout49>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc625) + %449 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc625) + %450 = "ttnn.multiply"(%446, %448, %449) <{operandSegmentSizes = array}> : (tensor<1x128x28x28xf32, #layout52>, tensor<128x1x1xf32, #layout49>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc625) + %451 = "ttnn.to_layout"(%arg44, %0) <{layout = #ttnn.layout}> : (tensor<128x1x1xf32, #layout3>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc626) + %452 = "ttnn.to_device"(%451, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<128x1x1xf32, #layout49>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc626) + %453 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc626) + %454 = "ttnn.add"(%450, %452, %453) <{operandSegmentSizes = array}> : (tensor<1x128x28x28xf32, #layout52>, tensor<128x1x1xf32, #layout49>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc626) + %455 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc627) + %456 = "ttnn.relu"(%454, %455) <{operandSegmentSizes = array}> : (tensor<1x128x28x28xf32, #layout52>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc627) + %457 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x128x28>}> : (!tt.device<#device>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc628) + %458 = "ttnn.transpose"(%456, %457) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x128x28x28xf32, #layout52>, tensor<1x28x128x28xf32, #layout51>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc628) + %459 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x128>}> : (!tt.device<#device>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc629) + %460 = "ttnn.transpose"(%458, %459) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x128x28xf32, #layout51>, tensor<1x28x28x128xf32, #layout50>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc629) + %461 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x128>}> : (!tt.device<#device>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc630) + %462 = "ttnn.conv2d"(%460, %arg129, %461, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 128 : i32, input_height = 28 : i32, input_width = 28 : i32, kernel_height = 3 : i32, kernel_width = 3 : i32, out_channels = 128 : i32, padding_height = 1 : i32, padding_width = 1 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x28x28x128xf32, #layout50>, tensor<128x128x3x3xf32, #layout13>, tensor<1x28x28x128xf32, #layout50>, !tt.device<#device>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc630) + %463 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x128x28>}> : (!tt.device<#device>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc631) + %464 = "ttnn.transpose"(%462, %463) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x28x128xf32, #layout50>, tensor<1x28x128x28xf32, #layout51>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc631) + %465 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc632) + %466 = "ttnn.transpose"(%464, %465) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x28x128x28xf32, #layout51>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc632) + %467 = "ttnn.to_layout"(%arg45, %0) <{layout = #ttnn.layout}> : (tensor<128x1x1xf32, #layout3>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc633) + %468 = "ttnn.to_device"(%467, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<128x1x1xf32, #layout49>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc633) + %469 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc633) + %470 = "ttnn.multiply"(%466, %468, %469) <{operandSegmentSizes = array}> : (tensor<1x128x28x28xf32, #layout52>, tensor<128x1x1xf32, #layout49>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc633) + %471 = "ttnn.to_layout"(%arg46, %0) <{layout = #ttnn.layout}> : (tensor<128x1x1xf32, #layout3>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc634) + %472 = "ttnn.to_device"(%471, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<128x1x1xf32, #layout49>, !tt.device<#device>) -> tensor<128x1x1xf32, #layout49> loc(#loc634) + %473 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc634) + %474 = "ttnn.add"(%470, %472, %473) <{operandSegmentSizes = array}> : (tensor<1x128x28x28xf32, #layout52>, tensor<128x1x1xf32, #layout49>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc634) + %475 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x128x28x28>}> : (!tt.device<#device>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc635) + %476 = "ttnn.relu"(%474, %475) <{operandSegmentSizes = array}> : (tensor<1x128x28x28xf32, #layout52>, tensor<1x128x28x28xf32, #layout52>) -> tensor<1x128x28x28xf32, #layout52> loc(#loc635) + %477 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x128x28>}> : (!tt.device<#device>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc636) + %478 = "ttnn.transpose"(%476, %477) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x128x28x28xf32, #layout52>, tensor<1x28x128x28xf32, #layout51>) -> tensor<1x28x128x28xf32, #layout51> loc(#loc636) + %479 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x128>}> : (!tt.device<#device>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc637) + %480 = "ttnn.transpose"(%478, %479) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x128x28xf32, #layout51>, tensor<1x28x28x128xf32, #layout50>) -> tensor<1x28x28x128xf32, #layout50> loc(#loc637) + %481 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x512>}> : (!tt.device<#device>) -> tensor<1x28x28x512xf32, #layout53> loc(#loc638) + %482 = "ttnn.conv2d"(%480, %arg130, %481, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 128 : i32, input_height = 28 : i32, input_width = 28 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 512 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x28x28x128xf32, #layout50>, tensor<512x128x1x1xf32, #layout14>, tensor<1x28x28x512xf32, #layout53>, !tt.device<#device>) -> tensor<1x28x28x512xf32, #layout53> loc(#loc638) + %483 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x512x28>}> : (!tt.device<#device>) -> tensor<1x28x512x28xf32, #layout54> loc(#loc639) + %484 = "ttnn.transpose"(%482, %483) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x28x512xf32, #layout53>, tensor<1x28x512x28xf32, #layout54>) -> tensor<1x28x512x28xf32, #layout54> loc(#loc639) + %485 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x28x28>}> : (!tt.device<#device>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc640) + %486 = "ttnn.transpose"(%484, %485) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x28x512x28xf32, #layout54>, tensor<1x512x28x28xf32, #layout55>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc640) + %487 = "ttnn.to_layout"(%arg47, %0) <{layout = #ttnn.layout}> : (tensor<512x1x1xf32, #layout4>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc641) + %488 = "ttnn.to_device"(%487, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<512x1x1xf32, #layout56>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc641) + %489 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x28x28>}> : (!tt.device<#device>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc641) + %490 = "ttnn.multiply"(%486, %488, %489) <{operandSegmentSizes = array}> : (tensor<1x512x28x28xf32, #layout55>, tensor<512x1x1xf32, #layout56>, tensor<1x512x28x28xf32, #layout55>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc641) + %491 = "ttnn.to_layout"(%arg48, %0) <{layout = #ttnn.layout}> : (tensor<512x1x1xf32, #layout4>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc642) + %492 = "ttnn.to_device"(%491, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<512x1x1xf32, #layout56>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc642) + %493 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x28x28>}> : (!tt.device<#device>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc642) + %494 = "ttnn.add"(%490, %492, %493) <{operandSegmentSizes = array}> : (tensor<1x512x28x28xf32, #layout55>, tensor<512x1x1xf32, #layout56>, tensor<1x512x28x28xf32, #layout55>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc642) + %495 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x28x28>}> : (!tt.device<#device>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc643) + %496 = "ttnn.add"(%494, %436, %495) <{operandSegmentSizes = array}> : (tensor<1x512x28x28xf32, #layout55>, tensor<1x512x28x28xf32, #layout55>, tensor<1x512x28x28xf32, #layout55>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc643) + %497 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x28x28>}> : (!tt.device<#device>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc644) + %498 = "ttnn.relu"(%496, %497) <{operandSegmentSizes = array}> : (tensor<1x512x28x28xf32, #layout55>, tensor<1x512x28x28xf32, #layout55>) -> tensor<1x512x28x28xf32, #layout55> loc(#loc644) + %499 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x512x28>}> : (!tt.device<#device>) -> tensor<1x28x512x28xf32, #layout54> loc(#loc645) + %500 = "ttnn.transpose"(%498, %499) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x512x28x28xf32, #layout55>, tensor<1x28x512x28xf32, #layout54>) -> tensor<1x28x512x28xf32, #layout54> loc(#loc645) + %501 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x512>}> : (!tt.device<#device>) -> tensor<1x28x28x512xf32, #layout53> loc(#loc646) + %502 = "ttnn.transpose"(%500, %501) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x512x28xf32, #layout54>, tensor<1x28x28x512xf32, #layout53>) -> tensor<1x28x28x512xf32, #layout53> loc(#loc646) + %503 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x256>}> : (!tt.device<#device>) -> tensor<1x28x28x256xf32, #layout57> loc(#loc647) + %504 = "ttnn.conv2d"(%502, %arg131, %503, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 512 : i32, input_height = 28 : i32, input_width = 28 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 256 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x28x28x512xf32, #layout53>, tensor<256x512x1x1xf32, #layout17>, tensor<1x28x28x256xf32, #layout57>, !tt.device<#device>) -> tensor<1x28x28x256xf32, #layout57> loc(#loc647) + %505 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x256x28>}> : (!tt.device<#device>) -> tensor<1x28x256x28xf32, #layout58> loc(#loc648) + %506 = "ttnn.transpose"(%504, %505) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x28x256xf32, #layout57>, tensor<1x28x256x28xf32, #layout58>) -> tensor<1x28x256x28xf32, #layout58> loc(#loc648) + %507 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x28x28>}> : (!tt.device<#device>) -> tensor<1x256x28x28xf32, #layout59> loc(#loc649) + %508 = "ttnn.transpose"(%506, %507) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x28x256x28xf32, #layout58>, tensor<1x256x28x28xf32, #layout59>) -> tensor<1x256x28x28xf32, #layout59> loc(#loc649) + %509 = "ttnn.to_layout"(%arg49, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc650) + %510 = "ttnn.to_device"(%509, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc650) + %511 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x28x28>}> : (!tt.device<#device>) -> tensor<1x256x28x28xf32, #layout59> loc(#loc650) + %512 = "ttnn.multiply"(%508, %510, %511) <{operandSegmentSizes = array}> : (tensor<1x256x28x28xf32, #layout59>, tensor<256x1x1xf32, #layout45>, tensor<1x256x28x28xf32, #layout59>) -> tensor<1x256x28x28xf32, #layout59> loc(#loc650) + %513 = "ttnn.to_layout"(%arg50, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc651) + %514 = "ttnn.to_device"(%513, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc651) + %515 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x28x28>}> : (!tt.device<#device>) -> tensor<1x256x28x28xf32, #layout59> loc(#loc651) + %516 = "ttnn.add"(%512, %514, %515) <{operandSegmentSizes = array}> : (tensor<1x256x28x28xf32, #layout59>, tensor<256x1x1xf32, #layout45>, tensor<1x256x28x28xf32, #layout59>) -> tensor<1x256x28x28xf32, #layout59> loc(#loc651) + %517 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x28x28>}> : (!tt.device<#device>) -> tensor<1x256x28x28xf32, #layout59> loc(#loc652) + %518 = "ttnn.relu"(%516, %517) <{operandSegmentSizes = array}> : (tensor<1x256x28x28xf32, #layout59>, tensor<1x256x28x28xf32, #layout59>) -> tensor<1x256x28x28xf32, #layout59> loc(#loc652) + %519 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x256x28>}> : (!tt.device<#device>) -> tensor<1x28x256x28xf32, #layout58> loc(#loc653) + %520 = "ttnn.transpose"(%518, %519) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x256x28x28xf32, #layout59>, tensor<1x28x256x28xf32, #layout58>) -> tensor<1x28x256x28xf32, #layout58> loc(#loc653) + %521 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x256>}> : (!tt.device<#device>) -> tensor<1x28x28x256xf32, #layout57> loc(#loc654) + %522 = "ttnn.transpose"(%520, %521) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x256x28xf32, #layout58>, tensor<1x28x28x256xf32, #layout57>) -> tensor<1x28x28x256xf32, #layout57> loc(#loc654) + %523 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x256>}> : (!tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc655) + %524 = "ttnn.conv2d"(%522, %arg132, %523, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 256 : i32, input_height = 28 : i32, input_width = 28 : i32, kernel_height = 3 : i32, kernel_width = 3 : i32, out_channels = 256 : i32, padding_height = 1 : i32, padding_width = 1 : i32, stride_height = 2 : i32, stride_width = 2 : i32}> : (tensor<1x28x28x256xf32, #layout57>, tensor<256x256x3x3xf32, #layout18>, tensor<1x14x14x256xf32, #layout60>, !tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc655) + %525 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x256x14>}> : (!tt.device<#device>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc656) + %526 = "ttnn.transpose"(%524, %525) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x14x256xf32, #layout60>, tensor<1x14x256x14xf32, #layout61>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc656) + %527 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc657) + %528 = "ttnn.transpose"(%526, %527) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x14x256x14xf32, #layout61>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc657) + %529 = "ttnn.to_layout"(%arg51, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc658) + %530 = "ttnn.to_device"(%529, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc658) + %531 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc658) + %532 = "ttnn.multiply"(%528, %530, %531) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<256x1x1xf32, #layout45>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc658) + %533 = "ttnn.to_layout"(%arg52, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc659) + %534 = "ttnn.to_device"(%533, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc659) + %535 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc659) + %536 = "ttnn.add"(%532, %534, %535) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<256x1x1xf32, #layout45>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc659) + %537 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc660) + %538 = "ttnn.relu"(%536, %537) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc660) + %539 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x256x14>}> : (!tt.device<#device>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc661) + %540 = "ttnn.transpose"(%538, %539) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x256x14x14xf32, #layout62>, tensor<1x14x256x14xf32, #layout61>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc661) + %541 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x256>}> : (!tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc662) + %542 = "ttnn.transpose"(%540, %541) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x256x14xf32, #layout61>, tensor<1x14x14x256xf32, #layout60>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc662) + %543 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x1024>}> : (!tt.device<#device>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc663) + %544 = "ttnn.conv2d"(%542, %arg133, %543, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 256 : i32, input_height = 14 : i32, input_width = 14 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 1024 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x14x14x256xf32, #layout60>, tensor<1024x256x1x1xf32, #layout19>, tensor<1x14x14x1024xf32, #layout63>, !tt.device<#device>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc663) + %545 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x1024x14>}> : (!tt.device<#device>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc664) + %546 = "ttnn.transpose"(%544, %545) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x14x1024xf32, #layout63>, tensor<1x14x1024x14xf32, #layout64>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc664) + %547 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc665) + %548 = "ttnn.transpose"(%546, %547) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x14x1024x14xf32, #layout64>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc665) + %549 = "ttnn.to_layout"(%arg53, %0) <{layout = #ttnn.layout}> : (tensor<1024x1x1xf32, #layout5>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc666) + %550 = "ttnn.to_device"(%549, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<1024x1x1xf32, #layout66>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc666) + %551 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc666) + %552 = "ttnn.multiply"(%548, %550, %551) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1024x1x1xf32, #layout66>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc666) + %553 = "ttnn.to_layout"(%arg54, %0) <{layout = #ttnn.layout}> : (tensor<1024x1x1xf32, #layout5>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc667) + %554 = "ttnn.to_device"(%553, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<1024x1x1xf32, #layout66>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc667) + %555 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc667) + %556 = "ttnn.add"(%552, %554, %555) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1024x1x1xf32, #layout66>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc667) + %557 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x512x28>}> : (!tt.device<#device>) -> tensor<1x28x512x28xf32, #layout54> loc(#loc668) + %558 = "ttnn.transpose"(%498, %557) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x512x28x28xf32, #layout55>, tensor<1x28x512x28xf32, #layout54>) -> tensor<1x28x512x28xf32, #layout54> loc(#loc668) + %559 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x28x28x512>}> : (!tt.device<#device>) -> tensor<1x28x28x512xf32, #layout53> loc(#loc669) + %560 = "ttnn.transpose"(%558, %559) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x28x512x28xf32, #layout54>, tensor<1x28x28x512xf32, #layout53>) -> tensor<1x28x28x512xf32, #layout53> loc(#loc669) + %561 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x1024>}> : (!tt.device<#device>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc670) + %562 = "ttnn.conv2d"(%560, %arg134, %561, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 512 : i32, input_height = 28 : i32, input_width = 28 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 1024 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 2 : i32, stride_width = 2 : i32}> : (tensor<1x28x28x512xf32, #layout53>, tensor<1024x512x1x1xf32, #layout20>, tensor<1x14x14x1024xf32, #layout63>, !tt.device<#device>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc670) + %563 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x1024x14>}> : (!tt.device<#device>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc671) + %564 = "ttnn.transpose"(%562, %563) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x14x1024xf32, #layout63>, tensor<1x14x1024x14xf32, #layout64>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc671) + %565 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc672) + %566 = "ttnn.transpose"(%564, %565) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x14x1024x14xf32, #layout64>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc672) + %567 = "ttnn.to_layout"(%arg55, %0) <{layout = #ttnn.layout}> : (tensor<1024x1x1xf32, #layout5>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc673) + %568 = "ttnn.to_device"(%567, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<1024x1x1xf32, #layout66>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc673) + %569 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc673) + %570 = "ttnn.multiply"(%566, %568, %569) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1024x1x1xf32, #layout66>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc673) + %571 = "ttnn.to_layout"(%arg56, %0) <{layout = #ttnn.layout}> : (tensor<1024x1x1xf32, #layout5>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc674) + %572 = "ttnn.to_device"(%571, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<1024x1x1xf32, #layout66>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc674) + %573 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc674) + %574 = "ttnn.add"(%570, %572, %573) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1024x1x1xf32, #layout66>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc674) + %575 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc675) + %576 = "ttnn.add"(%556, %574, %575) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1x1024x14x14xf32, #layout65>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc675) + %577 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc676) + %578 = "ttnn.relu"(%576, %577) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc676) + %579 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x1024x14>}> : (!tt.device<#device>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc677) + %580 = "ttnn.transpose"(%578, %579) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1x14x1024x14xf32, #layout64>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc677) + %581 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x1024>}> : (!tt.device<#device>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc678) + %582 = "ttnn.transpose"(%580, %581) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x1024x14xf32, #layout64>, tensor<1x14x14x1024xf32, #layout63>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc678) + %583 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x256>}> : (!tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc679) + %584 = "ttnn.conv2d"(%582, %arg135, %583, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 1024 : i32, input_height = 14 : i32, input_width = 14 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 256 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x14x14x1024xf32, #layout63>, tensor<256x1024x1x1xf32, #layout21>, tensor<1x14x14x256xf32, #layout60>, !tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc679) + %585 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x256x14>}> : (!tt.device<#device>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc680) + %586 = "ttnn.transpose"(%584, %585) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x14x256xf32, #layout60>, tensor<1x14x256x14xf32, #layout61>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc680) + %587 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc681) + %588 = "ttnn.transpose"(%586, %587) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x14x256x14xf32, #layout61>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc681) + %589 = "ttnn.to_layout"(%arg57, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc682) + %590 = "ttnn.to_device"(%589, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc682) + %591 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc682) + %592 = "ttnn.multiply"(%588, %590, %591) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<256x1x1xf32, #layout45>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc682) + %593 = "ttnn.to_layout"(%arg58, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc683) + %594 = "ttnn.to_device"(%593, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc683) + %595 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc683) + %596 = "ttnn.add"(%592, %594, %595) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<256x1x1xf32, #layout45>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc683) + %597 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc684) + %598 = "ttnn.relu"(%596, %597) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc684) + %599 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x256x14>}> : (!tt.device<#device>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc685) + %600 = "ttnn.transpose"(%598, %599) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x256x14x14xf32, #layout62>, tensor<1x14x256x14xf32, #layout61>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc685) + %601 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x256>}> : (!tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc686) + %602 = "ttnn.transpose"(%600, %601) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x256x14xf32, #layout61>, tensor<1x14x14x256xf32, #layout60>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc686) + %603 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x256>}> : (!tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc687) + %604 = "ttnn.conv2d"(%602, %arg136, %603, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 256 : i32, input_height = 14 : i32, input_width = 14 : i32, kernel_height = 3 : i32, kernel_width = 3 : i32, out_channels = 256 : i32, padding_height = 1 : i32, padding_width = 1 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x14x14x256xf32, #layout60>, tensor<256x256x3x3xf32, #layout18>, tensor<1x14x14x256xf32, #layout60>, !tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc687) + %605 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x256x14>}> : (!tt.device<#device>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc688) + %606 = "ttnn.transpose"(%604, %605) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x14x256xf32, #layout60>, tensor<1x14x256x14xf32, #layout61>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc688) + %607 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc689) + %608 = "ttnn.transpose"(%606, %607) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x14x256x14xf32, #layout61>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc689) + %609 = "ttnn.to_layout"(%arg59, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc690) + %610 = "ttnn.to_device"(%609, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc690) + %611 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc690) + %612 = "ttnn.multiply"(%608, %610, %611) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<256x1x1xf32, #layout45>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc690) + %613 = "ttnn.to_layout"(%arg60, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc691) + %614 = "ttnn.to_device"(%613, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc691) + %615 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc691) + %616 = "ttnn.add"(%612, %614, %615) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<256x1x1xf32, #layout45>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc691) + %617 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc692) + %618 = "ttnn.relu"(%616, %617) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc692) + %619 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x256x14>}> : (!tt.device<#device>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc693) + %620 = "ttnn.transpose"(%618, %619) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x256x14x14xf32, #layout62>, tensor<1x14x256x14xf32, #layout61>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc693) + %621 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x256>}> : (!tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc694) + %622 = "ttnn.transpose"(%620, %621) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x256x14xf32, #layout61>, tensor<1x14x14x256xf32, #layout60>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc694) + %623 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x1024>}> : (!tt.device<#device>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc695) + %624 = "ttnn.conv2d"(%622, %arg137, %623, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 256 : i32, input_height = 14 : i32, input_width = 14 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 1024 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x14x14x256xf32, #layout60>, tensor<1024x256x1x1xf32, #layout19>, tensor<1x14x14x1024xf32, #layout63>, !tt.device<#device>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc695) + %625 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x1024x14>}> : (!tt.device<#device>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc696) + %626 = "ttnn.transpose"(%624, %625) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x14x1024xf32, #layout63>, tensor<1x14x1024x14xf32, #layout64>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc696) + %627 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc697) + %628 = "ttnn.transpose"(%626, %627) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x14x1024x14xf32, #layout64>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc697) + %629 = "ttnn.to_layout"(%arg61, %0) <{layout = #ttnn.layout}> : (tensor<1024x1x1xf32, #layout5>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc698) + %630 = "ttnn.to_device"(%629, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<1024x1x1xf32, #layout66>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc698) + %631 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc698) + %632 = "ttnn.multiply"(%628, %630, %631) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1024x1x1xf32, #layout66>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc698) + %633 = "ttnn.to_layout"(%arg62, %0) <{layout = #ttnn.layout}> : (tensor<1024x1x1xf32, #layout5>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc699) + %634 = "ttnn.to_device"(%633, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<1024x1x1xf32, #layout66>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc699) + %635 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc699) + %636 = "ttnn.add"(%632, %634, %635) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1024x1x1xf32, #layout66>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc699) + %637 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc700) + %638 = "ttnn.add"(%636, %578, %637) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1x1024x14x14xf32, #layout65>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc700) + %639 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc701) + %640 = "ttnn.relu"(%638, %639) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc701) + %641 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x1024x14>}> : (!tt.device<#device>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc702) + %642 = "ttnn.transpose"(%640, %641) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1x14x1024x14xf32, #layout64>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc702) + %643 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x1024>}> : (!tt.device<#device>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc703) + %644 = "ttnn.transpose"(%642, %643) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x1024x14xf32, #layout64>, tensor<1x14x14x1024xf32, #layout63>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc703) + %645 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x256>}> : (!tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc704) + %646 = "ttnn.conv2d"(%644, %arg138, %645, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 1024 : i32, input_height = 14 : i32, input_width = 14 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 256 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x14x14x1024xf32, #layout63>, tensor<256x1024x1x1xf32, #layout21>, tensor<1x14x14x256xf32, #layout60>, !tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc704) + %647 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x256x14>}> : (!tt.device<#device>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc705) + %648 = "ttnn.transpose"(%646, %647) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x14x256xf32, #layout60>, tensor<1x14x256x14xf32, #layout61>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc705) + %649 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc706) + %650 = "ttnn.transpose"(%648, %649) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x14x256x14xf32, #layout61>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc706) + %651 = "ttnn.to_layout"(%arg63, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc707) + %652 = "ttnn.to_device"(%651, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc707) + %653 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc707) + %654 = "ttnn.multiply"(%650, %652, %653) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<256x1x1xf32, #layout45>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc707) + %655 = "ttnn.to_layout"(%arg64, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc708) + %656 = "ttnn.to_device"(%655, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc708) + %657 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc708) + %658 = "ttnn.add"(%654, %656, %657) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<256x1x1xf32, #layout45>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc708) + %659 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc709) + %660 = "ttnn.relu"(%658, %659) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc709) + %661 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x256x14>}> : (!tt.device<#device>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc710) + %662 = "ttnn.transpose"(%660, %661) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x256x14x14xf32, #layout62>, tensor<1x14x256x14xf32, #layout61>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc710) + %663 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x256>}> : (!tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc711) + %664 = "ttnn.transpose"(%662, %663) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x256x14xf32, #layout61>, tensor<1x14x14x256xf32, #layout60>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc711) + %665 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x256>}> : (!tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc712) + %666 = "ttnn.conv2d"(%664, %arg139, %665, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 256 : i32, input_height = 14 : i32, input_width = 14 : i32, kernel_height = 3 : i32, kernel_width = 3 : i32, out_channels = 256 : i32, padding_height = 1 : i32, padding_width = 1 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x14x14x256xf32, #layout60>, tensor<256x256x3x3xf32, #layout18>, tensor<1x14x14x256xf32, #layout60>, !tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc712) + %667 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x256x14>}> : (!tt.device<#device>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc713) + %668 = "ttnn.transpose"(%666, %667) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x14x256xf32, #layout60>, tensor<1x14x256x14xf32, #layout61>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc713) + %669 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc714) + %670 = "ttnn.transpose"(%668, %669) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x14x256x14xf32, #layout61>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc714) + %671 = "ttnn.to_layout"(%arg65, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc715) + %672 = "ttnn.to_device"(%671, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc715) + %673 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc715) + %674 = "ttnn.multiply"(%670, %672, %673) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<256x1x1xf32, #layout45>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc715) + %675 = "ttnn.to_layout"(%arg66, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc716) + %676 = "ttnn.to_device"(%675, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc716) + %677 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc716) + %678 = "ttnn.add"(%674, %676, %677) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<256x1x1xf32, #layout45>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc716) + %679 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc717) + %680 = "ttnn.relu"(%678, %679) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc717) + %681 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x256x14>}> : (!tt.device<#device>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc718) + %682 = "ttnn.transpose"(%680, %681) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x256x14x14xf32, #layout62>, tensor<1x14x256x14xf32, #layout61>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc718) + %683 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x256>}> : (!tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc719) + %684 = "ttnn.transpose"(%682, %683) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x256x14xf32, #layout61>, tensor<1x14x14x256xf32, #layout60>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc719) + %685 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x1024>}> : (!tt.device<#device>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc720) + %686 = "ttnn.conv2d"(%684, %arg140, %685, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 256 : i32, input_height = 14 : i32, input_width = 14 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 1024 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x14x14x256xf32, #layout60>, tensor<1024x256x1x1xf32, #layout19>, tensor<1x14x14x1024xf32, #layout63>, !tt.device<#device>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc720) + %687 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x1024x14>}> : (!tt.device<#device>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc721) + %688 = "ttnn.transpose"(%686, %687) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x14x1024xf32, #layout63>, tensor<1x14x1024x14xf32, #layout64>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc721) + %689 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc722) + %690 = "ttnn.transpose"(%688, %689) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x14x1024x14xf32, #layout64>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc722) + %691 = "ttnn.to_layout"(%arg67, %0) <{layout = #ttnn.layout}> : (tensor<1024x1x1xf32, #layout5>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc723) + %692 = "ttnn.to_device"(%691, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<1024x1x1xf32, #layout66>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc723) + %693 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc723) + %694 = "ttnn.multiply"(%690, %692, %693) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1024x1x1xf32, #layout66>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc723) + %695 = "ttnn.to_layout"(%arg68, %0) <{layout = #ttnn.layout}> : (tensor<1024x1x1xf32, #layout5>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc724) + %696 = "ttnn.to_device"(%695, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<1024x1x1xf32, #layout66>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc724) + %697 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc724) + %698 = "ttnn.add"(%694, %696, %697) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1024x1x1xf32, #layout66>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc724) + %699 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc725) + %700 = "ttnn.add"(%698, %640, %699) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1x1024x14x14xf32, #layout65>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc725) + %701 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc726) + %702 = "ttnn.relu"(%700, %701) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc726) + %703 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x1024x14>}> : (!tt.device<#device>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc727) + %704 = "ttnn.transpose"(%702, %703) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1x14x1024x14xf32, #layout64>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc727) + %705 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x1024>}> : (!tt.device<#device>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc728) + %706 = "ttnn.transpose"(%704, %705) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x1024x14xf32, #layout64>, tensor<1x14x14x1024xf32, #layout63>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc728) + %707 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x256>}> : (!tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc729) + %708 = "ttnn.conv2d"(%706, %arg141, %707, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 1024 : i32, input_height = 14 : i32, input_width = 14 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 256 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x14x14x1024xf32, #layout63>, tensor<256x1024x1x1xf32, #layout21>, tensor<1x14x14x256xf32, #layout60>, !tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc729) + %709 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x256x14>}> : (!tt.device<#device>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc730) + %710 = "ttnn.transpose"(%708, %709) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x14x256xf32, #layout60>, tensor<1x14x256x14xf32, #layout61>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc730) + %711 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc731) + %712 = "ttnn.transpose"(%710, %711) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x14x256x14xf32, #layout61>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc731) + %713 = "ttnn.to_layout"(%arg69, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc732) + %714 = "ttnn.to_device"(%713, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc732) + %715 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc732) + %716 = "ttnn.multiply"(%712, %714, %715) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<256x1x1xf32, #layout45>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc732) + %717 = "ttnn.to_layout"(%arg70, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc733) + %718 = "ttnn.to_device"(%717, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc733) + %719 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc733) + %720 = "ttnn.add"(%716, %718, %719) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<256x1x1xf32, #layout45>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc733) + %721 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc734) + %722 = "ttnn.relu"(%720, %721) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc734) + %723 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x256x14>}> : (!tt.device<#device>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc735) + %724 = "ttnn.transpose"(%722, %723) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x256x14x14xf32, #layout62>, tensor<1x14x256x14xf32, #layout61>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc735) + %725 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x256>}> : (!tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc736) + %726 = "ttnn.transpose"(%724, %725) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x256x14xf32, #layout61>, tensor<1x14x14x256xf32, #layout60>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc736) + %727 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x256>}> : (!tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc737) + %728 = "ttnn.conv2d"(%726, %arg142, %727, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 256 : i32, input_height = 14 : i32, input_width = 14 : i32, kernel_height = 3 : i32, kernel_width = 3 : i32, out_channels = 256 : i32, padding_height = 1 : i32, padding_width = 1 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x14x14x256xf32, #layout60>, tensor<256x256x3x3xf32, #layout18>, tensor<1x14x14x256xf32, #layout60>, !tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc737) + %729 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x256x14>}> : (!tt.device<#device>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc738) + %730 = "ttnn.transpose"(%728, %729) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x14x256xf32, #layout60>, tensor<1x14x256x14xf32, #layout61>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc738) + %731 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc739) + %732 = "ttnn.transpose"(%730, %731) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x14x256x14xf32, #layout61>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc739) + %733 = "ttnn.to_layout"(%arg71, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc740) + %734 = "ttnn.to_device"(%733, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc740) + %735 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc740) + %736 = "ttnn.multiply"(%732, %734, %735) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<256x1x1xf32, #layout45>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc740) + %737 = "ttnn.to_layout"(%arg72, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc741) + %738 = "ttnn.to_device"(%737, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc741) + %739 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc741) + %740 = "ttnn.add"(%736, %738, %739) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<256x1x1xf32, #layout45>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc741) + %741 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc742) + %742 = "ttnn.relu"(%740, %741) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc742) + %743 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x256x14>}> : (!tt.device<#device>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc743) + %744 = "ttnn.transpose"(%742, %743) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x256x14x14xf32, #layout62>, tensor<1x14x256x14xf32, #layout61>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc743) + %745 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x256>}> : (!tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc744) + %746 = "ttnn.transpose"(%744, %745) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x256x14xf32, #layout61>, tensor<1x14x14x256xf32, #layout60>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc744) + %747 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x1024>}> : (!tt.device<#device>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc745) + %748 = "ttnn.conv2d"(%746, %arg143, %747, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 256 : i32, input_height = 14 : i32, input_width = 14 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 1024 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x14x14x256xf32, #layout60>, tensor<1024x256x1x1xf32, #layout19>, tensor<1x14x14x1024xf32, #layout63>, !tt.device<#device>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc745) + %749 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x1024x14>}> : (!tt.device<#device>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc746) + %750 = "ttnn.transpose"(%748, %749) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x14x1024xf32, #layout63>, tensor<1x14x1024x14xf32, #layout64>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc746) + %751 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc747) + %752 = "ttnn.transpose"(%750, %751) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x14x1024x14xf32, #layout64>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc747) + %753 = "ttnn.to_layout"(%arg73, %0) <{layout = #ttnn.layout}> : (tensor<1024x1x1xf32, #layout5>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc748) + %754 = "ttnn.to_device"(%753, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<1024x1x1xf32, #layout66>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc748) + %755 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc748) + %756 = "ttnn.multiply"(%752, %754, %755) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1024x1x1xf32, #layout66>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc748) + %757 = "ttnn.to_layout"(%arg74, %0) <{layout = #ttnn.layout}> : (tensor<1024x1x1xf32, #layout5>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc749) + %758 = "ttnn.to_device"(%757, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<1024x1x1xf32, #layout66>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc749) + %759 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc749) + %760 = "ttnn.add"(%756, %758, %759) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1024x1x1xf32, #layout66>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc749) + %761 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc750) + %762 = "ttnn.add"(%760, %702, %761) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1x1024x14x14xf32, #layout65>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc750) + %763 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc751) + %764 = "ttnn.relu"(%762, %763) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc751) + %765 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x1024x14>}> : (!tt.device<#device>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc752) + %766 = "ttnn.transpose"(%764, %765) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1x14x1024x14xf32, #layout64>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc752) + %767 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x1024>}> : (!tt.device<#device>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc753) + %768 = "ttnn.transpose"(%766, %767) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x1024x14xf32, #layout64>, tensor<1x14x14x1024xf32, #layout63>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc753) + %769 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x256>}> : (!tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc754) + %770 = "ttnn.conv2d"(%768, %arg144, %769, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 1024 : i32, input_height = 14 : i32, input_width = 14 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 256 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x14x14x1024xf32, #layout63>, tensor<256x1024x1x1xf32, #layout21>, tensor<1x14x14x256xf32, #layout60>, !tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc754) + %771 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x256x14>}> : (!tt.device<#device>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc755) + %772 = "ttnn.transpose"(%770, %771) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x14x256xf32, #layout60>, tensor<1x14x256x14xf32, #layout61>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc755) + %773 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc756) + %774 = "ttnn.transpose"(%772, %773) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x14x256x14xf32, #layout61>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc756) + %775 = "ttnn.to_layout"(%arg75, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc757) + %776 = "ttnn.to_device"(%775, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc757) + %777 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc757) + %778 = "ttnn.multiply"(%774, %776, %777) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<256x1x1xf32, #layout45>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc757) + %779 = "ttnn.to_layout"(%arg76, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc758) + %780 = "ttnn.to_device"(%779, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc758) + %781 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc758) + %782 = "ttnn.add"(%778, %780, %781) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<256x1x1xf32, #layout45>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc758) + %783 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc759) + %784 = "ttnn.relu"(%782, %783) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc759) + %785 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x256x14>}> : (!tt.device<#device>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc760) + %786 = "ttnn.transpose"(%784, %785) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x256x14x14xf32, #layout62>, tensor<1x14x256x14xf32, #layout61>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc760) + %787 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x256>}> : (!tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc761) + %788 = "ttnn.transpose"(%786, %787) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x256x14xf32, #layout61>, tensor<1x14x14x256xf32, #layout60>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc761) + %789 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x256>}> : (!tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc762) + %790 = "ttnn.conv2d"(%788, %arg145, %789, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 256 : i32, input_height = 14 : i32, input_width = 14 : i32, kernel_height = 3 : i32, kernel_width = 3 : i32, out_channels = 256 : i32, padding_height = 1 : i32, padding_width = 1 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x14x14x256xf32, #layout60>, tensor<256x256x3x3xf32, #layout18>, tensor<1x14x14x256xf32, #layout60>, !tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc762) + %791 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x256x14>}> : (!tt.device<#device>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc763) + %792 = "ttnn.transpose"(%790, %791) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x14x256xf32, #layout60>, tensor<1x14x256x14xf32, #layout61>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc763) + %793 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc764) + %794 = "ttnn.transpose"(%792, %793) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x14x256x14xf32, #layout61>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc764) + %795 = "ttnn.to_layout"(%arg77, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc765) + %796 = "ttnn.to_device"(%795, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc765) + %797 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc765) + %798 = "ttnn.multiply"(%794, %796, %797) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<256x1x1xf32, #layout45>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc765) + %799 = "ttnn.to_layout"(%arg78, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc766) + %800 = "ttnn.to_device"(%799, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc766) + %801 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc766) + %802 = "ttnn.add"(%798, %800, %801) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<256x1x1xf32, #layout45>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc766) + %803 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc767) + %804 = "ttnn.relu"(%802, %803) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc767) + %805 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x256x14>}> : (!tt.device<#device>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc768) + %806 = "ttnn.transpose"(%804, %805) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x256x14x14xf32, #layout62>, tensor<1x14x256x14xf32, #layout61>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc768) + %807 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x256>}> : (!tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc769) + %808 = "ttnn.transpose"(%806, %807) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x256x14xf32, #layout61>, tensor<1x14x14x256xf32, #layout60>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc769) + %809 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x1024>}> : (!tt.device<#device>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc770) + %810 = "ttnn.conv2d"(%808, %arg146, %809, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 256 : i32, input_height = 14 : i32, input_width = 14 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 1024 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x14x14x256xf32, #layout60>, tensor<1024x256x1x1xf32, #layout19>, tensor<1x14x14x1024xf32, #layout63>, !tt.device<#device>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc770) + %811 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x1024x14>}> : (!tt.device<#device>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc771) + %812 = "ttnn.transpose"(%810, %811) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x14x1024xf32, #layout63>, tensor<1x14x1024x14xf32, #layout64>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc771) + %813 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc772) + %814 = "ttnn.transpose"(%812, %813) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x14x1024x14xf32, #layout64>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc772) + %815 = "ttnn.to_layout"(%arg79, %0) <{layout = #ttnn.layout}> : (tensor<1024x1x1xf32, #layout5>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc773) + %816 = "ttnn.to_device"(%815, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<1024x1x1xf32, #layout66>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc773) + %817 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc773) + %818 = "ttnn.multiply"(%814, %816, %817) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1024x1x1xf32, #layout66>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc773) + %819 = "ttnn.to_layout"(%arg80, %0) <{layout = #ttnn.layout}> : (tensor<1024x1x1xf32, #layout5>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc774) + %820 = "ttnn.to_device"(%819, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<1024x1x1xf32, #layout66>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc774) + %821 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc774) + %822 = "ttnn.add"(%818, %820, %821) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1024x1x1xf32, #layout66>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc774) + %823 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc775) + %824 = "ttnn.add"(%822, %764, %823) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1x1024x14x14xf32, #layout65>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc775) + %825 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc776) + %826 = "ttnn.relu"(%824, %825) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc776) + %827 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x1024x14>}> : (!tt.device<#device>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc777) + %828 = "ttnn.transpose"(%826, %827) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1x14x1024x14xf32, #layout64>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc777) + %829 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x1024>}> : (!tt.device<#device>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc778) + %830 = "ttnn.transpose"(%828, %829) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x1024x14xf32, #layout64>, tensor<1x14x14x1024xf32, #layout63>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc778) + %831 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x256>}> : (!tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc779) + %832 = "ttnn.conv2d"(%830, %arg147, %831, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 1024 : i32, input_height = 14 : i32, input_width = 14 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 256 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x14x14x1024xf32, #layout63>, tensor<256x1024x1x1xf32, #layout21>, tensor<1x14x14x256xf32, #layout60>, !tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc779) + %833 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x256x14>}> : (!tt.device<#device>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc780) + %834 = "ttnn.transpose"(%832, %833) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x14x256xf32, #layout60>, tensor<1x14x256x14xf32, #layout61>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc780) + %835 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc781) + %836 = "ttnn.transpose"(%834, %835) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x14x256x14xf32, #layout61>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc781) + %837 = "ttnn.to_layout"(%arg81, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc782) + %838 = "ttnn.to_device"(%837, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc782) + %839 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc782) + %840 = "ttnn.multiply"(%836, %838, %839) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<256x1x1xf32, #layout45>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc782) + %841 = "ttnn.to_layout"(%arg82, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc783) + %842 = "ttnn.to_device"(%841, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc783) + %843 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc783) + %844 = "ttnn.add"(%840, %842, %843) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<256x1x1xf32, #layout45>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc783) + %845 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc784) + %846 = "ttnn.relu"(%844, %845) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc784) + %847 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x256x14>}> : (!tt.device<#device>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc785) + %848 = "ttnn.transpose"(%846, %847) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x256x14x14xf32, #layout62>, tensor<1x14x256x14xf32, #layout61>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc785) + %849 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x256>}> : (!tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc786) + %850 = "ttnn.transpose"(%848, %849) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x256x14xf32, #layout61>, tensor<1x14x14x256xf32, #layout60>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc786) + %851 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x256>}> : (!tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc787) + %852 = "ttnn.conv2d"(%850, %arg148, %851, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 256 : i32, input_height = 14 : i32, input_width = 14 : i32, kernel_height = 3 : i32, kernel_width = 3 : i32, out_channels = 256 : i32, padding_height = 1 : i32, padding_width = 1 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x14x14x256xf32, #layout60>, tensor<256x256x3x3xf32, #layout18>, tensor<1x14x14x256xf32, #layout60>, !tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc787) + %853 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x256x14>}> : (!tt.device<#device>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc788) + %854 = "ttnn.transpose"(%852, %853) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x14x256xf32, #layout60>, tensor<1x14x256x14xf32, #layout61>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc788) + %855 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc789) + %856 = "ttnn.transpose"(%854, %855) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x14x256x14xf32, #layout61>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc789) + %857 = "ttnn.to_layout"(%arg83, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc790) + %858 = "ttnn.to_device"(%857, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc790) + %859 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc790) + %860 = "ttnn.multiply"(%856, %858, %859) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<256x1x1xf32, #layout45>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc790) + %861 = "ttnn.to_layout"(%arg84, %0) <{layout = #ttnn.layout}> : (tensor<256x1x1xf32, #layout2>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc791) + %862 = "ttnn.to_device"(%861, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<256x1x1xf32, #layout45>, !tt.device<#device>) -> tensor<256x1x1xf32, #layout45> loc(#loc791) + %863 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc791) + %864 = "ttnn.add"(%860, %862, %863) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<256x1x1xf32, #layout45>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc791) + %865 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x256x14x14>}> : (!tt.device<#device>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc792) + %866 = "ttnn.relu"(%864, %865) <{operandSegmentSizes = array}> : (tensor<1x256x14x14xf32, #layout62>, tensor<1x256x14x14xf32, #layout62>) -> tensor<1x256x14x14xf32, #layout62> loc(#loc792) + %867 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x256x14>}> : (!tt.device<#device>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc793) + %868 = "ttnn.transpose"(%866, %867) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x256x14x14xf32, #layout62>, tensor<1x14x256x14xf32, #layout61>) -> tensor<1x14x256x14xf32, #layout61> loc(#loc793) + %869 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x256>}> : (!tt.device<#device>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc794) + %870 = "ttnn.transpose"(%868, %869) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x256x14xf32, #layout61>, tensor<1x14x14x256xf32, #layout60>) -> tensor<1x14x14x256xf32, #layout60> loc(#loc794) + %871 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x1024>}> : (!tt.device<#device>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc795) + %872 = "ttnn.conv2d"(%870, %arg149, %871, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 256 : i32, input_height = 14 : i32, input_width = 14 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 1024 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x14x14x256xf32, #layout60>, tensor<1024x256x1x1xf32, #layout19>, tensor<1x14x14x1024xf32, #layout63>, !tt.device<#device>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc795) + %873 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x1024x14>}> : (!tt.device<#device>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc796) + %874 = "ttnn.transpose"(%872, %873) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x14x1024xf32, #layout63>, tensor<1x14x1024x14xf32, #layout64>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc796) + %875 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc797) + %876 = "ttnn.transpose"(%874, %875) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x14x1024x14xf32, #layout64>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc797) + %877 = "ttnn.to_layout"(%arg85, %0) <{layout = #ttnn.layout}> : (tensor<1024x1x1xf32, #layout5>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc798) + %878 = "ttnn.to_device"(%877, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<1024x1x1xf32, #layout66>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc798) + %879 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc798) + %880 = "ttnn.multiply"(%876, %878, %879) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1024x1x1xf32, #layout66>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc798) + %881 = "ttnn.to_layout"(%arg86, %0) <{layout = #ttnn.layout}> : (tensor<1024x1x1xf32, #layout5>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc799) + %882 = "ttnn.to_device"(%881, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<1024x1x1xf32, #layout66>, !tt.device<#device>) -> tensor<1024x1x1xf32, #layout66> loc(#loc799) + %883 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc799) + %884 = "ttnn.add"(%880, %882, %883) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1024x1x1xf32, #layout66>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc799) + %885 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc800) + %886 = "ttnn.add"(%884, %826, %885) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1x1024x14x14xf32, #layout65>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc800) + %887 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1024x14x14>}> : (!tt.device<#device>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc801) + %888 = "ttnn.relu"(%886, %887) <{operandSegmentSizes = array}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1x1024x14x14xf32, #layout65>) -> tensor<1x1024x14x14xf32, #layout65> loc(#loc801) + %889 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x1024x14>}> : (!tt.device<#device>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc802) + %890 = "ttnn.transpose"(%888, %889) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1x14x1024x14xf32, #layout64>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc802) + %891 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x1024>}> : (!tt.device<#device>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc803) + %892 = "ttnn.transpose"(%890, %891) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x1024x14xf32, #layout64>, tensor<1x14x14x1024xf32, #layout63>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc803) + %893 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x512>}> : (!tt.device<#device>) -> tensor<1x14x14x512xf32, #layout67> loc(#loc804) + %894 = "ttnn.conv2d"(%892, %arg150, %893, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 1024 : i32, input_height = 14 : i32, input_width = 14 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 512 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x14x14x1024xf32, #layout63>, tensor<512x1024x1x1xf32, #layout22>, tensor<1x14x14x512xf32, #layout67>, !tt.device<#device>) -> tensor<1x14x14x512xf32, #layout67> loc(#loc804) + %895 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x512x14>}> : (!tt.device<#device>) -> tensor<1x14x512x14xf32, #layout68> loc(#loc805) + %896 = "ttnn.transpose"(%894, %895) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x14x512xf32, #layout67>, tensor<1x14x512x14xf32, #layout68>) -> tensor<1x14x512x14xf32, #layout68> loc(#loc805) + %897 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x14x14>}> : (!tt.device<#device>) -> tensor<1x512x14x14xf32, #layout69> loc(#loc806) + %898 = "ttnn.transpose"(%896, %897) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x14x512x14xf32, #layout68>, tensor<1x512x14x14xf32, #layout69>) -> tensor<1x512x14x14xf32, #layout69> loc(#loc806) + %899 = "ttnn.to_layout"(%arg87, %0) <{layout = #ttnn.layout}> : (tensor<512x1x1xf32, #layout4>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc807) + %900 = "ttnn.to_device"(%899, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<512x1x1xf32, #layout56>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc807) + %901 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x14x14>}> : (!tt.device<#device>) -> tensor<1x512x14x14xf32, #layout69> loc(#loc807) + %902 = "ttnn.multiply"(%898, %900, %901) <{operandSegmentSizes = array}> : (tensor<1x512x14x14xf32, #layout69>, tensor<512x1x1xf32, #layout56>, tensor<1x512x14x14xf32, #layout69>) -> tensor<1x512x14x14xf32, #layout69> loc(#loc807) + %903 = "ttnn.to_layout"(%arg88, %0) <{layout = #ttnn.layout}> : (tensor<512x1x1xf32, #layout4>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc808) + %904 = "ttnn.to_device"(%903, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<512x1x1xf32, #layout56>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc808) + %905 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x14x14>}> : (!tt.device<#device>) -> tensor<1x512x14x14xf32, #layout69> loc(#loc808) + %906 = "ttnn.add"(%902, %904, %905) <{operandSegmentSizes = array}> : (tensor<1x512x14x14xf32, #layout69>, tensor<512x1x1xf32, #layout56>, tensor<1x512x14x14xf32, #layout69>) -> tensor<1x512x14x14xf32, #layout69> loc(#loc808) + %907 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x14x14>}> : (!tt.device<#device>) -> tensor<1x512x14x14xf32, #layout69> loc(#loc809) + %908 = "ttnn.relu"(%906, %907) <{operandSegmentSizes = array}> : (tensor<1x512x14x14xf32, #layout69>, tensor<1x512x14x14xf32, #layout69>) -> tensor<1x512x14x14xf32, #layout69> loc(#loc809) + %909 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x512x14>}> : (!tt.device<#device>) -> tensor<1x14x512x14xf32, #layout68> loc(#loc810) + %910 = "ttnn.transpose"(%908, %909) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x512x14x14xf32, #layout69>, tensor<1x14x512x14xf32, #layout68>) -> tensor<1x14x512x14xf32, #layout68> loc(#loc810) + %911 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x512>}> : (!tt.device<#device>) -> tensor<1x14x14x512xf32, #layout67> loc(#loc811) + %912 = "ttnn.transpose"(%910, %911) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x512x14xf32, #layout68>, tensor<1x14x14x512xf32, #layout67>) -> tensor<1x14x14x512xf32, #layout67> loc(#loc811) + %913 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x7x512>}> : (!tt.device<#device>) -> tensor<1x7x7x512xf32, #layout70> loc(#loc812) + %914 = "ttnn.conv2d"(%912, %arg151, %913, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 512 : i32, input_height = 14 : i32, input_width = 14 : i32, kernel_height = 3 : i32, kernel_width = 3 : i32, out_channels = 512 : i32, padding_height = 1 : i32, padding_width = 1 : i32, stride_height = 2 : i32, stride_width = 2 : i32}> : (tensor<1x14x14x512xf32, #layout67>, tensor<512x512x3x3xf32, #layout23>, tensor<1x7x7x512xf32, #layout70>, !tt.device<#device>) -> tensor<1x7x7x512xf32, #layout70> loc(#loc812) + %915 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x512x7>}> : (!tt.device<#device>) -> tensor<1x7x512x7xf32, #layout71> loc(#loc813) + %916 = "ttnn.transpose"(%914, %915) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x7x7x512xf32, #layout70>, tensor<1x7x512x7xf32, #layout71>) -> tensor<1x7x512x7xf32, #layout71> loc(#loc813) + %917 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x7x7>}> : (!tt.device<#device>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc814) + %918 = "ttnn.transpose"(%916, %917) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x7x512x7xf32, #layout71>, tensor<1x512x7x7xf32, #layout72>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc814) + %919 = "ttnn.to_layout"(%arg89, %0) <{layout = #ttnn.layout}> : (tensor<512x1x1xf32, #layout4>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc815) + %920 = "ttnn.to_device"(%919, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<512x1x1xf32, #layout56>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc815) + %921 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x7x7>}> : (!tt.device<#device>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc815) + %922 = "ttnn.multiply"(%918, %920, %921) <{operandSegmentSizes = array}> : (tensor<1x512x7x7xf32, #layout72>, tensor<512x1x1xf32, #layout56>, tensor<1x512x7x7xf32, #layout72>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc815) + %923 = "ttnn.to_layout"(%arg90, %0) <{layout = #ttnn.layout}> : (tensor<512x1x1xf32, #layout4>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc816) + %924 = "ttnn.to_device"(%923, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<512x1x1xf32, #layout56>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc816) + %925 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x7x7>}> : (!tt.device<#device>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc816) + %926 = "ttnn.add"(%922, %924, %925) <{operandSegmentSizes = array}> : (tensor<1x512x7x7xf32, #layout72>, tensor<512x1x1xf32, #layout56>, tensor<1x512x7x7xf32, #layout72>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc816) + %927 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x7x7>}> : (!tt.device<#device>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc817) + %928 = "ttnn.relu"(%926, %927) <{operandSegmentSizes = array}> : (tensor<1x512x7x7xf32, #layout72>, tensor<1x512x7x7xf32, #layout72>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc817) + %929 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x512x7>}> : (!tt.device<#device>) -> tensor<1x7x512x7xf32, #layout71> loc(#loc818) + %930 = "ttnn.transpose"(%928, %929) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x512x7x7xf32, #layout72>, tensor<1x7x512x7xf32, #layout71>) -> tensor<1x7x512x7xf32, #layout71> loc(#loc818) + %931 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x7x512>}> : (!tt.device<#device>) -> tensor<1x7x7x512xf32, #layout70> loc(#loc819) + %932 = "ttnn.transpose"(%930, %931) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x7x512x7xf32, #layout71>, tensor<1x7x7x512xf32, #layout70>) -> tensor<1x7x7x512xf32, #layout70> loc(#loc819) + %933 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x7x2048>}> : (!tt.device<#device>) -> tensor<1x7x7x2048xf32, #layout73> loc(#loc820) + %934 = "ttnn.conv2d"(%932, %arg152, %933, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 512 : i32, input_height = 7 : i32, input_width = 7 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 2048 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x7x7x512xf32, #layout70>, tensor<2048x512x1x1xf32, #layout24>, tensor<1x7x7x2048xf32, #layout73>, !tt.device<#device>) -> tensor<1x7x7x2048xf32, #layout73> loc(#loc820) + %935 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x2048x7>}> : (!tt.device<#device>) -> tensor<1x7x2048x7xf32, #layout74> loc(#loc821) + %936 = "ttnn.transpose"(%934, %935) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x7x7x2048xf32, #layout73>, tensor<1x7x2048x7xf32, #layout74>) -> tensor<1x7x2048x7xf32, #layout74> loc(#loc821) + %937 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x2048x7x7>}> : (!tt.device<#device>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc822) + %938 = "ttnn.transpose"(%936, %937) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x7x2048x7xf32, #layout74>, tensor<1x2048x7x7xf32, #layout75>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc822) + %939 = "ttnn.to_layout"(%arg91, %0) <{layout = #ttnn.layout}> : (tensor<2048x1x1xf32, #layout6>, !tt.device<#device>) -> tensor<2048x1x1xf32, #layout76> loc(#loc823) + %940 = "ttnn.to_device"(%939, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<2048x1x1xf32, #layout76>, !tt.device<#device>) -> tensor<2048x1x1xf32, #layout76> loc(#loc823) + %941 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x2048x7x7>}> : (!tt.device<#device>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc823) + %942 = "ttnn.multiply"(%938, %940, %941) <{operandSegmentSizes = array}> : (tensor<1x2048x7x7xf32, #layout75>, tensor<2048x1x1xf32, #layout76>, tensor<1x2048x7x7xf32, #layout75>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc823) + %943 = "ttnn.to_layout"(%arg92, %0) <{layout = #ttnn.layout}> : (tensor<2048x1x1xf32, #layout6>, !tt.device<#device>) -> tensor<2048x1x1xf32, #layout76> loc(#loc824) + %944 = "ttnn.to_device"(%943, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<2048x1x1xf32, #layout76>, !tt.device<#device>) -> tensor<2048x1x1xf32, #layout76> loc(#loc824) + %945 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x2048x7x7>}> : (!tt.device<#device>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc824) + %946 = "ttnn.add"(%942, %944, %945) <{operandSegmentSizes = array}> : (tensor<1x2048x7x7xf32, #layout75>, tensor<2048x1x1xf32, #layout76>, tensor<1x2048x7x7xf32, #layout75>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc824) + %947 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x1024x14>}> : (!tt.device<#device>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc825) + %948 = "ttnn.transpose"(%888, %947) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x1024x14x14xf32, #layout65>, tensor<1x14x1024x14xf32, #layout64>) -> tensor<1x14x1024x14xf32, #layout64> loc(#loc825) + %949 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x14x14x1024>}> : (!tt.device<#device>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc826) + %950 = "ttnn.transpose"(%948, %949) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x14x1024x14xf32, #layout64>, tensor<1x14x14x1024xf32, #layout63>) -> tensor<1x14x14x1024xf32, #layout63> loc(#loc826) + %951 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x7x2048>}> : (!tt.device<#device>) -> tensor<1x7x7x2048xf32, #layout73> loc(#loc827) + %952 = "ttnn.conv2d"(%950, %arg153, %951, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 1024 : i32, input_height = 14 : i32, input_width = 14 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 2048 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 2 : i32, stride_width = 2 : i32}> : (tensor<1x14x14x1024xf32, #layout63>, tensor<2048x1024x1x1xf32, #layout25>, tensor<1x7x7x2048xf32, #layout73>, !tt.device<#device>) -> tensor<1x7x7x2048xf32, #layout73> loc(#loc827) + %953 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x2048x7>}> : (!tt.device<#device>) -> tensor<1x7x2048x7xf32, #layout74> loc(#loc828) + %954 = "ttnn.transpose"(%952, %953) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x7x7x2048xf32, #layout73>, tensor<1x7x2048x7xf32, #layout74>) -> tensor<1x7x2048x7xf32, #layout74> loc(#loc828) + %955 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x2048x7x7>}> : (!tt.device<#device>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc829) + %956 = "ttnn.transpose"(%954, %955) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x7x2048x7xf32, #layout74>, tensor<1x2048x7x7xf32, #layout75>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc829) + %957 = "ttnn.to_layout"(%arg93, %0) <{layout = #ttnn.layout}> : (tensor<2048x1x1xf32, #layout6>, !tt.device<#device>) -> tensor<2048x1x1xf32, #layout76> loc(#loc830) + %958 = "ttnn.to_device"(%957, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<2048x1x1xf32, #layout76>, !tt.device<#device>) -> tensor<2048x1x1xf32, #layout76> loc(#loc830) + %959 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x2048x7x7>}> : (!tt.device<#device>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc830) + %960 = "ttnn.multiply"(%956, %958, %959) <{operandSegmentSizes = array}> : (tensor<1x2048x7x7xf32, #layout75>, tensor<2048x1x1xf32, #layout76>, tensor<1x2048x7x7xf32, #layout75>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc830) + %961 = "ttnn.to_layout"(%arg94, %0) <{layout = #ttnn.layout}> : (tensor<2048x1x1xf32, #layout6>, !tt.device<#device>) -> tensor<2048x1x1xf32, #layout76> loc(#loc831) + %962 = "ttnn.to_device"(%961, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<2048x1x1xf32, #layout76>, !tt.device<#device>) -> tensor<2048x1x1xf32, #layout76> loc(#loc831) + %963 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x2048x7x7>}> : (!tt.device<#device>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc831) + %964 = "ttnn.add"(%960, %962, %963) <{operandSegmentSizes = array}> : (tensor<1x2048x7x7xf32, #layout75>, tensor<2048x1x1xf32, #layout76>, tensor<1x2048x7x7xf32, #layout75>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc831) + %965 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x2048x7x7>}> : (!tt.device<#device>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc832) + %966 = "ttnn.add"(%946, %964, %965) <{operandSegmentSizes = array}> : (tensor<1x2048x7x7xf32, #layout75>, tensor<1x2048x7x7xf32, #layout75>, tensor<1x2048x7x7xf32, #layout75>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc832) + %967 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x2048x7x7>}> : (!tt.device<#device>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc833) + %968 = "ttnn.relu"(%966, %967) <{operandSegmentSizes = array}> : (tensor<1x2048x7x7xf32, #layout75>, tensor<1x2048x7x7xf32, #layout75>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc833) + %969 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x2048x7>}> : (!tt.device<#device>) -> tensor<1x7x2048x7xf32, #layout74> loc(#loc834) + %970 = "ttnn.transpose"(%968, %969) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x2048x7x7xf32, #layout75>, tensor<1x7x2048x7xf32, #layout74>) -> tensor<1x7x2048x7xf32, #layout74> loc(#loc834) + %971 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x7x2048>}> : (!tt.device<#device>) -> tensor<1x7x7x2048xf32, #layout73> loc(#loc835) + %972 = "ttnn.transpose"(%970, %971) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x7x2048x7xf32, #layout74>, tensor<1x7x7x2048xf32, #layout73>) -> tensor<1x7x7x2048xf32, #layout73> loc(#loc835) + %973 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x7x512>}> : (!tt.device<#device>) -> tensor<1x7x7x512xf32, #layout70> loc(#loc836) + %974 = "ttnn.conv2d"(%972, %arg154, %973, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 2048 : i32, input_height = 7 : i32, input_width = 7 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 512 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x7x7x2048xf32, #layout73>, tensor<512x2048x1x1xf32, #layout26>, tensor<1x7x7x512xf32, #layout70>, !tt.device<#device>) -> tensor<1x7x7x512xf32, #layout70> loc(#loc836) + %975 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x512x7>}> : (!tt.device<#device>) -> tensor<1x7x512x7xf32, #layout71> loc(#loc837) + %976 = "ttnn.transpose"(%974, %975) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x7x7x512xf32, #layout70>, tensor<1x7x512x7xf32, #layout71>) -> tensor<1x7x512x7xf32, #layout71> loc(#loc837) + %977 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x7x7>}> : (!tt.device<#device>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc838) + %978 = "ttnn.transpose"(%976, %977) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x7x512x7xf32, #layout71>, tensor<1x512x7x7xf32, #layout72>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc838) + %979 = "ttnn.to_layout"(%arg95, %0) <{layout = #ttnn.layout}> : (tensor<512x1x1xf32, #layout4>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc839) + %980 = "ttnn.to_device"(%979, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<512x1x1xf32, #layout56>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc839) + %981 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x7x7>}> : (!tt.device<#device>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc839) + %982 = "ttnn.multiply"(%978, %980, %981) <{operandSegmentSizes = array}> : (tensor<1x512x7x7xf32, #layout72>, tensor<512x1x1xf32, #layout56>, tensor<1x512x7x7xf32, #layout72>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc839) + %983 = "ttnn.to_layout"(%arg96, %0) <{layout = #ttnn.layout}> : (tensor<512x1x1xf32, #layout4>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc840) + %984 = "ttnn.to_device"(%983, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<512x1x1xf32, #layout56>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc840) + %985 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x7x7>}> : (!tt.device<#device>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc840) + %986 = "ttnn.add"(%982, %984, %985) <{operandSegmentSizes = array}> : (tensor<1x512x7x7xf32, #layout72>, tensor<512x1x1xf32, #layout56>, tensor<1x512x7x7xf32, #layout72>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc840) + %987 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x7x7>}> : (!tt.device<#device>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc841) + %988 = "ttnn.relu"(%986, %987) <{operandSegmentSizes = array}> : (tensor<1x512x7x7xf32, #layout72>, tensor<1x512x7x7xf32, #layout72>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc841) + %989 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x512x7>}> : (!tt.device<#device>) -> tensor<1x7x512x7xf32, #layout71> loc(#loc842) + %990 = "ttnn.transpose"(%988, %989) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x512x7x7xf32, #layout72>, tensor<1x7x512x7xf32, #layout71>) -> tensor<1x7x512x7xf32, #layout71> loc(#loc842) + %991 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x7x512>}> : (!tt.device<#device>) -> tensor<1x7x7x512xf32, #layout70> loc(#loc843) + %992 = "ttnn.transpose"(%990, %991) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x7x512x7xf32, #layout71>, tensor<1x7x7x512xf32, #layout70>) -> tensor<1x7x7x512xf32, #layout70> loc(#loc843) + %993 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x7x512>}> : (!tt.device<#device>) -> tensor<1x7x7x512xf32, #layout70> loc(#loc844) + %994 = "ttnn.conv2d"(%992, %arg155, %993, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 512 : i32, input_height = 7 : i32, input_width = 7 : i32, kernel_height = 3 : i32, kernel_width = 3 : i32, out_channels = 512 : i32, padding_height = 1 : i32, padding_width = 1 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x7x7x512xf32, #layout70>, tensor<512x512x3x3xf32, #layout23>, tensor<1x7x7x512xf32, #layout70>, !tt.device<#device>) -> tensor<1x7x7x512xf32, #layout70> loc(#loc844) + %995 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x512x7>}> : (!tt.device<#device>) -> tensor<1x7x512x7xf32, #layout71> loc(#loc845) + %996 = "ttnn.transpose"(%994, %995) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x7x7x512xf32, #layout70>, tensor<1x7x512x7xf32, #layout71>) -> tensor<1x7x512x7xf32, #layout71> loc(#loc845) + %997 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x7x7>}> : (!tt.device<#device>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc846) + %998 = "ttnn.transpose"(%996, %997) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x7x512x7xf32, #layout71>, tensor<1x512x7x7xf32, #layout72>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc846) + %999 = "ttnn.to_layout"(%arg97, %0) <{layout = #ttnn.layout}> : (tensor<512x1x1xf32, #layout4>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc847) + %1000 = "ttnn.to_device"(%999, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<512x1x1xf32, #layout56>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc847) + %1001 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x7x7>}> : (!tt.device<#device>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc847) + %1002 = "ttnn.multiply"(%998, %1000, %1001) <{operandSegmentSizes = array}> : (tensor<1x512x7x7xf32, #layout72>, tensor<512x1x1xf32, #layout56>, tensor<1x512x7x7xf32, #layout72>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc847) + %1003 = "ttnn.to_layout"(%arg98, %0) <{layout = #ttnn.layout}> : (tensor<512x1x1xf32, #layout4>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc848) + %1004 = "ttnn.to_device"(%1003, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<512x1x1xf32, #layout56>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc848) + %1005 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x7x7>}> : (!tt.device<#device>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc848) + %1006 = "ttnn.add"(%1002, %1004, %1005) <{operandSegmentSizes = array}> : (tensor<1x512x7x7xf32, #layout72>, tensor<512x1x1xf32, #layout56>, tensor<1x512x7x7xf32, #layout72>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc848) + %1007 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x7x7>}> : (!tt.device<#device>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc849) + %1008 = "ttnn.relu"(%1006, %1007) <{operandSegmentSizes = array}> : (tensor<1x512x7x7xf32, #layout72>, tensor<1x512x7x7xf32, #layout72>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc849) + %1009 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x512x7>}> : (!tt.device<#device>) -> tensor<1x7x512x7xf32, #layout71> loc(#loc850) + %1010 = "ttnn.transpose"(%1008, %1009) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x512x7x7xf32, #layout72>, tensor<1x7x512x7xf32, #layout71>) -> tensor<1x7x512x7xf32, #layout71> loc(#loc850) + %1011 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x7x512>}> : (!tt.device<#device>) -> tensor<1x7x7x512xf32, #layout70> loc(#loc851) + %1012 = "ttnn.transpose"(%1010, %1011) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x7x512x7xf32, #layout71>, tensor<1x7x7x512xf32, #layout70>) -> tensor<1x7x7x512xf32, #layout70> loc(#loc851) + %1013 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x7x2048>}> : (!tt.device<#device>) -> tensor<1x7x7x2048xf32, #layout73> loc(#loc852) + %1014 = "ttnn.conv2d"(%1012, %arg156, %1013, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 512 : i32, input_height = 7 : i32, input_width = 7 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 2048 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x7x7x512xf32, #layout70>, tensor<2048x512x1x1xf32, #layout24>, tensor<1x7x7x2048xf32, #layout73>, !tt.device<#device>) -> tensor<1x7x7x2048xf32, #layout73> loc(#loc852) + %1015 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x2048x7>}> : (!tt.device<#device>) -> tensor<1x7x2048x7xf32, #layout74> loc(#loc853) + %1016 = "ttnn.transpose"(%1014, %1015) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x7x7x2048xf32, #layout73>, tensor<1x7x2048x7xf32, #layout74>) -> tensor<1x7x2048x7xf32, #layout74> loc(#loc853) + %1017 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x2048x7x7>}> : (!tt.device<#device>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc854) + %1018 = "ttnn.transpose"(%1016, %1017) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x7x2048x7xf32, #layout74>, tensor<1x2048x7x7xf32, #layout75>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc854) + %1019 = "ttnn.to_layout"(%arg99, %0) <{layout = #ttnn.layout}> : (tensor<2048x1x1xf32, #layout6>, !tt.device<#device>) -> tensor<2048x1x1xf32, #layout76> loc(#loc855) + %1020 = "ttnn.to_device"(%1019, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<2048x1x1xf32, #layout76>, !tt.device<#device>) -> tensor<2048x1x1xf32, #layout76> loc(#loc855) + %1021 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x2048x7x7>}> : (!tt.device<#device>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc855) + %1022 = "ttnn.multiply"(%1018, %1020, %1021) <{operandSegmentSizes = array}> : (tensor<1x2048x7x7xf32, #layout75>, tensor<2048x1x1xf32, #layout76>, tensor<1x2048x7x7xf32, #layout75>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc855) + %1023 = "ttnn.to_layout"(%arg100, %0) <{layout = #ttnn.layout}> : (tensor<2048x1x1xf32, #layout6>, !tt.device<#device>) -> tensor<2048x1x1xf32, #layout76> loc(#loc856) + %1024 = "ttnn.to_device"(%1023, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<2048x1x1xf32, #layout76>, !tt.device<#device>) -> tensor<2048x1x1xf32, #layout76> loc(#loc856) + %1025 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x2048x7x7>}> : (!tt.device<#device>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc856) + %1026 = "ttnn.add"(%1022, %1024, %1025) <{operandSegmentSizes = array}> : (tensor<1x2048x7x7xf32, #layout75>, tensor<2048x1x1xf32, #layout76>, tensor<1x2048x7x7xf32, #layout75>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc856) + %1027 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x2048x7x7>}> : (!tt.device<#device>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc857) + %1028 = "ttnn.add"(%1026, %968, %1027) <{operandSegmentSizes = array}> : (tensor<1x2048x7x7xf32, #layout75>, tensor<1x2048x7x7xf32, #layout75>, tensor<1x2048x7x7xf32, #layout75>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc857) + %1029 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x2048x7x7>}> : (!tt.device<#device>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc858) + %1030 = "ttnn.relu"(%1028, %1029) <{operandSegmentSizes = array}> : (tensor<1x2048x7x7xf32, #layout75>, tensor<1x2048x7x7xf32, #layout75>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc858) + %1031 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x2048x7>}> : (!tt.device<#device>) -> tensor<1x7x2048x7xf32, #layout74> loc(#loc859) + %1032 = "ttnn.transpose"(%1030, %1031) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x2048x7x7xf32, #layout75>, tensor<1x7x2048x7xf32, #layout74>) -> tensor<1x7x2048x7xf32, #layout74> loc(#loc859) + %1033 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x7x2048>}> : (!tt.device<#device>) -> tensor<1x7x7x2048xf32, #layout73> loc(#loc860) + %1034 = "ttnn.transpose"(%1032, %1033) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x7x2048x7xf32, #layout74>, tensor<1x7x7x2048xf32, #layout73>) -> tensor<1x7x7x2048xf32, #layout73> loc(#loc860) + %1035 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x7x512>}> : (!tt.device<#device>) -> tensor<1x7x7x512xf32, #layout70> loc(#loc861) + %1036 = "ttnn.conv2d"(%1034, %arg157, %1035, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 2048 : i32, input_height = 7 : i32, input_width = 7 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 512 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x7x7x2048xf32, #layout73>, tensor<512x2048x1x1xf32, #layout26>, tensor<1x7x7x512xf32, #layout70>, !tt.device<#device>) -> tensor<1x7x7x512xf32, #layout70> loc(#loc861) + %1037 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x512x7>}> : (!tt.device<#device>) -> tensor<1x7x512x7xf32, #layout71> loc(#loc862) + %1038 = "ttnn.transpose"(%1036, %1037) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x7x7x512xf32, #layout70>, tensor<1x7x512x7xf32, #layout71>) -> tensor<1x7x512x7xf32, #layout71> loc(#loc862) + %1039 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x7x7>}> : (!tt.device<#device>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc863) + %1040 = "ttnn.transpose"(%1038, %1039) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x7x512x7xf32, #layout71>, tensor<1x512x7x7xf32, #layout72>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc863) + %1041 = "ttnn.to_layout"(%arg101, %0) <{layout = #ttnn.layout}> : (tensor<512x1x1xf32, #layout4>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc864) + %1042 = "ttnn.to_device"(%1041, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<512x1x1xf32, #layout56>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc864) + %1043 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x7x7>}> : (!tt.device<#device>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc864) + %1044 = "ttnn.multiply"(%1040, %1042, %1043) <{operandSegmentSizes = array}> : (tensor<1x512x7x7xf32, #layout72>, tensor<512x1x1xf32, #layout56>, tensor<1x512x7x7xf32, #layout72>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc864) + %1045 = "ttnn.to_layout"(%arg102, %0) <{layout = #ttnn.layout}> : (tensor<512x1x1xf32, #layout4>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc865) + %1046 = "ttnn.to_device"(%1045, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<512x1x1xf32, #layout56>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc865) + %1047 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x7x7>}> : (!tt.device<#device>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc865) + %1048 = "ttnn.add"(%1044, %1046, %1047) <{operandSegmentSizes = array}> : (tensor<1x512x7x7xf32, #layout72>, tensor<512x1x1xf32, #layout56>, tensor<1x512x7x7xf32, #layout72>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc865) + %1049 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x7x7>}> : (!tt.device<#device>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc866) + %1050 = "ttnn.relu"(%1048, %1049) <{operandSegmentSizes = array}> : (tensor<1x512x7x7xf32, #layout72>, tensor<1x512x7x7xf32, #layout72>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc866) + %1051 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x512x7>}> : (!tt.device<#device>) -> tensor<1x7x512x7xf32, #layout71> loc(#loc867) + %1052 = "ttnn.transpose"(%1050, %1051) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x512x7x7xf32, #layout72>, tensor<1x7x512x7xf32, #layout71>) -> tensor<1x7x512x7xf32, #layout71> loc(#loc867) + %1053 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x7x512>}> : (!tt.device<#device>) -> tensor<1x7x7x512xf32, #layout70> loc(#loc868) + %1054 = "ttnn.transpose"(%1052, %1053) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x7x512x7xf32, #layout71>, tensor<1x7x7x512xf32, #layout70>) -> tensor<1x7x7x512xf32, #layout70> loc(#loc868) + %1055 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x7x512>}> : (!tt.device<#device>) -> tensor<1x7x7x512xf32, #layout70> loc(#loc869) + %1056 = "ttnn.conv2d"(%1054, %arg158, %1055, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 512 : i32, input_height = 7 : i32, input_width = 7 : i32, kernel_height = 3 : i32, kernel_width = 3 : i32, out_channels = 512 : i32, padding_height = 1 : i32, padding_width = 1 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x7x7x512xf32, #layout70>, tensor<512x512x3x3xf32, #layout23>, tensor<1x7x7x512xf32, #layout70>, !tt.device<#device>) -> tensor<1x7x7x512xf32, #layout70> loc(#loc869) + %1057 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x512x7>}> : (!tt.device<#device>) -> tensor<1x7x512x7xf32, #layout71> loc(#loc870) + %1058 = "ttnn.transpose"(%1056, %1057) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x7x7x512xf32, #layout70>, tensor<1x7x512x7xf32, #layout71>) -> tensor<1x7x512x7xf32, #layout71> loc(#loc870) + %1059 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x7x7>}> : (!tt.device<#device>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc871) + %1060 = "ttnn.transpose"(%1058, %1059) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x7x512x7xf32, #layout71>, tensor<1x512x7x7xf32, #layout72>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc871) + %1061 = "ttnn.to_layout"(%arg103, %0) <{layout = #ttnn.layout}> : (tensor<512x1x1xf32, #layout4>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc872) + %1062 = "ttnn.to_device"(%1061, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<512x1x1xf32, #layout56>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc872) + %1063 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x7x7>}> : (!tt.device<#device>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc872) + %1064 = "ttnn.multiply"(%1060, %1062, %1063) <{operandSegmentSizes = array}> : (tensor<1x512x7x7xf32, #layout72>, tensor<512x1x1xf32, #layout56>, tensor<1x512x7x7xf32, #layout72>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc872) + %1065 = "ttnn.to_layout"(%arg104, %0) <{layout = #ttnn.layout}> : (tensor<512x1x1xf32, #layout4>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc873) + %1066 = "ttnn.to_device"(%1065, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<512x1x1xf32, #layout56>, !tt.device<#device>) -> tensor<512x1x1xf32, #layout56> loc(#loc873) + %1067 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x7x7>}> : (!tt.device<#device>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc873) + %1068 = "ttnn.add"(%1064, %1066, %1067) <{operandSegmentSizes = array}> : (tensor<1x512x7x7xf32, #layout72>, tensor<512x1x1xf32, #layout56>, tensor<1x512x7x7xf32, #layout72>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc873) + %1069 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x512x7x7>}> : (!tt.device<#device>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc874) + %1070 = "ttnn.relu"(%1068, %1069) <{operandSegmentSizes = array}> : (tensor<1x512x7x7xf32, #layout72>, tensor<1x512x7x7xf32, #layout72>) -> tensor<1x512x7x7xf32, #layout72> loc(#loc874) + %1071 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x512x7>}> : (!tt.device<#device>) -> tensor<1x7x512x7xf32, #layout71> loc(#loc875) + %1072 = "ttnn.transpose"(%1070, %1071) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x512x7x7xf32, #layout72>, tensor<1x7x512x7xf32, #layout71>) -> tensor<1x7x512x7xf32, #layout71> loc(#loc875) + %1073 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x7x512>}> : (!tt.device<#device>) -> tensor<1x7x7x512xf32, #layout70> loc(#loc876) + %1074 = "ttnn.transpose"(%1072, %1073) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x7x512x7xf32, #layout71>, tensor<1x7x7x512xf32, #layout70>) -> tensor<1x7x7x512xf32, #layout70> loc(#loc876) + %1075 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x7x2048>}> : (!tt.device<#device>) -> tensor<1x7x7x2048xf32, #layout73> loc(#loc877) + %1076 = "ttnn.conv2d"(%1074, %arg159, %1075, %0) <{batch_size = 1 : i32, dilation_height = 1 : i32, dilation_width = 1 : i32, groups = 1 : i32, in_channels = 512 : i32, input_height = 7 : i32, input_width = 7 : i32, kernel_height = 1 : i32, kernel_width = 1 : i32, out_channels = 2048 : i32, padding_height = 0 : i32, padding_width = 0 : i32, stride_height = 1 : i32, stride_width = 1 : i32}> : (tensor<1x7x7x512xf32, #layout70>, tensor<2048x512x1x1xf32, #layout24>, tensor<1x7x7x2048xf32, #layout73>, !tt.device<#device>) -> tensor<1x7x7x2048xf32, #layout73> loc(#loc877) + %1077 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x7x2048x7>}> : (!tt.device<#device>) -> tensor<1x7x2048x7xf32, #layout74> loc(#loc878) + %1078 = "ttnn.transpose"(%1076, %1077) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x7x7x2048xf32, #layout73>, tensor<1x7x2048x7xf32, #layout74>) -> tensor<1x7x2048x7xf32, #layout74> loc(#loc878) + %1079 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x2048x7x7>}> : (!tt.device<#device>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc879) + %1080 = "ttnn.transpose"(%1078, %1079) <{dim0 = -3 : si32, dim1 = -2 : si32}> : (tensor<1x7x2048x7xf32, #layout74>, tensor<1x2048x7x7xf32, #layout75>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc879) + %1081 = "ttnn.to_layout"(%arg105, %0) <{layout = #ttnn.layout}> : (tensor<2048x1x1xf32, #layout6>, !tt.device<#device>) -> tensor<2048x1x1xf32, #layout76> loc(#loc880) + %1082 = "ttnn.to_device"(%1081, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<2048x1x1xf32, #layout76>, !tt.device<#device>) -> tensor<2048x1x1xf32, #layout76> loc(#loc880) + %1083 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x2048x7x7>}> : (!tt.device<#device>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc880) + %1084 = "ttnn.multiply"(%1080, %1082, %1083) <{operandSegmentSizes = array}> : (tensor<1x2048x7x7xf32, #layout75>, tensor<2048x1x1xf32, #layout76>, tensor<1x2048x7x7xf32, #layout75>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc880) + %1085 = "ttnn.to_layout"(%arg106, %0) <{layout = #ttnn.layout}> : (tensor<2048x1x1xf32, #layout6>, !tt.device<#device>) -> tensor<2048x1x1xf32, #layout76> loc(#loc881) + %1086 = "ttnn.to_device"(%1085, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<2048x1x1xf32, #layout76>, !tt.device<#device>) -> tensor<2048x1x1xf32, #layout76> loc(#loc881) + %1087 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x2048x7x7>}> : (!tt.device<#device>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc881) + %1088 = "ttnn.add"(%1084, %1086, %1087) <{operandSegmentSizes = array}> : (tensor<1x2048x7x7xf32, #layout75>, tensor<2048x1x1xf32, #layout76>, tensor<1x2048x7x7xf32, #layout75>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc881) + %1089 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x2048x7x7>}> : (!tt.device<#device>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc882) + %1090 = "ttnn.add"(%1088, %1030, %1089) <{operandSegmentSizes = array}> : (tensor<1x2048x7x7xf32, #layout75>, tensor<1x2048x7x7xf32, #layout75>, tensor<1x2048x7x7xf32, #layout75>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc882) + %1091 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x2048x7x7>}> : (!tt.device<#device>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc883) + %1092 = "ttnn.relu"(%1090, %1091) <{operandSegmentSizes = array}> : (tensor<1x2048x7x7xf32, #layout75>, tensor<1x2048x7x7xf32, #layout75>) -> tensor<1x2048x7x7xf32, #layout75> loc(#loc883) + %1093 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1x2048x49>}> : (!tt.device<#device>) -> tensor<1x1x2048x49xf32, #layout77> loc(#loc884) + %1094 = "ttnn.reshape"(%1092, %1093) <{shape = [1 : i32, 1 : i32, 2048 : i32, 49 : i32]}> : (tensor<1x2048x7x7xf32, #layout75>, tensor<1x1x2048x49xf32, #layout77>) -> tensor<1x1x2048x49xf32, #layout77> loc(#loc884) + %1095 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1x49x2048>}> : (!tt.device<#device>) -> tensor<1x1x49x2048xf32, #layout78> loc(#loc885) + %1096 = "ttnn.transpose"(%1094, %1095) <{dim0 = -2 : si32, dim1 = -1 : si32}> : (tensor<1x1x2048x49xf32, #layout77>, tensor<1x1x49x2048xf32, #layout78>) -> tensor<1x1x49x2048xf32, #layout78> loc(#loc885) + %1097 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1x1x2048>}> : (!tt.device<#device>) -> tensor<1x1x1x2048xf32, #layout79> loc(#loc886) + %1098 = "ttnn.mean"(%1096, %1097) <{keep_dim = true}> : (tensor<1x1x49x2048xf32, #layout78>, tensor<1x1x1x2048xf32, #layout79>) -> tensor<1x1x1x2048xf32, #layout79> loc(#loc886) + %1099 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x2048x1x1>}> : (!tt.device<#device>) -> tensor<1x2048x1x1xf32, #layout80> loc(#loc887) + %1100 = "ttnn.reshape"(%1098, %1099) <{shape = [1 : i32, 2048 : i32, 1 : i32, 1 : i32]}> : (tensor<1x1x1x2048xf32, #layout79>, tensor<1x2048x1x1xf32, #layout80>) -> tensor<1x2048x1x1xf32, #layout80> loc(#loc887) + %1101 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x2048x1>}> : (!tt.device<#device>) -> tensor<1x2048x1xf32, #layout81> loc(#loc888) + %1102 = "ttnn.reshape"(%1100, %1101) <{shape = [1 : i32, 2048 : i32, 1 : i32]}> : (tensor<1x2048x1x1xf32, #layout80>, tensor<1x2048x1xf32, #layout81>) -> tensor<1x2048x1xf32, #layout81> loc(#loc888) + %1103 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x2048>}> : (!tt.device<#device>) -> tensor<1x2048xf32, #layout82> loc(#loc889) + %1104 = "ttnn.reshape"(%1102, %1103) <{shape = [1 : i32, 2048 : i32]}> : (tensor<1x2048x1xf32, #layout81>, tensor<1x2048xf32, #layout82>) -> tensor<1x2048xf32, #layout82> loc(#loc889) + %1105 = "ttnn.to_layout"(%arg160, %0) <{layout = #ttnn.layout}> : (tensor<2048x1000xf32, #layout27>, !tt.device<#device>) -> tensor<2048x1000xf32, #layout83> loc(#loc890) + %1106 = "ttnn.to_device"(%1105, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<2048x1000xf32, #layout83>, !tt.device<#device>) -> tensor<2048x1000xf32, #layout83> loc(#loc890) + %1107 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1000>}> : (!tt.device<#device>) -> tensor<1x1000xf32, #layout84> loc(#loc890) + %1108 = "ttnn.matmul"(%1104, %1106, %1107) : (tensor<1x2048xf32, #layout82>, tensor<2048x1000xf32, #layout83>, tensor<1x1000xf32, #layout84>) -> tensor<1x1000xf32, #layout84> loc(#loc890) + %1109 = "ttnn.to_layout"(%arg161, %0) <{layout = #ttnn.layout}> : (tensor<1000xf32, #layout28>, !tt.device<#device>) -> tensor<1000xf32, #layout85> loc(#loc891) + %1110 = "ttnn.to_device"(%1109, %0) <{memory_config = #ttnn.memory_config<, >}> : (tensor<1000xf32, #layout85>, !tt.device<#device>) -> tensor<1000xf32, #layout85> loc(#loc891) + %1111 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes, layout = #ttnn.layout, memory_config = #ttnn.memory_config<, >, shape = #ttnn.shape<1x1000>}> : (!tt.device<#device>) -> tensor<1x1000xf32, #layout84> loc(#loc891) + %1112 = "ttnn.add"(%1108, %1110, %1111) <{operandSegmentSizes = array}> : (tensor<1x1000xf32, #layout84>, tensor<1000xf32, #layout85>, tensor<1x1000xf32, #layout84>) -> tensor<1x1000xf32, #layout84> loc(#loc891) + %1113 = "ttnn.to_memory_config"(%1112, %0) : (tensor<1x1000xf32, #layout84>, !tt.device<#device>) -> tensor<1x1000xf32, #layout29> loc(#loc446) + return %1113 : tensor<1x1000xf32, #layout29> loc(#loc446) + } loc(#loc) +} loc(#loc) +#loc1 = loc("forward":4294967295:2951) +#loc2 = loc("forward":4294967295:2952) +#loc3 = loc("forward":4294967295:2954) +#loc4 = loc("forward":4294967295:2955) +#loc5 = loc("forward":4294967295:2956) +#loc6 = loc("forward":4294967295:2958) +#loc7 = loc("forward":4294967295:2960) +#loc8 = loc("forward":4294967295:2961) +#loc9 = loc("forward":4294967295:2962) +#loc10 = loc("forward":4294967295:2963) +#loc11 = loc("forward":4294967295:2964) +#loc12 = loc("forward":4294967295:2966) +#loc13 = loc("forward":4294967295:2967) +#loc14 = loc("forward":4294967295:2968) +#loc15 = loc("forward":4294967295:2970) +#loc16 = loc("forward":4294967295:2972) +#loc17 = loc("forward":4294967295:2973) +#loc18 = loc("forward":4294967295:2974) +#loc19 = loc("forward":4294967295:2975) +#loc20 = loc("forward":4294967295:2977) +#loc21 = loc("forward":4294967295:2978) +#loc22 = loc("forward":4294967295:2979) +#loc23 = loc("forward":4294967295:2981) +#loc24 = loc("forward":4294967295:2983) +#loc25 = loc("forward":4294967295:2984) +#loc26 = loc("forward":4294967295:2985) +#loc27 = loc("forward":4294967295:2986) +#loc28 = loc("forward":4294967295:2988) +#loc29 = loc("forward":4294967295:2989) +#loc30 = loc("forward":4294967295:2990) +#loc31 = loc("forward":4294967295:2992) +#loc32 = loc("forward":4294967295:2994) +#loc33 = loc("forward":4294967295:2995) +#loc34 = loc("forward":4294967295:2996) +#loc35 = loc("forward":4294967295:2998) +#loc36 = loc("forward":4294967295:2999) +#loc37 = loc("forward":4294967295:3000) +#loc38 = loc("forward":4294967295:3002) +#loc39 = loc("forward":4294967295:3004) +#loc40 = loc("forward":4294967295:3005) +#loc41 = loc("forward":4294967295:3006) +#loc42 = loc("forward":4294967295:3007) +#loc43 = loc("forward":4294967295:3008) +#loc44 = loc("forward":4294967295:3010) +#loc45 = loc("forward":4294967295:3011) +#loc46 = loc("forward":4294967295:3012) +#loc47 = loc("forward":4294967295:3014) +#loc48 = loc("forward":4294967295:3016) +#loc49 = loc("forward":4294967295:3017) +#loc50 = loc("forward":4294967295:3018) +#loc51 = loc("forward":4294967295:3019) +#loc52 = loc("forward":4294967295:3021) +#loc53 = loc("forward":4294967295:3022) +#loc54 = loc("forward":4294967295:3023) +#loc55 = loc("forward":4294967295:3025) +#loc56 = loc("forward":4294967295:3027) +#loc57 = loc("forward":4294967295:3028) +#loc58 = loc("forward":4294967295:3029) +#loc59 = loc("forward":4294967295:3030) +#loc60 = loc("forward":4294967295:3032) +#loc61 = loc("forward":4294967295:3033) +#loc62 = loc("forward":4294967295:3034) +#loc63 = loc("forward":4294967295:3036) +#loc64 = loc("forward":4294967295:3038) +#loc65 = loc("forward":4294967295:3039) +#loc66 = loc("forward":4294967295:3040) +#loc67 = loc("forward":4294967295:3041) +#loc68 = loc("forward":4294967295:3042) +#loc69 = loc("forward":4294967295:3044) +#loc70 = loc("forward":4294967295:3045) +#loc71 = loc("forward":4294967295:3046) +#loc72 = loc("forward":4294967295:3048) +#loc73 = loc("forward":4294967295:3050) +#loc74 = loc("forward":4294967295:3051) +#loc75 = loc("forward":4294967295:3052) +#loc76 = loc("forward":4294967295:3053) +#loc77 = loc("forward":4294967295:3055) +#loc78 = loc("forward":4294967295:3056) +#loc79 = loc("forward":4294967295:3057) +#loc80 = loc("forward":4294967295:3059) +#loc81 = loc("forward":4294967295:3061) +#loc82 = loc("forward":4294967295:3062) +#loc83 = loc("forward":4294967295:3063) +#loc84 = loc("forward":4294967295:3064) +#loc85 = loc("forward":4294967295:3066) +#loc86 = loc("forward":4294967295:3067) +#loc87 = loc("forward":4294967295:3068) +#loc88 = loc("forward":4294967295:3070) +#loc89 = loc("forward":4294967295:3072) +#loc90 = loc("forward":4294967295:3073) +#loc91 = loc("forward":4294967295:3074) +#loc92 = loc("forward":4294967295:3075) +#loc93 = loc("forward":4294967295:3076) +#loc94 = loc("forward":4294967295:3078) +#loc95 = loc("forward":4294967295:3079) +#loc96 = loc("forward":4294967295:3080) +#loc97 = loc("forward":4294967295:3082) +#loc98 = loc("forward":4294967295:3084) +#loc99 = loc("forward":4294967295:3085) +#loc100 = loc("forward":4294967295:3086) +#loc101 = loc("forward":4294967295:3087) +#loc102 = loc("forward":4294967295:3089) +#loc103 = loc("forward":4294967295:3090) +#loc104 = loc("forward":4294967295:3091) +#loc105 = loc("forward":4294967295:3093) +#loc106 = loc("forward":4294967295:3095) +#loc107 = loc("forward":4294967295:3096) +#loc108 = loc("forward":4294967295:3097) +#loc109 = loc("forward":4294967295:3098) +#loc110 = loc("forward":4294967295:3100) +#loc111 = loc("forward":4294967295:3101) +#loc112 = loc("forward":4294967295:3102) +#loc113 = loc("forward":4294967295:3104) +#loc114 = loc("forward":4294967295:3106) +#loc115 = loc("forward":4294967295:3107) +#loc116 = loc("forward":4294967295:3108) +#loc117 = loc("forward":4294967295:3110) +#loc118 = loc("forward":4294967295:3111) +#loc119 = loc("forward":4294967295:3112) +#loc120 = loc("forward":4294967295:3114) +#loc121 = loc("forward":4294967295:3116) +#loc122 = loc("forward":4294967295:3117) +#loc123 = loc("forward":4294967295:3118) +#loc124 = loc("forward":4294967295:3119) +#loc125 = loc("forward":4294967295:3120) +#loc126 = loc("forward":4294967295:3122) +#loc127 = loc("forward":4294967295:3123) +#loc128 = loc("forward":4294967295:3124) +#loc129 = loc("forward":4294967295:3126) +#loc130 = loc("forward":4294967295:3128) +#loc131 = loc("forward":4294967295:3129) +#loc132 = loc("forward":4294967295:3130) +#loc133 = loc("forward":4294967295:3131) +#loc134 = loc("forward":4294967295:3133) +#loc135 = loc("forward":4294967295:3134) +#loc136 = loc("forward":4294967295:3135) +#loc137 = loc("forward":4294967295:3137) +#loc138 = loc("forward":4294967295:3139) +#loc139 = loc("forward":4294967295:3140) +#loc140 = loc("forward":4294967295:3141) +#loc141 = loc("forward":4294967295:3142) +#loc142 = loc("forward":4294967295:3144) +#loc143 = loc("forward":4294967295:3145) +#loc144 = loc("forward":4294967295:3146) +#loc145 = loc("forward":4294967295:3148) +#loc146 = loc("forward":4294967295:3150) +#loc147 = loc("forward":4294967295:3151) +#loc148 = loc("forward":4294967295:3152) +#loc149 = loc("forward":4294967295:3153) +#loc150 = loc("forward":4294967295:3154) +#loc151 = loc("forward":4294967295:3156) +#loc152 = loc("forward":4294967295:3157) +#loc153 = loc("forward":4294967295:3158) +#loc154 = loc("forward":4294967295:3160) +#loc155 = loc("forward":4294967295:3162) +#loc156 = loc("forward":4294967295:3163) +#loc157 = loc("forward":4294967295:3164) +#loc158 = loc("forward":4294967295:3165) +#loc159 = loc("forward":4294967295:3167) +#loc160 = loc("forward":4294967295:3168) +#loc161 = loc("forward":4294967295:3169) +#loc162 = loc("forward":4294967295:3171) +#loc163 = loc("forward":4294967295:3173) +#loc164 = loc("forward":4294967295:3174) +#loc165 = loc("forward":4294967295:3175) +#loc166 = loc("forward":4294967295:3176) +#loc167 = loc("forward":4294967295:3178) +#loc168 = loc("forward":4294967295:3179) +#loc169 = loc("forward":4294967295:3180) +#loc170 = loc("forward":4294967295:3182) +#loc171 = loc("forward":4294967295:3184) +#loc172 = loc("forward":4294967295:3185) +#loc173 = loc("forward":4294967295:3186) +#loc174 = loc("forward":4294967295:3187) +#loc175 = loc("forward":4294967295:3188) +#loc176 = loc("forward":4294967295:3190) +#loc177 = loc("forward":4294967295:3191) +#loc178 = loc("forward":4294967295:3192) +#loc179 = loc("forward":4294967295:3194) +#loc180 = loc("forward":4294967295:3196) +#loc181 = loc("forward":4294967295:3197) +#loc182 = loc("forward":4294967295:3198) +#loc183 = loc("forward":4294967295:3199) +#loc184 = loc("forward":4294967295:3201) +#loc185 = loc("forward":4294967295:3202) +#loc186 = loc("forward":4294967295:3203) +#loc187 = loc("forward":4294967295:3205) +#loc188 = loc("forward":4294967295:3207) +#loc189 = loc("forward":4294967295:3208) +#loc190 = loc("forward":4294967295:3209) +#loc191 = loc("forward":4294967295:3210) +#loc192 = loc("forward":4294967295:3212) +#loc193 = loc("forward":4294967295:3213) +#loc194 = loc("forward":4294967295:3214) +#loc195 = loc("forward":4294967295:3216) +#loc196 = loc("forward":4294967295:3218) +#loc197 = loc("forward":4294967295:3219) +#loc198 = loc("forward":4294967295:3220) +#loc199 = loc("forward":4294967295:3221) +#loc200 = loc("forward":4294967295:3222) +#loc201 = loc("forward":4294967295:3224) +#loc202 = loc("forward":4294967295:3225) +#loc203 = loc("forward":4294967295:3226) +#loc204 = loc("forward":4294967295:3228) +#loc205 = loc("forward":4294967295:3230) +#loc206 = loc("forward":4294967295:3231) +#loc207 = loc("forward":4294967295:3232) +#loc208 = loc("forward":4294967295:3233) +#loc209 = loc("forward":4294967295:3235) +#loc210 = loc("forward":4294967295:3236) +#loc211 = loc("forward":4294967295:3237) +#loc212 = loc("forward":4294967295:3239) +#loc213 = loc("forward":4294967295:3241) +#loc214 = loc("forward":4294967295:3242) +#loc215 = loc("forward":4294967295:3243) +#loc216 = loc("forward":4294967295:3244) +#loc217 = loc("forward":4294967295:3246) +#loc218 = loc("forward":4294967295:3247) +#loc219 = loc("forward":4294967295:3248) +#loc220 = loc("forward":4294967295:3250) +#loc221 = loc("forward":4294967295:3252) +#loc222 = loc("forward":4294967295:3253) +#loc223 = loc("forward":4294967295:3254) +#loc224 = loc("forward":4294967295:3256) +#loc225 = loc("forward":4294967295:3257) +#loc226 = loc("forward":4294967295:3258) +#loc227 = loc("forward":4294967295:3260) +#loc228 = loc("forward":4294967295:3262) +#loc229 = loc("forward":4294967295:3263) +#loc230 = loc("forward":4294967295:3264) +#loc231 = loc("forward":4294967295:3265) +#loc232 = loc("forward":4294967295:3266) +#loc233 = loc("forward":4294967295:3268) +#loc234 = loc("forward":4294967295:3269) +#loc235 = loc("forward":4294967295:3270) +#loc236 = loc("forward":4294967295:3272) +#loc237 = loc("forward":4294967295:3274) +#loc238 = loc("forward":4294967295:3275) +#loc239 = loc("forward":4294967295:3276) +#loc240 = loc("forward":4294967295:3277) +#loc241 = loc("forward":4294967295:3279) +#loc242 = loc("forward":4294967295:3280) +#loc243 = loc("forward":4294967295:3281) +#loc244 = loc("forward":4294967295:3283) +#loc245 = loc("forward":4294967295:3285) +#loc246 = loc("forward":4294967295:3286) +#loc247 = loc("forward":4294967295:3287) +#loc248 = loc("forward":4294967295:3288) +#loc249 = loc("forward":4294967295:3290) +#loc250 = loc("forward":4294967295:3291) +#loc251 = loc("forward":4294967295:3292) +#loc252 = loc("forward":4294967295:3294) +#loc253 = loc("forward":4294967295:3296) +#loc254 = loc("forward":4294967295:3297) +#loc255 = loc("forward":4294967295:3298) +#loc256 = loc("forward":4294967295:3299) +#loc257 = loc("forward":4294967295:3300) +#loc258 = loc("forward":4294967295:3302) +#loc259 = loc("forward":4294967295:3303) +#loc260 = loc("forward":4294967295:3304) +#loc261 = loc("forward":4294967295:3306) +#loc262 = loc("forward":4294967295:3308) +#loc263 = loc("forward":4294967295:3309) +#loc264 = loc("forward":4294967295:3310) +#loc265 = loc("forward":4294967295:3311) +#loc266 = loc("forward":4294967295:3313) +#loc267 = loc("forward":4294967295:3314) +#loc268 = loc("forward":4294967295:3315) +#loc269 = loc("forward":4294967295:3317) +#loc270 = loc("forward":4294967295:3319) +#loc271 = loc("forward":4294967295:3320) +#loc272 = loc("forward":4294967295:3321) +#loc273 = loc("forward":4294967295:3322) +#loc274 = loc("forward":4294967295:3324) +#loc275 = loc("forward":4294967295:3325) +#loc276 = loc("forward":4294967295:3326) +#loc277 = loc("forward":4294967295:3328) +#loc278 = loc("forward":4294967295:3330) +#loc279 = loc("forward":4294967295:3331) +#loc280 = loc("forward":4294967295:3332) +#loc281 = loc("forward":4294967295:3333) +#loc282 = loc("forward":4294967295:3334) +#loc283 = loc("forward":4294967295:3336) +#loc284 = loc("forward":4294967295:3337) +#loc285 = loc("forward":4294967295:3338) +#loc286 = loc("forward":4294967295:3340) +#loc287 = loc("forward":4294967295:3342) +#loc288 = loc("forward":4294967295:3343) +#loc289 = loc("forward":4294967295:3344) +#loc290 = loc("forward":4294967295:3345) +#loc291 = loc("forward":4294967295:3347) +#loc292 = loc("forward":4294967295:3348) +#loc293 = loc("forward":4294967295:3349) +#loc294 = loc("forward":4294967295:3351) +#loc295 = loc("forward":4294967295:3353) +#loc296 = loc("forward":4294967295:3354) +#loc297 = loc("forward":4294967295:3355) +#loc298 = loc("forward":4294967295:3356) +#loc299 = loc("forward":4294967295:3358) +#loc300 = loc("forward":4294967295:3359) +#loc301 = loc("forward":4294967295:3360) +#loc302 = loc("forward":4294967295:3362) +#loc303 = loc("forward":4294967295:3364) +#loc304 = loc("forward":4294967295:3365) +#loc305 = loc("forward":4294967295:3366) +#loc306 = loc("forward":4294967295:3367) +#loc307 = loc("forward":4294967295:3368) +#loc308 = loc("forward":4294967295:3370) +#loc309 = loc("forward":4294967295:3371) +#loc310 = loc("forward":4294967295:3372) +#loc311 = loc("forward":4294967295:3374) +#loc312 = loc("forward":4294967295:3376) +#loc313 = loc("forward":4294967295:3377) +#loc314 = loc("forward":4294967295:3378) +#loc315 = loc("forward":4294967295:3379) +#loc316 = loc("forward":4294967295:3381) +#loc317 = loc("forward":4294967295:3382) +#loc318 = loc("forward":4294967295:3383) +#loc319 = loc("forward":4294967295:3385) +#loc320 = loc("forward":4294967295:3387) +#loc321 = loc("forward":4294967295:3388) +#loc322 = loc("forward":4294967295:3389) +#loc323 = loc("forward":4294967295:3390) +#loc324 = loc("forward":4294967295:3392) +#loc325 = loc("forward":4294967295:3393) +#loc326 = loc("forward":4294967295:3394) +#loc327 = loc("forward":4294967295:3396) +#loc328 = loc("forward":4294967295:3398) +#loc329 = loc("forward":4294967295:3399) +#loc330 = loc("forward":4294967295:3400) +#loc331 = loc("forward":4294967295:3401) +#loc332 = loc("forward":4294967295:3402) +#loc333 = loc("forward":4294967295:3404) +#loc334 = loc("forward":4294967295:3405) +#loc335 = loc("forward":4294967295:3406) +#loc336 = loc("forward":4294967295:3408) +#loc337 = loc("forward":4294967295:3410) +#loc338 = loc("forward":4294967295:3411) +#loc339 = loc("forward":4294967295:3412) +#loc340 = loc("forward":4294967295:3413) +#loc341 = loc("forward":4294967295:3415) +#loc342 = loc("forward":4294967295:3416) +#loc343 = loc("forward":4294967295:3417) +#loc344 = loc("forward":4294967295:3419) +#loc345 = loc("forward":4294967295:3421) +#loc346 = loc("forward":4294967295:3422) +#loc347 = loc("forward":4294967295:3423) +#loc348 = loc("forward":4294967295:3424) +#loc349 = loc("forward":4294967295:3426) +#loc350 = loc("forward":4294967295:3427) +#loc351 = loc("forward":4294967295:3428) +#loc352 = loc("forward":4294967295:3430) +#loc353 = loc("forward":4294967295:3432) +#loc354 = loc("forward":4294967295:3433) +#loc355 = loc("forward":4294967295:3434) +#loc356 = loc("forward":4294967295:3435) +#loc357 = loc("forward":4294967295:3436) +#loc358 = loc("forward":4294967295:3438) +#loc359 = loc("forward":4294967295:3439) +#loc360 = loc("forward":4294967295:3440) +#loc361 = loc("forward":4294967295:3442) +#loc362 = loc("forward":4294967295:3444) +#loc363 = loc("forward":4294967295:3445) +#loc364 = loc("forward":4294967295:3446) +#loc365 = loc("forward":4294967295:3447) +#loc366 = loc("forward":4294967295:3449) +#loc367 = loc("forward":4294967295:3450) +#loc368 = loc("forward":4294967295:3451) +#loc369 = loc("forward":4294967295:3453) +#loc370 = loc("forward":4294967295:3455) +#loc371 = loc("forward":4294967295:3456) +#loc372 = loc("forward":4294967295:3457) +#loc373 = loc("forward":4294967295:3458) +#loc374 = loc("forward":4294967295:3460) +#loc375 = loc("forward":4294967295:3461) +#loc376 = loc("forward":4294967295:3462) +#loc377 = loc("forward":4294967295:3464) +#loc378 = loc("forward":4294967295:3466) +#loc379 = loc("forward":4294967295:3467) +#loc380 = loc("forward":4294967295:3468) +#loc381 = loc("forward":4294967295:3470) +#loc382 = loc("forward":4294967295:3471) +#loc383 = loc("forward":4294967295:3472) +#loc384 = loc("forward":4294967295:3474) +#loc385 = loc("forward":4294967295:3476) +#loc386 = loc("forward":4294967295:3477) +#loc387 = loc("forward":4294967295:3478) +#loc388 = loc("forward":4294967295:3479) +#loc389 = loc("forward":4294967295:3480) +#loc390 = loc("forward":4294967295:3482) +#loc391 = loc("forward":4294967295:3483) +#loc392 = loc("forward":4294967295:3484) +#loc393 = loc("forward":4294967295:3486) +#loc394 = loc("forward":4294967295:3488) +#loc395 = loc("forward":4294967295:3489) +#loc396 = loc("forward":4294967295:3490) +#loc397 = loc("forward":4294967295:3491) +#loc398 = loc("forward":4294967295:3493) +#loc399 = loc("forward":4294967295:3494) +#loc400 = loc("forward":4294967295:3495) +#loc401 = loc("forward":4294967295:3497) +#loc402 = loc("forward":4294967295:3499) +#loc403 = loc("forward":4294967295:3500) +#loc404 = loc("forward":4294967295:3501) +#loc405 = loc("forward":4294967295:3502) +#loc406 = loc("forward":4294967295:3504) +#loc407 = loc("forward":4294967295:3505) +#loc408 = loc("forward":4294967295:3506) +#loc409 = loc("forward":4294967295:3508) +#loc410 = loc("forward":4294967295:3510) +#loc411 = loc("forward":4294967295:3511) +#loc412 = loc("forward":4294967295:3512) +#loc413 = loc("forward":4294967295:3513) +#loc414 = loc("forward":4294967295:3514) +#loc415 = loc("forward":4294967295:3516) +#loc416 = loc("forward":4294967295:3517) +#loc417 = loc("forward":4294967295:3518) +#loc418 = loc("forward":4294967295:3520) +#loc419 = loc("forward":4294967295:3522) +#loc420 = loc("forward":4294967295:3523) +#loc421 = loc("forward":4294967295:3524) +#loc422 = loc("forward":4294967295:3525) +#loc423 = loc("forward":4294967295:3527) +#loc424 = loc("forward":4294967295:3528) +#loc425 = loc("forward":4294967295:3529) +#loc426 = loc("forward":4294967295:3531) +#loc427 = loc("forward":4294967295:3533) +#loc428 = loc("forward":4294967295:3534) +#loc429 = loc("forward":4294967295:3535) +#loc430 = loc("forward":4294967295:3536) +#loc431 = loc("forward":4294967295:3538) +#loc432 = loc("forward":4294967295:3539) +#loc433 = loc("forward":4294967295:3540) +#loc434 = loc("forward":4294967295:3542) +#loc435 = loc("forward":4294967295:3544) +#loc436 = loc("forward":4294967295:3545) +#loc437 = loc("forward":4294967295:3546) +#loc438 = loc("forward":4294967295:3547) +#loc439 = loc("forward":4294967295:3548) +#loc440 = loc("forward":4294967295:3549) +#loc441 = loc("forward":4294967295:3550) +#loc442 = loc("forward":4294967295:3551) +#loc443 = loc("forward":4294967295:3552) +#loc444 = loc("forward":4294967295:3554) +#loc445 = loc("forward":4294967295:3556) +#loc446 = loc(unknown) +#loc447 = loc("conv2d_0.dc.transpose.0"(#loc1)) +#loc448 = loc("conv2d_0.dc.transpose.1"(#loc2)) +#loc449 = loc("conv2d_0.dc.conv2d.2"(#loc3)) +#loc450 = loc("conv2d_0.dc.transpose.3"(#loc4)) +#loc451 = loc("conv2d_0.dc.transpose.4"(#loc5)) +#loc452 = loc("multiply_8"(#loc6)) +#loc453 = loc("add_14"(#loc7)) +#loc454 = loc("relu_15"(#loc8)) +#loc455 = loc("max_pool2d_16"(#loc9)) +#loc456 = loc("conv2d_17.dc.transpose.0"(#loc10)) +#loc457 = loc("conv2d_17.dc.transpose.1"(#loc11)) +#loc458 = loc("conv2d_17.dc.conv2d.2"(#loc12)) +#loc459 = loc("conv2d_17.dc.transpose.3"(#loc13)) +#loc460 = loc("conv2d_17.dc.transpose.4"(#loc14)) +#loc461 = loc("multiply_25"(#loc15)) +#loc462 = loc("add_31"(#loc16)) +#loc463 = loc("relu_32"(#loc17)) +#loc464 = loc("conv2d_33.dc.transpose.0"(#loc18)) +#loc465 = loc("conv2d_33.dc.transpose.1"(#loc19)) +#loc466 = loc("conv2d_33.dc.conv2d.2"(#loc20)) +#loc467 = loc("conv2d_33.dc.transpose.3"(#loc21)) +#loc468 = loc("conv2d_33.dc.transpose.4"(#loc22)) +#loc469 = loc("multiply_41"(#loc23)) +#loc470 = loc("add_47"(#loc24)) +#loc471 = loc("relu_48"(#loc25)) +#loc472 = loc("conv2d_49.dc.transpose.0"(#loc26)) +#loc473 = loc("conv2d_49.dc.transpose.1"(#loc27)) +#loc474 = loc("conv2d_49.dc.conv2d.2"(#loc28)) +#loc475 = loc("conv2d_49.dc.transpose.3"(#loc29)) +#loc476 = loc("conv2d_49.dc.transpose.4"(#loc30)) +#loc477 = loc("multiply_57"(#loc31)) +#loc478 = loc("add_63"(#loc32)) +#loc479 = loc("conv2d_64.dc.transpose.0"(#loc33)) +#loc480 = loc("conv2d_64.dc.transpose.1"(#loc34)) +#loc481 = loc("conv2d_64.dc.conv2d.2"(#loc35)) +#loc482 = loc("conv2d_64.dc.transpose.3"(#loc36)) +#loc483 = loc("conv2d_64.dc.transpose.4"(#loc37)) +#loc484 = loc("multiply_72"(#loc38)) +#loc485 = loc("add_78"(#loc39)) +#loc486 = loc("add_79"(#loc40)) +#loc487 = loc("relu_80"(#loc41)) +#loc488 = loc("conv2d_81.dc.transpose.0"(#loc42)) +#loc489 = loc("conv2d_81.dc.transpose.1"(#loc43)) +#loc490 = loc("conv2d_81.dc.conv2d.2"(#loc44)) +#loc491 = loc("conv2d_81.dc.transpose.3"(#loc45)) +#loc492 = loc("conv2d_81.dc.transpose.4"(#loc46)) +#loc493 = loc("multiply_89"(#loc47)) +#loc494 = loc("add_95"(#loc48)) +#loc495 = loc("relu_96"(#loc49)) +#loc496 = loc("conv2d_97.dc.transpose.0"(#loc50)) +#loc497 = loc("conv2d_97.dc.transpose.1"(#loc51)) +#loc498 = loc("conv2d_97.dc.conv2d.2"(#loc52)) +#loc499 = loc("conv2d_97.dc.transpose.3"(#loc53)) +#loc500 = loc("conv2d_97.dc.transpose.4"(#loc54)) +#loc501 = loc("multiply_105"(#loc55)) +#loc502 = loc("add_111"(#loc56)) +#loc503 = loc("relu_112"(#loc57)) +#loc504 = loc("conv2d_113.dc.transpose.0"(#loc58)) +#loc505 = loc("conv2d_113.dc.transpose.1"(#loc59)) +#loc506 = loc("conv2d_113.dc.conv2d.2"(#loc60)) +#loc507 = loc("conv2d_113.dc.transpose.3"(#loc61)) +#loc508 = loc("conv2d_113.dc.transpose.4"(#loc62)) +#loc509 = loc("multiply_121"(#loc63)) +#loc510 = loc("add_127"(#loc64)) +#loc511 = loc("add_128"(#loc65)) +#loc512 = loc("relu_129"(#loc66)) +#loc513 = loc("conv2d_130.dc.transpose.0"(#loc67)) +#loc514 = loc("conv2d_130.dc.transpose.1"(#loc68)) +#loc515 = loc("conv2d_130.dc.conv2d.2"(#loc69)) +#loc516 = loc("conv2d_130.dc.transpose.3"(#loc70)) +#loc517 = loc("conv2d_130.dc.transpose.4"(#loc71)) +#loc518 = loc("multiply_138"(#loc72)) +#loc519 = loc("add_144"(#loc73)) +#loc520 = loc("relu_145"(#loc74)) +#loc521 = loc("conv2d_146.dc.transpose.0"(#loc75)) +#loc522 = loc("conv2d_146.dc.transpose.1"(#loc76)) +#loc523 = loc("conv2d_146.dc.conv2d.2"(#loc77)) +#loc524 = loc("conv2d_146.dc.transpose.3"(#loc78)) +#loc525 = loc("conv2d_146.dc.transpose.4"(#loc79)) +#loc526 = loc("multiply_154"(#loc80)) +#loc527 = loc("add_160"(#loc81)) +#loc528 = loc("relu_161"(#loc82)) +#loc529 = loc("conv2d_162.dc.transpose.0"(#loc83)) +#loc530 = loc("conv2d_162.dc.transpose.1"(#loc84)) +#loc531 = loc("conv2d_162.dc.conv2d.2"(#loc85)) +#loc532 = loc("conv2d_162.dc.transpose.3"(#loc86)) +#loc533 = loc("conv2d_162.dc.transpose.4"(#loc87)) +#loc534 = loc("multiply_170"(#loc88)) +#loc535 = loc("add_176"(#loc89)) +#loc536 = loc("add_177"(#loc90)) +#loc537 = loc("relu_178"(#loc91)) +#loc538 = loc("conv2d_179.dc.transpose.0"(#loc92)) +#loc539 = loc("conv2d_179.dc.transpose.1"(#loc93)) +#loc540 = loc("conv2d_179.dc.conv2d.2"(#loc94)) +#loc541 = loc("conv2d_179.dc.transpose.3"(#loc95)) +#loc542 = loc("conv2d_179.dc.transpose.4"(#loc96)) +#loc543 = loc("multiply_187"(#loc97)) +#loc544 = loc("add_193"(#loc98)) +#loc545 = loc("relu_194"(#loc99)) +#loc546 = loc("conv2d_195.dc.transpose.0"(#loc100)) +#loc547 = loc("conv2d_195.dc.transpose.1"(#loc101)) +#loc548 = loc("conv2d_195.dc.conv2d.2"(#loc102)) +#loc549 = loc("conv2d_195.dc.transpose.3"(#loc103)) +#loc550 = loc("conv2d_195.dc.transpose.4"(#loc104)) +#loc551 = loc("multiply_203"(#loc105)) +#loc552 = loc("add_209"(#loc106)) +#loc553 = loc("relu_210"(#loc107)) +#loc554 = loc("conv2d_211.dc.transpose.0"(#loc108)) +#loc555 = loc("conv2d_211.dc.transpose.1"(#loc109)) +#loc556 = loc("conv2d_211.dc.conv2d.2"(#loc110)) +#loc557 = loc("conv2d_211.dc.transpose.3"(#loc111)) +#loc558 = loc("conv2d_211.dc.transpose.4"(#loc112)) +#loc559 = loc("multiply_219"(#loc113)) +#loc560 = loc("add_225"(#loc114)) +#loc561 = loc("conv2d_226.dc.transpose.0"(#loc115)) +#loc562 = loc("conv2d_226.dc.transpose.1"(#loc116)) +#loc563 = loc("conv2d_226.dc.conv2d.2"(#loc117)) +#loc564 = loc("conv2d_226.dc.transpose.3"(#loc118)) +#loc565 = loc("conv2d_226.dc.transpose.4"(#loc119)) +#loc566 = loc("multiply_234"(#loc120)) +#loc567 = loc("add_240"(#loc121)) +#loc568 = loc("add_241"(#loc122)) +#loc569 = loc("relu_242"(#loc123)) +#loc570 = loc("conv2d_243.dc.transpose.0"(#loc124)) +#loc571 = loc("conv2d_243.dc.transpose.1"(#loc125)) +#loc572 = loc("conv2d_243.dc.conv2d.2"(#loc126)) +#loc573 = loc("conv2d_243.dc.transpose.3"(#loc127)) +#loc574 = loc("conv2d_243.dc.transpose.4"(#loc128)) +#loc575 = loc("multiply_251"(#loc129)) +#loc576 = loc("add_257"(#loc130)) +#loc577 = loc("relu_258"(#loc131)) +#loc578 = loc("conv2d_259.dc.transpose.0"(#loc132)) +#loc579 = loc("conv2d_259.dc.transpose.1"(#loc133)) +#loc580 = loc("conv2d_259.dc.conv2d.2"(#loc134)) +#loc581 = loc("conv2d_259.dc.transpose.3"(#loc135)) +#loc582 = loc("conv2d_259.dc.transpose.4"(#loc136)) +#loc583 = loc("multiply_267"(#loc137)) +#loc584 = loc("add_273"(#loc138)) +#loc585 = loc("relu_274"(#loc139)) +#loc586 = loc("conv2d_275.dc.transpose.0"(#loc140)) +#loc587 = loc("conv2d_275.dc.transpose.1"(#loc141)) +#loc588 = loc("conv2d_275.dc.conv2d.2"(#loc142)) +#loc589 = loc("conv2d_275.dc.transpose.3"(#loc143)) +#loc590 = loc("conv2d_275.dc.transpose.4"(#loc144)) +#loc591 = loc("multiply_283"(#loc145)) +#loc592 = loc("add_289"(#loc146)) +#loc593 = loc("add_290"(#loc147)) +#loc594 = loc("relu_291"(#loc148)) +#loc595 = loc("conv2d_292.dc.transpose.0"(#loc149)) +#loc596 = loc("conv2d_292.dc.transpose.1"(#loc150)) +#loc597 = loc("conv2d_292.dc.conv2d.2"(#loc151)) +#loc598 = loc("conv2d_292.dc.transpose.3"(#loc152)) +#loc599 = loc("conv2d_292.dc.transpose.4"(#loc153)) +#loc600 = loc("multiply_300"(#loc154)) +#loc601 = loc("add_306"(#loc155)) +#loc602 = loc("relu_307"(#loc156)) +#loc603 = loc("conv2d_308.dc.transpose.0"(#loc157)) +#loc604 = loc("conv2d_308.dc.transpose.1"(#loc158)) +#loc605 = loc("conv2d_308.dc.conv2d.2"(#loc159)) +#loc606 = loc("conv2d_308.dc.transpose.3"(#loc160)) +#loc607 = loc("conv2d_308.dc.transpose.4"(#loc161)) +#loc608 = loc("multiply_316"(#loc162)) +#loc609 = loc("add_322"(#loc163)) +#loc610 = loc("relu_323"(#loc164)) +#loc611 = loc("conv2d_324.dc.transpose.0"(#loc165)) +#loc612 = loc("conv2d_324.dc.transpose.1"(#loc166)) +#loc613 = loc("conv2d_324.dc.conv2d.2"(#loc167)) +#loc614 = loc("conv2d_324.dc.transpose.3"(#loc168)) +#loc615 = loc("conv2d_324.dc.transpose.4"(#loc169)) +#loc616 = loc("multiply_332"(#loc170)) +#loc617 = loc("add_338"(#loc171)) +#loc618 = loc("add_339"(#loc172)) +#loc619 = loc("relu_340"(#loc173)) +#loc620 = loc("conv2d_341.dc.transpose.0"(#loc174)) +#loc621 = loc("conv2d_341.dc.transpose.1"(#loc175)) +#loc622 = loc("conv2d_341.dc.conv2d.2"(#loc176)) +#loc623 = loc("conv2d_341.dc.transpose.3"(#loc177)) +#loc624 = loc("conv2d_341.dc.transpose.4"(#loc178)) +#loc625 = loc("multiply_349"(#loc179)) +#loc626 = loc("add_355"(#loc180)) +#loc627 = loc("relu_356"(#loc181)) +#loc628 = loc("conv2d_357.dc.transpose.0"(#loc182)) +#loc629 = loc("conv2d_357.dc.transpose.1"(#loc183)) +#loc630 = loc("conv2d_357.dc.conv2d.2"(#loc184)) +#loc631 = loc("conv2d_357.dc.transpose.3"(#loc185)) +#loc632 = loc("conv2d_357.dc.transpose.4"(#loc186)) +#loc633 = loc("multiply_365"(#loc187)) +#loc634 = loc("add_371"(#loc188)) +#loc635 = loc("relu_372"(#loc189)) +#loc636 = loc("conv2d_373.dc.transpose.0"(#loc190)) +#loc637 = loc("conv2d_373.dc.transpose.1"(#loc191)) +#loc638 = loc("conv2d_373.dc.conv2d.2"(#loc192)) +#loc639 = loc("conv2d_373.dc.transpose.3"(#loc193)) +#loc640 = loc("conv2d_373.dc.transpose.4"(#loc194)) +#loc641 = loc("multiply_381"(#loc195)) +#loc642 = loc("add_387"(#loc196)) +#loc643 = loc("add_388"(#loc197)) +#loc644 = loc("relu_389"(#loc198)) +#loc645 = loc("conv2d_390.dc.transpose.0"(#loc199)) +#loc646 = loc("conv2d_390.dc.transpose.1"(#loc200)) +#loc647 = loc("conv2d_390.dc.conv2d.2"(#loc201)) +#loc648 = loc("conv2d_390.dc.transpose.3"(#loc202)) +#loc649 = loc("conv2d_390.dc.transpose.4"(#loc203)) +#loc650 = loc("multiply_398"(#loc204)) +#loc651 = loc("add_404"(#loc205)) +#loc652 = loc("relu_405"(#loc206)) +#loc653 = loc("conv2d_406.dc.transpose.0"(#loc207)) +#loc654 = loc("conv2d_406.dc.transpose.1"(#loc208)) +#loc655 = loc("conv2d_406.dc.conv2d.2"(#loc209)) +#loc656 = loc("conv2d_406.dc.transpose.3"(#loc210)) +#loc657 = loc("conv2d_406.dc.transpose.4"(#loc211)) +#loc658 = loc("multiply_414"(#loc212)) +#loc659 = loc("add_420"(#loc213)) +#loc660 = loc("relu_421"(#loc214)) +#loc661 = loc("conv2d_422.dc.transpose.0"(#loc215)) +#loc662 = loc("conv2d_422.dc.transpose.1"(#loc216)) +#loc663 = loc("conv2d_422.dc.conv2d.2"(#loc217)) +#loc664 = loc("conv2d_422.dc.transpose.3"(#loc218)) +#loc665 = loc("conv2d_422.dc.transpose.4"(#loc219)) +#loc666 = loc("multiply_430"(#loc220)) +#loc667 = loc("add_436"(#loc221)) +#loc668 = loc("conv2d_437.dc.transpose.0"(#loc222)) +#loc669 = loc("conv2d_437.dc.transpose.1"(#loc223)) +#loc670 = loc("conv2d_437.dc.conv2d.2"(#loc224)) +#loc671 = loc("conv2d_437.dc.transpose.3"(#loc225)) +#loc672 = loc("conv2d_437.dc.transpose.4"(#loc226)) +#loc673 = loc("multiply_445"(#loc227)) +#loc674 = loc("add_451"(#loc228)) +#loc675 = loc("add_452"(#loc229)) +#loc676 = loc("relu_453"(#loc230)) +#loc677 = loc("conv2d_454.dc.transpose.0"(#loc231)) +#loc678 = loc("conv2d_454.dc.transpose.1"(#loc232)) +#loc679 = loc("conv2d_454.dc.conv2d.2"(#loc233)) +#loc680 = loc("conv2d_454.dc.transpose.3"(#loc234)) +#loc681 = loc("conv2d_454.dc.transpose.4"(#loc235)) +#loc682 = loc("multiply_462"(#loc236)) +#loc683 = loc("add_468"(#loc237)) +#loc684 = loc("relu_469"(#loc238)) +#loc685 = loc("conv2d_470.dc.transpose.0"(#loc239)) +#loc686 = loc("conv2d_470.dc.transpose.1"(#loc240)) +#loc687 = loc("conv2d_470.dc.conv2d.2"(#loc241)) +#loc688 = loc("conv2d_470.dc.transpose.3"(#loc242)) +#loc689 = loc("conv2d_470.dc.transpose.4"(#loc243)) +#loc690 = loc("multiply_478"(#loc244)) +#loc691 = loc("add_484"(#loc245)) +#loc692 = loc("relu_485"(#loc246)) +#loc693 = loc("conv2d_486.dc.transpose.0"(#loc247)) +#loc694 = loc("conv2d_486.dc.transpose.1"(#loc248)) +#loc695 = loc("conv2d_486.dc.conv2d.2"(#loc249)) +#loc696 = loc("conv2d_486.dc.transpose.3"(#loc250)) +#loc697 = loc("conv2d_486.dc.transpose.4"(#loc251)) +#loc698 = loc("multiply_494"(#loc252)) +#loc699 = loc("add_500"(#loc253)) +#loc700 = loc("add_501"(#loc254)) +#loc701 = loc("relu_502"(#loc255)) +#loc702 = loc("conv2d_503.dc.transpose.0"(#loc256)) +#loc703 = loc("conv2d_503.dc.transpose.1"(#loc257)) +#loc704 = loc("conv2d_503.dc.conv2d.2"(#loc258)) +#loc705 = loc("conv2d_503.dc.transpose.3"(#loc259)) +#loc706 = loc("conv2d_503.dc.transpose.4"(#loc260)) +#loc707 = loc("multiply_511"(#loc261)) +#loc708 = loc("add_517"(#loc262)) +#loc709 = loc("relu_518"(#loc263)) +#loc710 = loc("conv2d_519.dc.transpose.0"(#loc264)) +#loc711 = loc("conv2d_519.dc.transpose.1"(#loc265)) +#loc712 = loc("conv2d_519.dc.conv2d.2"(#loc266)) +#loc713 = loc("conv2d_519.dc.transpose.3"(#loc267)) +#loc714 = loc("conv2d_519.dc.transpose.4"(#loc268)) +#loc715 = loc("multiply_527"(#loc269)) +#loc716 = loc("add_533"(#loc270)) +#loc717 = loc("relu_534"(#loc271)) +#loc718 = loc("conv2d_535.dc.transpose.0"(#loc272)) +#loc719 = loc("conv2d_535.dc.transpose.1"(#loc273)) +#loc720 = loc("conv2d_535.dc.conv2d.2"(#loc274)) +#loc721 = loc("conv2d_535.dc.transpose.3"(#loc275)) +#loc722 = loc("conv2d_535.dc.transpose.4"(#loc276)) +#loc723 = loc("multiply_543"(#loc277)) +#loc724 = loc("add_549"(#loc278)) +#loc725 = loc("add_550"(#loc279)) +#loc726 = loc("relu_551"(#loc280)) +#loc727 = loc("conv2d_552.dc.transpose.0"(#loc281)) +#loc728 = loc("conv2d_552.dc.transpose.1"(#loc282)) +#loc729 = loc("conv2d_552.dc.conv2d.2"(#loc283)) +#loc730 = loc("conv2d_552.dc.transpose.3"(#loc284)) +#loc731 = loc("conv2d_552.dc.transpose.4"(#loc285)) +#loc732 = loc("multiply_560"(#loc286)) +#loc733 = loc("add_566"(#loc287)) +#loc734 = loc("relu_567"(#loc288)) +#loc735 = loc("conv2d_568.dc.transpose.0"(#loc289)) +#loc736 = loc("conv2d_568.dc.transpose.1"(#loc290)) +#loc737 = loc("conv2d_568.dc.conv2d.2"(#loc291)) +#loc738 = loc("conv2d_568.dc.transpose.3"(#loc292)) +#loc739 = loc("conv2d_568.dc.transpose.4"(#loc293)) +#loc740 = loc("multiply_576"(#loc294)) +#loc741 = loc("add_582"(#loc295)) +#loc742 = loc("relu_583"(#loc296)) +#loc743 = loc("conv2d_584.dc.transpose.0"(#loc297)) +#loc744 = loc("conv2d_584.dc.transpose.1"(#loc298)) +#loc745 = loc("conv2d_584.dc.conv2d.2"(#loc299)) +#loc746 = loc("conv2d_584.dc.transpose.3"(#loc300)) +#loc747 = loc("conv2d_584.dc.transpose.4"(#loc301)) +#loc748 = loc("multiply_592"(#loc302)) +#loc749 = loc("add_598"(#loc303)) +#loc750 = loc("add_599"(#loc304)) +#loc751 = loc("relu_600"(#loc305)) +#loc752 = loc("conv2d_601.dc.transpose.0"(#loc306)) +#loc753 = loc("conv2d_601.dc.transpose.1"(#loc307)) +#loc754 = loc("conv2d_601.dc.conv2d.2"(#loc308)) +#loc755 = loc("conv2d_601.dc.transpose.3"(#loc309)) +#loc756 = loc("conv2d_601.dc.transpose.4"(#loc310)) +#loc757 = loc("multiply_609"(#loc311)) +#loc758 = loc("add_615"(#loc312)) +#loc759 = loc("relu_616"(#loc313)) +#loc760 = loc("conv2d_617.dc.transpose.0"(#loc314)) +#loc761 = loc("conv2d_617.dc.transpose.1"(#loc315)) +#loc762 = loc("conv2d_617.dc.conv2d.2"(#loc316)) +#loc763 = loc("conv2d_617.dc.transpose.3"(#loc317)) +#loc764 = loc("conv2d_617.dc.transpose.4"(#loc318)) +#loc765 = loc("multiply_625"(#loc319)) +#loc766 = loc("add_631"(#loc320)) +#loc767 = loc("relu_632"(#loc321)) +#loc768 = loc("conv2d_633.dc.transpose.0"(#loc322)) +#loc769 = loc("conv2d_633.dc.transpose.1"(#loc323)) +#loc770 = loc("conv2d_633.dc.conv2d.2"(#loc324)) +#loc771 = loc("conv2d_633.dc.transpose.3"(#loc325)) +#loc772 = loc("conv2d_633.dc.transpose.4"(#loc326)) +#loc773 = loc("multiply_641"(#loc327)) +#loc774 = loc("add_647"(#loc328)) +#loc775 = loc("add_648"(#loc329)) +#loc776 = loc("relu_649"(#loc330)) +#loc777 = loc("conv2d_650.dc.transpose.0"(#loc331)) +#loc778 = loc("conv2d_650.dc.transpose.1"(#loc332)) +#loc779 = loc("conv2d_650.dc.conv2d.2"(#loc333)) +#loc780 = loc("conv2d_650.dc.transpose.3"(#loc334)) +#loc781 = loc("conv2d_650.dc.transpose.4"(#loc335)) +#loc782 = loc("multiply_658"(#loc336)) +#loc783 = loc("add_664"(#loc337)) +#loc784 = loc("relu_665"(#loc338)) +#loc785 = loc("conv2d_666.dc.transpose.0"(#loc339)) +#loc786 = loc("conv2d_666.dc.transpose.1"(#loc340)) +#loc787 = loc("conv2d_666.dc.conv2d.2"(#loc341)) +#loc788 = loc("conv2d_666.dc.transpose.3"(#loc342)) +#loc789 = loc("conv2d_666.dc.transpose.4"(#loc343)) +#loc790 = loc("multiply_674"(#loc344)) +#loc791 = loc("add_680"(#loc345)) +#loc792 = loc("relu_681"(#loc346)) +#loc793 = loc("conv2d_682.dc.transpose.0"(#loc347)) +#loc794 = loc("conv2d_682.dc.transpose.1"(#loc348)) +#loc795 = loc("conv2d_682.dc.conv2d.2"(#loc349)) +#loc796 = loc("conv2d_682.dc.transpose.3"(#loc350)) +#loc797 = loc("conv2d_682.dc.transpose.4"(#loc351)) +#loc798 = loc("multiply_690"(#loc352)) +#loc799 = loc("add_696"(#loc353)) +#loc800 = loc("add_697"(#loc354)) +#loc801 = loc("relu_698"(#loc355)) +#loc802 = loc("conv2d_699.dc.transpose.0"(#loc356)) +#loc803 = loc("conv2d_699.dc.transpose.1"(#loc357)) +#loc804 = loc("conv2d_699.dc.conv2d.2"(#loc358)) +#loc805 = loc("conv2d_699.dc.transpose.3"(#loc359)) +#loc806 = loc("conv2d_699.dc.transpose.4"(#loc360)) +#loc807 = loc("multiply_707"(#loc361)) +#loc808 = loc("add_713"(#loc362)) +#loc809 = loc("relu_714"(#loc363)) +#loc810 = loc("conv2d_715.dc.transpose.0"(#loc364)) +#loc811 = loc("conv2d_715.dc.transpose.1"(#loc365)) +#loc812 = loc("conv2d_715.dc.conv2d.2"(#loc366)) +#loc813 = loc("conv2d_715.dc.transpose.3"(#loc367)) +#loc814 = loc("conv2d_715.dc.transpose.4"(#loc368)) +#loc815 = loc("multiply_723"(#loc369)) +#loc816 = loc("add_729"(#loc370)) +#loc817 = loc("relu_730"(#loc371)) +#loc818 = loc("conv2d_731.dc.transpose.0"(#loc372)) +#loc819 = loc("conv2d_731.dc.transpose.1"(#loc373)) +#loc820 = loc("conv2d_731.dc.conv2d.2"(#loc374)) +#loc821 = loc("conv2d_731.dc.transpose.3"(#loc375)) +#loc822 = loc("conv2d_731.dc.transpose.4"(#loc376)) +#loc823 = loc("multiply_739"(#loc377)) +#loc824 = loc("add_745"(#loc378)) +#loc825 = loc("conv2d_746.dc.transpose.0"(#loc379)) +#loc826 = loc("conv2d_746.dc.transpose.1"(#loc380)) +#loc827 = loc("conv2d_746.dc.conv2d.2"(#loc381)) +#loc828 = loc("conv2d_746.dc.transpose.3"(#loc382)) +#loc829 = loc("conv2d_746.dc.transpose.4"(#loc383)) +#loc830 = loc("multiply_754"(#loc384)) +#loc831 = loc("add_760"(#loc385)) +#loc832 = loc("add_761"(#loc386)) +#loc833 = loc("relu_762"(#loc387)) +#loc834 = loc("conv2d_763.dc.transpose.0"(#loc388)) +#loc835 = loc("conv2d_763.dc.transpose.1"(#loc389)) +#loc836 = loc("conv2d_763.dc.conv2d.2"(#loc390)) +#loc837 = loc("conv2d_763.dc.transpose.3"(#loc391)) +#loc838 = loc("conv2d_763.dc.transpose.4"(#loc392)) +#loc839 = loc("multiply_771"(#loc393)) +#loc840 = loc("add_777"(#loc394)) +#loc841 = loc("relu_778"(#loc395)) +#loc842 = loc("conv2d_779.dc.transpose.0"(#loc396)) +#loc843 = loc("conv2d_779.dc.transpose.1"(#loc397)) +#loc844 = loc("conv2d_779.dc.conv2d.2"(#loc398)) +#loc845 = loc("conv2d_779.dc.transpose.3"(#loc399)) +#loc846 = loc("conv2d_779.dc.transpose.4"(#loc400)) +#loc847 = loc("multiply_787"(#loc401)) +#loc848 = loc("add_793"(#loc402)) +#loc849 = loc("relu_794"(#loc403)) +#loc850 = loc("conv2d_795.dc.transpose.0"(#loc404)) +#loc851 = loc("conv2d_795.dc.transpose.1"(#loc405)) +#loc852 = loc("conv2d_795.dc.conv2d.2"(#loc406)) +#loc853 = loc("conv2d_795.dc.transpose.3"(#loc407)) +#loc854 = loc("conv2d_795.dc.transpose.4"(#loc408)) +#loc855 = loc("multiply_803"(#loc409)) +#loc856 = loc("add_809"(#loc410)) +#loc857 = loc("add_810"(#loc411)) +#loc858 = loc("relu_811"(#loc412)) +#loc859 = loc("conv2d_812.dc.transpose.0"(#loc413)) +#loc860 = loc("conv2d_812.dc.transpose.1"(#loc414)) +#loc861 = loc("conv2d_812.dc.conv2d.2"(#loc415)) +#loc862 = loc("conv2d_812.dc.transpose.3"(#loc416)) +#loc863 = loc("conv2d_812.dc.transpose.4"(#loc417)) +#loc864 = loc("multiply_820"(#loc418)) +#loc865 = loc("add_826"(#loc419)) +#loc866 = loc("relu_827"(#loc420)) +#loc867 = loc("conv2d_828.dc.transpose.0"(#loc421)) +#loc868 = loc("conv2d_828.dc.transpose.1"(#loc422)) +#loc869 = loc("conv2d_828.dc.conv2d.2"(#loc423)) +#loc870 = loc("conv2d_828.dc.transpose.3"(#loc424)) +#loc871 = loc("conv2d_828.dc.transpose.4"(#loc425)) +#loc872 = loc("multiply_836"(#loc426)) +#loc873 = loc("add_842"(#loc427)) +#loc874 = loc("relu_843"(#loc428)) +#loc875 = loc("conv2d_844.dc.transpose.0"(#loc429)) +#loc876 = loc("conv2d_844.dc.transpose.1"(#loc430)) +#loc877 = loc("conv2d_844.dc.conv2d.2"(#loc431)) +#loc878 = loc("conv2d_844.dc.transpose.3"(#loc432)) +#loc879 = loc("conv2d_844.dc.transpose.4"(#loc433)) +#loc880 = loc("multiply_852"(#loc434)) +#loc881 = loc("add_858"(#loc435)) +#loc882 = loc("add_859"(#loc436)) +#loc883 = loc("relu_860"(#loc437)) +#loc884 = loc("avg_pool2d_861.dc.reshape.0"(#loc438)) +#loc885 = loc("avg_pool2d_861.dc.transpose.1.dc.transpose.0"(#loc439)) +#loc886 = loc("avg_pool2d_861.dc.reduce_avg.2"(#loc440)) +#loc887 = loc("avg_pool2d_861.dc.reshape.4"(#loc441)) +#loc888 = loc("squeeze_863"(#loc442)) +#loc889 = loc("squeeze_864"(#loc443)) +#loc890 = loc("matmul_866"(#loc444)) +#loc891 = loc("add_867"(#loc445)) diff --git a/tools/explorer/test/models/test_10k_ops.mlir b/tools/explorer/test/models/test_10k_ops.mlir new file mode 100644 index 000000000..0c94560df --- /dev/null +++ b/tools/explorer/test/models/test_10k_ops.mlir @@ -0,0 +1,10005 @@ +module @Test10k attributes {} { + func.func @forward(%arg0: tensor<1x64xf32> {ttir.name = "input_1"}) -> (tensor<1x64xf32> {ttir.name = "TEST10k"}) { + %0 = tensor.empty() : tensor<1x64xf32> + %1 = "ttir.relu"(%arg0, %0) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2 = tensor.empty() : tensor<1x64xf32> + %3 = "ttir.relu"(%1, %2) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4 = tensor.empty() : tensor<1x64xf32> + %5 = "ttir.relu"(%3, %4) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6 = tensor.empty() : tensor<1x64xf32> + %7 = "ttir.relu"(%5, %6) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8 = tensor.empty() : tensor<1x64xf32> + %9 = "ttir.relu"(%7, %8) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %10 = tensor.empty() : tensor<1x64xf32> + %11 = "ttir.relu"(%9, %10) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %12 = tensor.empty() : tensor<1x64xf32> + %13 = "ttir.relu"(%11, %12) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %14 = tensor.empty() : tensor<1x64xf32> + %15 = "ttir.relu"(%13, %14) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %16 = tensor.empty() : tensor<1x64xf32> + %17 = "ttir.relu"(%15, %16) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %18 = tensor.empty() : tensor<1x64xf32> + %19 = "ttir.relu"(%17, %18) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %20 = tensor.empty() : tensor<1x64xf32> + %21 = "ttir.relu"(%19, %20) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %22 = tensor.empty() : tensor<1x64xf32> + %23 = "ttir.relu"(%21, %22) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %24 = tensor.empty() : tensor<1x64xf32> + %25 = "ttir.relu"(%23, %24) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %26 = tensor.empty() : tensor<1x64xf32> + %27 = "ttir.relu"(%25, %26) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %28 = tensor.empty() : tensor<1x64xf32> + %29 = "ttir.relu"(%27, %28) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %30 = tensor.empty() : tensor<1x64xf32> + %31 = "ttir.relu"(%29, %30) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %32 = tensor.empty() : tensor<1x64xf32> + %33 = "ttir.relu"(%31, %32) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %34 = tensor.empty() : tensor<1x64xf32> + %35 = "ttir.relu"(%33, %34) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %36 = tensor.empty() : tensor<1x64xf32> + %37 = "ttir.relu"(%35, %36) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %38 = tensor.empty() : tensor<1x64xf32> + %39 = "ttir.relu"(%37, %38) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %40 = tensor.empty() : tensor<1x64xf32> + %41 = "ttir.relu"(%39, %40) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %42 = tensor.empty() : tensor<1x64xf32> + %43 = "ttir.relu"(%41, %42) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %44 = tensor.empty() : tensor<1x64xf32> + %45 = "ttir.relu"(%43, %44) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %46 = tensor.empty() : tensor<1x64xf32> + %47 = "ttir.relu"(%45, %46) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %48 = tensor.empty() : tensor<1x64xf32> + %49 = "ttir.relu"(%47, %48) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %50 = tensor.empty() : tensor<1x64xf32> + %51 = "ttir.relu"(%49, %50) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %52 = tensor.empty() : tensor<1x64xf32> + %53 = "ttir.relu"(%51, %52) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %54 = tensor.empty() : tensor<1x64xf32> + %55 = "ttir.relu"(%53, %54) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %56 = tensor.empty() : tensor<1x64xf32> + %57 = "ttir.relu"(%55, %56) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %58 = tensor.empty() : tensor<1x64xf32> + %59 = "ttir.relu"(%57, %58) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %60 = tensor.empty() : tensor<1x64xf32> + %61 = "ttir.relu"(%59, %60) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %62 = tensor.empty() : tensor<1x64xf32> + %63 = "ttir.relu"(%61, %62) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %64 = tensor.empty() : tensor<1x64xf32> + %65 = "ttir.relu"(%63, %64) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %66 = tensor.empty() : tensor<1x64xf32> + %67 = "ttir.relu"(%65, %66) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %68 = tensor.empty() : tensor<1x64xf32> + %69 = "ttir.relu"(%67, %68) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %70 = tensor.empty() : tensor<1x64xf32> + %71 = "ttir.relu"(%69, %70) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %72 = tensor.empty() : tensor<1x64xf32> + %73 = "ttir.relu"(%71, %72) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %74 = tensor.empty() : tensor<1x64xf32> + %75 = "ttir.relu"(%73, %74) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %76 = tensor.empty() : tensor<1x64xf32> + %77 = "ttir.relu"(%75, %76) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %78 = tensor.empty() : tensor<1x64xf32> + %79 = "ttir.relu"(%77, %78) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %80 = tensor.empty() : tensor<1x64xf32> + %81 = "ttir.relu"(%79, %80) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %82 = tensor.empty() : tensor<1x64xf32> + %83 = "ttir.relu"(%81, %82) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %84 = tensor.empty() : tensor<1x64xf32> + %85 = "ttir.relu"(%83, %84) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %86 = tensor.empty() : tensor<1x64xf32> + %87 = "ttir.relu"(%85, %86) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %88 = tensor.empty() : tensor<1x64xf32> + %89 = "ttir.relu"(%87, %88) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %90 = tensor.empty() : tensor<1x64xf32> + %91 = "ttir.relu"(%89, %90) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %92 = tensor.empty() : tensor<1x64xf32> + %93 = "ttir.relu"(%91, %92) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %94 = tensor.empty() : tensor<1x64xf32> + %95 = "ttir.relu"(%93, %94) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %96 = tensor.empty() : tensor<1x64xf32> + %97 = "ttir.relu"(%95, %96) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %98 = tensor.empty() : tensor<1x64xf32> + %99 = "ttir.relu"(%97, %98) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %100 = tensor.empty() : tensor<1x64xf32> + %101 = "ttir.relu"(%99, %100) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %102 = tensor.empty() : tensor<1x64xf32> + %103 = "ttir.relu"(%101, %102) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %104 = tensor.empty() : tensor<1x64xf32> + %105 = "ttir.relu"(%103, %104) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %106 = tensor.empty() : tensor<1x64xf32> + %107 = "ttir.relu"(%105, %106) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %108 = tensor.empty() : tensor<1x64xf32> + %109 = "ttir.relu"(%107, %108) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %110 = tensor.empty() : tensor<1x64xf32> + %111 = "ttir.relu"(%109, %110) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %112 = tensor.empty() : tensor<1x64xf32> + %113 = "ttir.relu"(%111, %112) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %114 = tensor.empty() : tensor<1x64xf32> + %115 = "ttir.relu"(%113, %114) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %116 = tensor.empty() : tensor<1x64xf32> + %117 = "ttir.relu"(%115, %116) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %118 = tensor.empty() : tensor<1x64xf32> + %119 = "ttir.relu"(%117, %118) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %120 = tensor.empty() : tensor<1x64xf32> + %121 = "ttir.relu"(%119, %120) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %122 = tensor.empty() : tensor<1x64xf32> + %123 = "ttir.relu"(%121, %122) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %124 = tensor.empty() : tensor<1x64xf32> + %125 = "ttir.relu"(%123, %124) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %126 = tensor.empty() : tensor<1x64xf32> + %127 = "ttir.relu"(%125, %126) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %128 = tensor.empty() : tensor<1x64xf32> + %129 = "ttir.relu"(%127, %128) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %130 = tensor.empty() : tensor<1x64xf32> + %131 = "ttir.relu"(%129, %130) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %132 = tensor.empty() : tensor<1x64xf32> + %133 = "ttir.relu"(%131, %132) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %134 = tensor.empty() : tensor<1x64xf32> + %135 = "ttir.relu"(%133, %134) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %136 = tensor.empty() : tensor<1x64xf32> + %137 = "ttir.relu"(%135, %136) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %138 = tensor.empty() : tensor<1x64xf32> + %139 = "ttir.relu"(%137, %138) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %140 = tensor.empty() : tensor<1x64xf32> + %141 = "ttir.relu"(%139, %140) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %142 = tensor.empty() : tensor<1x64xf32> + %143 = "ttir.relu"(%141, %142) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %144 = tensor.empty() : tensor<1x64xf32> + %145 = "ttir.relu"(%143, %144) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %146 = tensor.empty() : tensor<1x64xf32> + %147 = "ttir.relu"(%145, %146) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %148 = tensor.empty() : tensor<1x64xf32> + %149 = "ttir.relu"(%147, %148) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %150 = tensor.empty() : tensor<1x64xf32> + %151 = "ttir.relu"(%149, %150) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %152 = tensor.empty() : tensor<1x64xf32> + %153 = "ttir.relu"(%151, %152) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %154 = tensor.empty() : tensor<1x64xf32> + %155 = "ttir.relu"(%153, %154) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %156 = tensor.empty() : tensor<1x64xf32> + %157 = "ttir.relu"(%155, %156) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %158 = tensor.empty() : tensor<1x64xf32> + %159 = "ttir.relu"(%157, %158) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %160 = tensor.empty() : tensor<1x64xf32> + %161 = "ttir.relu"(%159, %160) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %162 = tensor.empty() : tensor<1x64xf32> + %163 = "ttir.relu"(%161, %162) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %164 = tensor.empty() : tensor<1x64xf32> + %165 = "ttir.relu"(%163, %164) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %166 = tensor.empty() : tensor<1x64xf32> + %167 = "ttir.relu"(%165, %166) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %168 = tensor.empty() : tensor<1x64xf32> + %169 = "ttir.relu"(%167, %168) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %170 = tensor.empty() : tensor<1x64xf32> + %171 = "ttir.relu"(%169, %170) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %172 = tensor.empty() : tensor<1x64xf32> + %173 = "ttir.relu"(%171, %172) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %174 = tensor.empty() : tensor<1x64xf32> + %175 = "ttir.relu"(%173, %174) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %176 = tensor.empty() : tensor<1x64xf32> + %177 = "ttir.relu"(%175, %176) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %178 = tensor.empty() : tensor<1x64xf32> + %179 = "ttir.relu"(%177, %178) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %180 = tensor.empty() : tensor<1x64xf32> + %181 = "ttir.relu"(%179, %180) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %182 = tensor.empty() : tensor<1x64xf32> + %183 = "ttir.relu"(%181, %182) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %184 = tensor.empty() : tensor<1x64xf32> + %185 = "ttir.relu"(%183, %184) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %186 = tensor.empty() : tensor<1x64xf32> + %187 = "ttir.relu"(%185, %186) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %188 = tensor.empty() : tensor<1x64xf32> + %189 = "ttir.relu"(%187, %188) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %190 = tensor.empty() : tensor<1x64xf32> + %191 = "ttir.relu"(%189, %190) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %192 = tensor.empty() : tensor<1x64xf32> + %193 = "ttir.relu"(%191, %192) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %194 = tensor.empty() : tensor<1x64xf32> + %195 = "ttir.relu"(%193, %194) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %196 = tensor.empty() : tensor<1x64xf32> + %197 = "ttir.relu"(%195, %196) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %198 = tensor.empty() : tensor<1x64xf32> + %199 = "ttir.relu"(%197, %198) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %200 = tensor.empty() : tensor<1x64xf32> + %201 = "ttir.relu"(%199, %200) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %202 = tensor.empty() : tensor<1x64xf32> + %203 = "ttir.relu"(%201, %202) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %204 = tensor.empty() : tensor<1x64xf32> + %205 = "ttir.relu"(%203, %204) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %206 = tensor.empty() : tensor<1x64xf32> + %207 = "ttir.relu"(%205, %206) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %208 = tensor.empty() : tensor<1x64xf32> + %209 = "ttir.relu"(%207, %208) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %210 = tensor.empty() : tensor<1x64xf32> + %211 = "ttir.relu"(%209, %210) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %212 = tensor.empty() : tensor<1x64xf32> + %213 = "ttir.relu"(%211, %212) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %214 = tensor.empty() : tensor<1x64xf32> + %215 = "ttir.relu"(%213, %214) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %216 = tensor.empty() : tensor<1x64xf32> + %217 = "ttir.relu"(%215, %216) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %218 = tensor.empty() : tensor<1x64xf32> + %219 = "ttir.relu"(%217, %218) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %220 = tensor.empty() : tensor<1x64xf32> + %221 = "ttir.relu"(%219, %220) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %222 = tensor.empty() : tensor<1x64xf32> + %223 = "ttir.relu"(%221, %222) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %224 = tensor.empty() : tensor<1x64xf32> + %225 = "ttir.relu"(%223, %224) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %226 = tensor.empty() : tensor<1x64xf32> + %227 = "ttir.relu"(%225, %226) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %228 = tensor.empty() : tensor<1x64xf32> + %229 = "ttir.relu"(%227, %228) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %230 = tensor.empty() : tensor<1x64xf32> + %231 = "ttir.relu"(%229, %230) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %232 = tensor.empty() : tensor<1x64xf32> + %233 = "ttir.relu"(%231, %232) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %234 = tensor.empty() : tensor<1x64xf32> + %235 = "ttir.relu"(%233, %234) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %236 = tensor.empty() : tensor<1x64xf32> + %237 = "ttir.relu"(%235, %236) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %238 = tensor.empty() : tensor<1x64xf32> + %239 = "ttir.relu"(%237, %238) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %240 = tensor.empty() : tensor<1x64xf32> + %241 = "ttir.relu"(%239, %240) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %242 = tensor.empty() : tensor<1x64xf32> + %243 = "ttir.relu"(%241, %242) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %244 = tensor.empty() : tensor<1x64xf32> + %245 = "ttir.relu"(%243, %244) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %246 = tensor.empty() : tensor<1x64xf32> + %247 = "ttir.relu"(%245, %246) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %248 = tensor.empty() : tensor<1x64xf32> + %249 = "ttir.relu"(%247, %248) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %250 = tensor.empty() : tensor<1x64xf32> + %251 = "ttir.relu"(%249, %250) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %252 = tensor.empty() : tensor<1x64xf32> + %253 = "ttir.relu"(%251, %252) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %254 = tensor.empty() : tensor<1x64xf32> + %255 = "ttir.relu"(%253, %254) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %256 = tensor.empty() : tensor<1x64xf32> + %257 = "ttir.relu"(%255, %256) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %258 = tensor.empty() : tensor<1x64xf32> + %259 = "ttir.relu"(%257, %258) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %260 = tensor.empty() : tensor<1x64xf32> + %261 = "ttir.relu"(%259, %260) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %262 = tensor.empty() : tensor<1x64xf32> + %263 = "ttir.relu"(%261, %262) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %264 = tensor.empty() : tensor<1x64xf32> + %265 = "ttir.relu"(%263, %264) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %266 = tensor.empty() : tensor<1x64xf32> + %267 = "ttir.relu"(%265, %266) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %268 = tensor.empty() : tensor<1x64xf32> + %269 = "ttir.relu"(%267, %268) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %270 = tensor.empty() : tensor<1x64xf32> + %271 = "ttir.relu"(%269, %270) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %272 = tensor.empty() : tensor<1x64xf32> + %273 = "ttir.relu"(%271, %272) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %274 = tensor.empty() : tensor<1x64xf32> + %275 = "ttir.relu"(%273, %274) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %276 = tensor.empty() : tensor<1x64xf32> + %277 = "ttir.relu"(%275, %276) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %278 = tensor.empty() : tensor<1x64xf32> + %279 = "ttir.relu"(%277, %278) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %280 = tensor.empty() : tensor<1x64xf32> + %281 = "ttir.relu"(%279, %280) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %282 = tensor.empty() : tensor<1x64xf32> + %283 = "ttir.relu"(%281, %282) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %284 = tensor.empty() : tensor<1x64xf32> + %285 = "ttir.relu"(%283, %284) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %286 = tensor.empty() : tensor<1x64xf32> + %287 = "ttir.relu"(%285, %286) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %288 = tensor.empty() : tensor<1x64xf32> + %289 = "ttir.relu"(%287, %288) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %290 = tensor.empty() : tensor<1x64xf32> + %291 = "ttir.relu"(%289, %290) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %292 = tensor.empty() : tensor<1x64xf32> + %293 = "ttir.relu"(%291, %292) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %294 = tensor.empty() : tensor<1x64xf32> + %295 = "ttir.relu"(%293, %294) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %296 = tensor.empty() : tensor<1x64xf32> + %297 = "ttir.relu"(%295, %296) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %298 = tensor.empty() : tensor<1x64xf32> + %299 = "ttir.relu"(%297, %298) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %300 = tensor.empty() : tensor<1x64xf32> + %301 = "ttir.relu"(%299, %300) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %302 = tensor.empty() : tensor<1x64xf32> + %303 = "ttir.relu"(%301, %302) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %304 = tensor.empty() : tensor<1x64xf32> + %305 = "ttir.relu"(%303, %304) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %306 = tensor.empty() : tensor<1x64xf32> + %307 = "ttir.relu"(%305, %306) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %308 = tensor.empty() : tensor<1x64xf32> + %309 = "ttir.relu"(%307, %308) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %310 = tensor.empty() : tensor<1x64xf32> + %311 = "ttir.relu"(%309, %310) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %312 = tensor.empty() : tensor<1x64xf32> + %313 = "ttir.relu"(%311, %312) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %314 = tensor.empty() : tensor<1x64xf32> + %315 = "ttir.relu"(%313, %314) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %316 = tensor.empty() : tensor<1x64xf32> + %317 = "ttir.relu"(%315, %316) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %318 = tensor.empty() : tensor<1x64xf32> + %319 = "ttir.relu"(%317, %318) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %320 = tensor.empty() : tensor<1x64xf32> + %321 = "ttir.relu"(%319, %320) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %322 = tensor.empty() : tensor<1x64xf32> + %323 = "ttir.relu"(%321, %322) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %324 = tensor.empty() : tensor<1x64xf32> + %325 = "ttir.relu"(%323, %324) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %326 = tensor.empty() : tensor<1x64xf32> + %327 = "ttir.relu"(%325, %326) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %328 = tensor.empty() : tensor<1x64xf32> + %329 = "ttir.relu"(%327, %328) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %330 = tensor.empty() : tensor<1x64xf32> + %331 = "ttir.relu"(%329, %330) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %332 = tensor.empty() : tensor<1x64xf32> + %333 = "ttir.relu"(%331, %332) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %334 = tensor.empty() : tensor<1x64xf32> + %335 = "ttir.relu"(%333, %334) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %336 = tensor.empty() : tensor<1x64xf32> + %337 = "ttir.relu"(%335, %336) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %338 = tensor.empty() : tensor<1x64xf32> + %339 = "ttir.relu"(%337, %338) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %340 = tensor.empty() : tensor<1x64xf32> + %341 = "ttir.relu"(%339, %340) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %342 = tensor.empty() : tensor<1x64xf32> + %343 = "ttir.relu"(%341, %342) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %344 = tensor.empty() : tensor<1x64xf32> + %345 = "ttir.relu"(%343, %344) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %346 = tensor.empty() : tensor<1x64xf32> + %347 = "ttir.relu"(%345, %346) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %348 = tensor.empty() : tensor<1x64xf32> + %349 = "ttir.relu"(%347, %348) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %350 = tensor.empty() : tensor<1x64xf32> + %351 = "ttir.relu"(%349, %350) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %352 = tensor.empty() : tensor<1x64xf32> + %353 = "ttir.relu"(%351, %352) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %354 = tensor.empty() : tensor<1x64xf32> + %355 = "ttir.relu"(%353, %354) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %356 = tensor.empty() : tensor<1x64xf32> + %357 = "ttir.relu"(%355, %356) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %358 = tensor.empty() : tensor<1x64xf32> + %359 = "ttir.relu"(%357, %358) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %360 = tensor.empty() : tensor<1x64xf32> + %361 = "ttir.relu"(%359, %360) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %362 = tensor.empty() : tensor<1x64xf32> + %363 = "ttir.relu"(%361, %362) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %364 = tensor.empty() : tensor<1x64xf32> + %365 = "ttir.relu"(%363, %364) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %366 = tensor.empty() : tensor<1x64xf32> + %367 = "ttir.relu"(%365, %366) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %368 = tensor.empty() : tensor<1x64xf32> + %369 = "ttir.relu"(%367, %368) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %370 = tensor.empty() : tensor<1x64xf32> + %371 = "ttir.relu"(%369, %370) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %372 = tensor.empty() : tensor<1x64xf32> + %373 = "ttir.relu"(%371, %372) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %374 = tensor.empty() : tensor<1x64xf32> + %375 = "ttir.relu"(%373, %374) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %376 = tensor.empty() : tensor<1x64xf32> + %377 = "ttir.relu"(%375, %376) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %378 = tensor.empty() : tensor<1x64xf32> + %379 = "ttir.relu"(%377, %378) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %380 = tensor.empty() : tensor<1x64xf32> + %381 = "ttir.relu"(%379, %380) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %382 = tensor.empty() : tensor<1x64xf32> + %383 = "ttir.relu"(%381, %382) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %384 = tensor.empty() : tensor<1x64xf32> + %385 = "ttir.relu"(%383, %384) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %386 = tensor.empty() : tensor<1x64xf32> + %387 = "ttir.relu"(%385, %386) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %388 = tensor.empty() : tensor<1x64xf32> + %389 = "ttir.relu"(%387, %388) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %390 = tensor.empty() : tensor<1x64xf32> + %391 = "ttir.relu"(%389, %390) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %392 = tensor.empty() : tensor<1x64xf32> + %393 = "ttir.relu"(%391, %392) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %394 = tensor.empty() : tensor<1x64xf32> + %395 = "ttir.relu"(%393, %394) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %396 = tensor.empty() : tensor<1x64xf32> + %397 = "ttir.relu"(%395, %396) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %398 = tensor.empty() : tensor<1x64xf32> + %399 = "ttir.relu"(%397, %398) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %400 = tensor.empty() : tensor<1x64xf32> + %401 = "ttir.relu"(%399, %400) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %402 = tensor.empty() : tensor<1x64xf32> + %403 = "ttir.relu"(%401, %402) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %404 = tensor.empty() : tensor<1x64xf32> + %405 = "ttir.relu"(%403, %404) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %406 = tensor.empty() : tensor<1x64xf32> + %407 = "ttir.relu"(%405, %406) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %408 = tensor.empty() : tensor<1x64xf32> + %409 = "ttir.relu"(%407, %408) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %410 = tensor.empty() : tensor<1x64xf32> + %411 = "ttir.relu"(%409, %410) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %412 = tensor.empty() : tensor<1x64xf32> + %413 = "ttir.relu"(%411, %412) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %414 = tensor.empty() : tensor<1x64xf32> + %415 = "ttir.relu"(%413, %414) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %416 = tensor.empty() : tensor<1x64xf32> + %417 = "ttir.relu"(%415, %416) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %418 = tensor.empty() : tensor<1x64xf32> + %419 = "ttir.relu"(%417, %418) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %420 = tensor.empty() : tensor<1x64xf32> + %421 = "ttir.relu"(%419, %420) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %422 = tensor.empty() : tensor<1x64xf32> + %423 = "ttir.relu"(%421, %422) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %424 = tensor.empty() : tensor<1x64xf32> + %425 = "ttir.relu"(%423, %424) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %426 = tensor.empty() : tensor<1x64xf32> + %427 = "ttir.relu"(%425, %426) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %428 = tensor.empty() : tensor<1x64xf32> + %429 = "ttir.relu"(%427, %428) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %430 = tensor.empty() : tensor<1x64xf32> + %431 = "ttir.relu"(%429, %430) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %432 = tensor.empty() : tensor<1x64xf32> + %433 = "ttir.relu"(%431, %432) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %434 = tensor.empty() : tensor<1x64xf32> + %435 = "ttir.relu"(%433, %434) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %436 = tensor.empty() : tensor<1x64xf32> + %437 = "ttir.relu"(%435, %436) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %438 = tensor.empty() : tensor<1x64xf32> + %439 = "ttir.relu"(%437, %438) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %440 = tensor.empty() : tensor<1x64xf32> + %441 = "ttir.relu"(%439, %440) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %442 = tensor.empty() : tensor<1x64xf32> + %443 = "ttir.relu"(%441, %442) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %444 = tensor.empty() : tensor<1x64xf32> + %445 = "ttir.relu"(%443, %444) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %446 = tensor.empty() : tensor<1x64xf32> + %447 = "ttir.relu"(%445, %446) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %448 = tensor.empty() : tensor<1x64xf32> + %449 = "ttir.relu"(%447, %448) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %450 = tensor.empty() : tensor<1x64xf32> + %451 = "ttir.relu"(%449, %450) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %452 = tensor.empty() : tensor<1x64xf32> + %453 = "ttir.relu"(%451, %452) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %454 = tensor.empty() : tensor<1x64xf32> + %455 = "ttir.relu"(%453, %454) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %456 = tensor.empty() : tensor<1x64xf32> + %457 = "ttir.relu"(%455, %456) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %458 = tensor.empty() : tensor<1x64xf32> + %459 = "ttir.relu"(%457, %458) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %460 = tensor.empty() : tensor<1x64xf32> + %461 = "ttir.relu"(%459, %460) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %462 = tensor.empty() : tensor<1x64xf32> + %463 = "ttir.relu"(%461, %462) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %464 = tensor.empty() : tensor<1x64xf32> + %465 = "ttir.relu"(%463, %464) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %466 = tensor.empty() : tensor<1x64xf32> + %467 = "ttir.relu"(%465, %466) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %468 = tensor.empty() : tensor<1x64xf32> + %469 = "ttir.relu"(%467, %468) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %470 = tensor.empty() : tensor<1x64xf32> + %471 = "ttir.relu"(%469, %470) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %472 = tensor.empty() : tensor<1x64xf32> + %473 = "ttir.relu"(%471, %472) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %474 = tensor.empty() : tensor<1x64xf32> + %475 = "ttir.relu"(%473, %474) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %476 = tensor.empty() : tensor<1x64xf32> + %477 = "ttir.relu"(%475, %476) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %478 = tensor.empty() : tensor<1x64xf32> + %479 = "ttir.relu"(%477, %478) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %480 = tensor.empty() : tensor<1x64xf32> + %481 = "ttir.relu"(%479, %480) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %482 = tensor.empty() : tensor<1x64xf32> + %483 = "ttir.relu"(%481, %482) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %484 = tensor.empty() : tensor<1x64xf32> + %485 = "ttir.relu"(%483, %484) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %486 = tensor.empty() : tensor<1x64xf32> + %487 = "ttir.relu"(%485, %486) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %488 = tensor.empty() : tensor<1x64xf32> + %489 = "ttir.relu"(%487, %488) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %490 = tensor.empty() : tensor<1x64xf32> + %491 = "ttir.relu"(%489, %490) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %492 = tensor.empty() : tensor<1x64xf32> + %493 = "ttir.relu"(%491, %492) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %494 = tensor.empty() : tensor<1x64xf32> + %495 = "ttir.relu"(%493, %494) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %496 = tensor.empty() : tensor<1x64xf32> + %497 = "ttir.relu"(%495, %496) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %498 = tensor.empty() : tensor<1x64xf32> + %499 = "ttir.relu"(%497, %498) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %500 = tensor.empty() : tensor<1x64xf32> + %501 = "ttir.relu"(%499, %500) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %502 = tensor.empty() : tensor<1x64xf32> + %503 = "ttir.relu"(%501, %502) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %504 = tensor.empty() : tensor<1x64xf32> + %505 = "ttir.relu"(%503, %504) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %506 = tensor.empty() : tensor<1x64xf32> + %507 = "ttir.relu"(%505, %506) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %508 = tensor.empty() : tensor<1x64xf32> + %509 = "ttir.relu"(%507, %508) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %510 = tensor.empty() : tensor<1x64xf32> + %511 = "ttir.relu"(%509, %510) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %512 = tensor.empty() : tensor<1x64xf32> + %513 = "ttir.relu"(%511, %512) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %514 = tensor.empty() : tensor<1x64xf32> + %515 = "ttir.relu"(%513, %514) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %516 = tensor.empty() : tensor<1x64xf32> + %517 = "ttir.relu"(%515, %516) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %518 = tensor.empty() : tensor<1x64xf32> + %519 = "ttir.relu"(%517, %518) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %520 = tensor.empty() : tensor<1x64xf32> + %521 = "ttir.relu"(%519, %520) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %522 = tensor.empty() : tensor<1x64xf32> + %523 = "ttir.relu"(%521, %522) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %524 = tensor.empty() : tensor<1x64xf32> + %525 = "ttir.relu"(%523, %524) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %526 = tensor.empty() : tensor<1x64xf32> + %527 = "ttir.relu"(%525, %526) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %528 = tensor.empty() : tensor<1x64xf32> + %529 = "ttir.relu"(%527, %528) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %530 = tensor.empty() : tensor<1x64xf32> + %531 = "ttir.relu"(%529, %530) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %532 = tensor.empty() : tensor<1x64xf32> + %533 = "ttir.relu"(%531, %532) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %534 = tensor.empty() : tensor<1x64xf32> + %535 = "ttir.relu"(%533, %534) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %536 = tensor.empty() : tensor<1x64xf32> + %537 = "ttir.relu"(%535, %536) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %538 = tensor.empty() : tensor<1x64xf32> + %539 = "ttir.relu"(%537, %538) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %540 = tensor.empty() : tensor<1x64xf32> + %541 = "ttir.relu"(%539, %540) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %542 = tensor.empty() : tensor<1x64xf32> + %543 = "ttir.relu"(%541, %542) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %544 = tensor.empty() : tensor<1x64xf32> + %545 = "ttir.relu"(%543, %544) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %546 = tensor.empty() : tensor<1x64xf32> + %547 = "ttir.relu"(%545, %546) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %548 = tensor.empty() : tensor<1x64xf32> + %549 = "ttir.relu"(%547, %548) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %550 = tensor.empty() : tensor<1x64xf32> + %551 = "ttir.relu"(%549, %550) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %552 = tensor.empty() : tensor<1x64xf32> + %553 = "ttir.relu"(%551, %552) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %554 = tensor.empty() : tensor<1x64xf32> + %555 = "ttir.relu"(%553, %554) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %556 = tensor.empty() : tensor<1x64xf32> + %557 = "ttir.relu"(%555, %556) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %558 = tensor.empty() : tensor<1x64xf32> + %559 = "ttir.relu"(%557, %558) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %560 = tensor.empty() : tensor<1x64xf32> + %561 = "ttir.relu"(%559, %560) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %562 = tensor.empty() : tensor<1x64xf32> + %563 = "ttir.relu"(%561, %562) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %564 = tensor.empty() : tensor<1x64xf32> + %565 = "ttir.relu"(%563, %564) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %566 = tensor.empty() : tensor<1x64xf32> + %567 = "ttir.relu"(%565, %566) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %568 = tensor.empty() : tensor<1x64xf32> + %569 = "ttir.relu"(%567, %568) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %570 = tensor.empty() : tensor<1x64xf32> + %571 = "ttir.relu"(%569, %570) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %572 = tensor.empty() : tensor<1x64xf32> + %573 = "ttir.relu"(%571, %572) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %574 = tensor.empty() : tensor<1x64xf32> + %575 = "ttir.relu"(%573, %574) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %576 = tensor.empty() : tensor<1x64xf32> + %577 = "ttir.relu"(%575, %576) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %578 = tensor.empty() : tensor<1x64xf32> + %579 = "ttir.relu"(%577, %578) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %580 = tensor.empty() : tensor<1x64xf32> + %581 = "ttir.relu"(%579, %580) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %582 = tensor.empty() : tensor<1x64xf32> + %583 = "ttir.relu"(%581, %582) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %584 = tensor.empty() : tensor<1x64xf32> + %585 = "ttir.relu"(%583, %584) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %586 = tensor.empty() : tensor<1x64xf32> + %587 = "ttir.relu"(%585, %586) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %588 = tensor.empty() : tensor<1x64xf32> + %589 = "ttir.relu"(%587, %588) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %590 = tensor.empty() : tensor<1x64xf32> + %591 = "ttir.relu"(%589, %590) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %592 = tensor.empty() : tensor<1x64xf32> + %593 = "ttir.relu"(%591, %592) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %594 = tensor.empty() : tensor<1x64xf32> + %595 = "ttir.relu"(%593, %594) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %596 = tensor.empty() : tensor<1x64xf32> + %597 = "ttir.relu"(%595, %596) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %598 = tensor.empty() : tensor<1x64xf32> + %599 = "ttir.relu"(%597, %598) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %600 = tensor.empty() : tensor<1x64xf32> + %601 = "ttir.relu"(%599, %600) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %602 = tensor.empty() : tensor<1x64xf32> + %603 = "ttir.relu"(%601, %602) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %604 = tensor.empty() : tensor<1x64xf32> + %605 = "ttir.relu"(%603, %604) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %606 = tensor.empty() : tensor<1x64xf32> + %607 = "ttir.relu"(%605, %606) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %608 = tensor.empty() : tensor<1x64xf32> + %609 = "ttir.relu"(%607, %608) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %610 = tensor.empty() : tensor<1x64xf32> + %611 = "ttir.relu"(%609, %610) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %612 = tensor.empty() : tensor<1x64xf32> + %613 = "ttir.relu"(%611, %612) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %614 = tensor.empty() : tensor<1x64xf32> + %615 = "ttir.relu"(%613, %614) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %616 = tensor.empty() : tensor<1x64xf32> + %617 = "ttir.relu"(%615, %616) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %618 = tensor.empty() : tensor<1x64xf32> + %619 = "ttir.relu"(%617, %618) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %620 = tensor.empty() : tensor<1x64xf32> + %621 = "ttir.relu"(%619, %620) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %622 = tensor.empty() : tensor<1x64xf32> + %623 = "ttir.relu"(%621, %622) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %624 = tensor.empty() : tensor<1x64xf32> + %625 = "ttir.relu"(%623, %624) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %626 = tensor.empty() : tensor<1x64xf32> + %627 = "ttir.relu"(%625, %626) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %628 = tensor.empty() : tensor<1x64xf32> + %629 = "ttir.relu"(%627, %628) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %630 = tensor.empty() : tensor<1x64xf32> + %631 = "ttir.relu"(%629, %630) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %632 = tensor.empty() : tensor<1x64xf32> + %633 = "ttir.relu"(%631, %632) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %634 = tensor.empty() : tensor<1x64xf32> + %635 = "ttir.relu"(%633, %634) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %636 = tensor.empty() : tensor<1x64xf32> + %637 = "ttir.relu"(%635, %636) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %638 = tensor.empty() : tensor<1x64xf32> + %639 = "ttir.relu"(%637, %638) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %640 = tensor.empty() : tensor<1x64xf32> + %641 = "ttir.relu"(%639, %640) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %642 = tensor.empty() : tensor<1x64xf32> + %643 = "ttir.relu"(%641, %642) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %644 = tensor.empty() : tensor<1x64xf32> + %645 = "ttir.relu"(%643, %644) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %646 = tensor.empty() : tensor<1x64xf32> + %647 = "ttir.relu"(%645, %646) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %648 = tensor.empty() : tensor<1x64xf32> + %649 = "ttir.relu"(%647, %648) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %650 = tensor.empty() : tensor<1x64xf32> + %651 = "ttir.relu"(%649, %650) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %652 = tensor.empty() : tensor<1x64xf32> + %653 = "ttir.relu"(%651, %652) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %654 = tensor.empty() : tensor<1x64xf32> + %655 = "ttir.relu"(%653, %654) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %656 = tensor.empty() : tensor<1x64xf32> + %657 = "ttir.relu"(%655, %656) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %658 = tensor.empty() : tensor<1x64xf32> + %659 = "ttir.relu"(%657, %658) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %660 = tensor.empty() : tensor<1x64xf32> + %661 = "ttir.relu"(%659, %660) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %662 = tensor.empty() : tensor<1x64xf32> + %663 = "ttir.relu"(%661, %662) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %664 = tensor.empty() : tensor<1x64xf32> + %665 = "ttir.relu"(%663, %664) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %666 = tensor.empty() : tensor<1x64xf32> + %667 = "ttir.relu"(%665, %666) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %668 = tensor.empty() : tensor<1x64xf32> + %669 = "ttir.relu"(%667, %668) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %670 = tensor.empty() : tensor<1x64xf32> + %671 = "ttir.relu"(%669, %670) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %672 = tensor.empty() : tensor<1x64xf32> + %673 = "ttir.relu"(%671, %672) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %674 = tensor.empty() : tensor<1x64xf32> + %675 = "ttir.relu"(%673, %674) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %676 = tensor.empty() : tensor<1x64xf32> + %677 = "ttir.relu"(%675, %676) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %678 = tensor.empty() : tensor<1x64xf32> + %679 = "ttir.relu"(%677, %678) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %680 = tensor.empty() : tensor<1x64xf32> + %681 = "ttir.relu"(%679, %680) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %682 = tensor.empty() : tensor<1x64xf32> + %683 = "ttir.relu"(%681, %682) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %684 = tensor.empty() : tensor<1x64xf32> + %685 = "ttir.relu"(%683, %684) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %686 = tensor.empty() : tensor<1x64xf32> + %687 = "ttir.relu"(%685, %686) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %688 = tensor.empty() : tensor<1x64xf32> + %689 = "ttir.relu"(%687, %688) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %690 = tensor.empty() : tensor<1x64xf32> + %691 = "ttir.relu"(%689, %690) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %692 = tensor.empty() : tensor<1x64xf32> + %693 = "ttir.relu"(%691, %692) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %694 = tensor.empty() : tensor<1x64xf32> + %695 = "ttir.relu"(%693, %694) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %696 = tensor.empty() : tensor<1x64xf32> + %697 = "ttir.relu"(%695, %696) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %698 = tensor.empty() : tensor<1x64xf32> + %699 = "ttir.relu"(%697, %698) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %700 = tensor.empty() : tensor<1x64xf32> + %701 = "ttir.relu"(%699, %700) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %702 = tensor.empty() : tensor<1x64xf32> + %703 = "ttir.relu"(%701, %702) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %704 = tensor.empty() : tensor<1x64xf32> + %705 = "ttir.relu"(%703, %704) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %706 = tensor.empty() : tensor<1x64xf32> + %707 = "ttir.relu"(%705, %706) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %708 = tensor.empty() : tensor<1x64xf32> + %709 = "ttir.relu"(%707, %708) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %710 = tensor.empty() : tensor<1x64xf32> + %711 = "ttir.relu"(%709, %710) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %712 = tensor.empty() : tensor<1x64xf32> + %713 = "ttir.relu"(%711, %712) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %714 = tensor.empty() : tensor<1x64xf32> + %715 = "ttir.relu"(%713, %714) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %716 = tensor.empty() : tensor<1x64xf32> + %717 = "ttir.relu"(%715, %716) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %718 = tensor.empty() : tensor<1x64xf32> + %719 = "ttir.relu"(%717, %718) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %720 = tensor.empty() : tensor<1x64xf32> + %721 = "ttir.relu"(%719, %720) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %722 = tensor.empty() : tensor<1x64xf32> + %723 = "ttir.relu"(%721, %722) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %724 = tensor.empty() : tensor<1x64xf32> + %725 = "ttir.relu"(%723, %724) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %726 = tensor.empty() : tensor<1x64xf32> + %727 = "ttir.relu"(%725, %726) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %728 = tensor.empty() : tensor<1x64xf32> + %729 = "ttir.relu"(%727, %728) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %730 = tensor.empty() : tensor<1x64xf32> + %731 = "ttir.relu"(%729, %730) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %732 = tensor.empty() : tensor<1x64xf32> + %733 = "ttir.relu"(%731, %732) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %734 = tensor.empty() : tensor<1x64xf32> + %735 = "ttir.relu"(%733, %734) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %736 = tensor.empty() : tensor<1x64xf32> + %737 = "ttir.relu"(%735, %736) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %738 = tensor.empty() : tensor<1x64xf32> + %739 = "ttir.relu"(%737, %738) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %740 = tensor.empty() : tensor<1x64xf32> + %741 = "ttir.relu"(%739, %740) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %742 = tensor.empty() : tensor<1x64xf32> + %743 = "ttir.relu"(%741, %742) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %744 = tensor.empty() : tensor<1x64xf32> + %745 = "ttir.relu"(%743, %744) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %746 = tensor.empty() : tensor<1x64xf32> + %747 = "ttir.relu"(%745, %746) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %748 = tensor.empty() : tensor<1x64xf32> + %749 = "ttir.relu"(%747, %748) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %750 = tensor.empty() : tensor<1x64xf32> + %751 = "ttir.relu"(%749, %750) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %752 = tensor.empty() : tensor<1x64xf32> + %753 = "ttir.relu"(%751, %752) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %754 = tensor.empty() : tensor<1x64xf32> + %755 = "ttir.relu"(%753, %754) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %756 = tensor.empty() : tensor<1x64xf32> + %757 = "ttir.relu"(%755, %756) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %758 = tensor.empty() : tensor<1x64xf32> + %759 = "ttir.relu"(%757, %758) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %760 = tensor.empty() : tensor<1x64xf32> + %761 = "ttir.relu"(%759, %760) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %762 = tensor.empty() : tensor<1x64xf32> + %763 = "ttir.relu"(%761, %762) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %764 = tensor.empty() : tensor<1x64xf32> + %765 = "ttir.relu"(%763, %764) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %766 = tensor.empty() : tensor<1x64xf32> + %767 = "ttir.relu"(%765, %766) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %768 = tensor.empty() : tensor<1x64xf32> + %769 = "ttir.relu"(%767, %768) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %770 = tensor.empty() : tensor<1x64xf32> + %771 = "ttir.relu"(%769, %770) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %772 = tensor.empty() : tensor<1x64xf32> + %773 = "ttir.relu"(%771, %772) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %774 = tensor.empty() : tensor<1x64xf32> + %775 = "ttir.relu"(%773, %774) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %776 = tensor.empty() : tensor<1x64xf32> + %777 = "ttir.relu"(%775, %776) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %778 = tensor.empty() : tensor<1x64xf32> + %779 = "ttir.relu"(%777, %778) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %780 = tensor.empty() : tensor<1x64xf32> + %781 = "ttir.relu"(%779, %780) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %782 = tensor.empty() : tensor<1x64xf32> + %783 = "ttir.relu"(%781, %782) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %784 = tensor.empty() : tensor<1x64xf32> + %785 = "ttir.relu"(%783, %784) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %786 = tensor.empty() : tensor<1x64xf32> + %787 = "ttir.relu"(%785, %786) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %788 = tensor.empty() : tensor<1x64xf32> + %789 = "ttir.relu"(%787, %788) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %790 = tensor.empty() : tensor<1x64xf32> + %791 = "ttir.relu"(%789, %790) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %792 = tensor.empty() : tensor<1x64xf32> + %793 = "ttir.relu"(%791, %792) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %794 = tensor.empty() : tensor<1x64xf32> + %795 = "ttir.relu"(%793, %794) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %796 = tensor.empty() : tensor<1x64xf32> + %797 = "ttir.relu"(%795, %796) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %798 = tensor.empty() : tensor<1x64xf32> + %799 = "ttir.relu"(%797, %798) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %800 = tensor.empty() : tensor<1x64xf32> + %801 = "ttir.relu"(%799, %800) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %802 = tensor.empty() : tensor<1x64xf32> + %803 = "ttir.relu"(%801, %802) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %804 = tensor.empty() : tensor<1x64xf32> + %805 = "ttir.relu"(%803, %804) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %806 = tensor.empty() : tensor<1x64xf32> + %807 = "ttir.relu"(%805, %806) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %808 = tensor.empty() : tensor<1x64xf32> + %809 = "ttir.relu"(%807, %808) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %810 = tensor.empty() : tensor<1x64xf32> + %811 = "ttir.relu"(%809, %810) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %812 = tensor.empty() : tensor<1x64xf32> + %813 = "ttir.relu"(%811, %812) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %814 = tensor.empty() : tensor<1x64xf32> + %815 = "ttir.relu"(%813, %814) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %816 = tensor.empty() : tensor<1x64xf32> + %817 = "ttir.relu"(%815, %816) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %818 = tensor.empty() : tensor<1x64xf32> + %819 = "ttir.relu"(%817, %818) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %820 = tensor.empty() : tensor<1x64xf32> + %821 = "ttir.relu"(%819, %820) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %822 = tensor.empty() : tensor<1x64xf32> + %823 = "ttir.relu"(%821, %822) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %824 = tensor.empty() : tensor<1x64xf32> + %825 = "ttir.relu"(%823, %824) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %826 = tensor.empty() : tensor<1x64xf32> + %827 = "ttir.relu"(%825, %826) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %828 = tensor.empty() : tensor<1x64xf32> + %829 = "ttir.relu"(%827, %828) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %830 = tensor.empty() : tensor<1x64xf32> + %831 = "ttir.relu"(%829, %830) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %832 = tensor.empty() : tensor<1x64xf32> + %833 = "ttir.relu"(%831, %832) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %834 = tensor.empty() : tensor<1x64xf32> + %835 = "ttir.relu"(%833, %834) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %836 = tensor.empty() : tensor<1x64xf32> + %837 = "ttir.relu"(%835, %836) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %838 = tensor.empty() : tensor<1x64xf32> + %839 = "ttir.relu"(%837, %838) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %840 = tensor.empty() : tensor<1x64xf32> + %841 = "ttir.relu"(%839, %840) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %842 = tensor.empty() : tensor<1x64xf32> + %843 = "ttir.relu"(%841, %842) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %844 = tensor.empty() : tensor<1x64xf32> + %845 = "ttir.relu"(%843, %844) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %846 = tensor.empty() : tensor<1x64xf32> + %847 = "ttir.relu"(%845, %846) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %848 = tensor.empty() : tensor<1x64xf32> + %849 = "ttir.relu"(%847, %848) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %850 = tensor.empty() : tensor<1x64xf32> + %851 = "ttir.relu"(%849, %850) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %852 = tensor.empty() : tensor<1x64xf32> + %853 = "ttir.relu"(%851, %852) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %854 = tensor.empty() : tensor<1x64xf32> + %855 = "ttir.relu"(%853, %854) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %856 = tensor.empty() : tensor<1x64xf32> + %857 = "ttir.relu"(%855, %856) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %858 = tensor.empty() : tensor<1x64xf32> + %859 = "ttir.relu"(%857, %858) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %860 = tensor.empty() : tensor<1x64xf32> + %861 = "ttir.relu"(%859, %860) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %862 = tensor.empty() : tensor<1x64xf32> + %863 = "ttir.relu"(%861, %862) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %864 = tensor.empty() : tensor<1x64xf32> + %865 = "ttir.relu"(%863, %864) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %866 = tensor.empty() : tensor<1x64xf32> + %867 = "ttir.relu"(%865, %866) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %868 = tensor.empty() : tensor<1x64xf32> + %869 = "ttir.relu"(%867, %868) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %870 = tensor.empty() : tensor<1x64xf32> + %871 = "ttir.relu"(%869, %870) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %872 = tensor.empty() : tensor<1x64xf32> + %873 = "ttir.relu"(%871, %872) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %874 = tensor.empty() : tensor<1x64xf32> + %875 = "ttir.relu"(%873, %874) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %876 = tensor.empty() : tensor<1x64xf32> + %877 = "ttir.relu"(%875, %876) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %878 = tensor.empty() : tensor<1x64xf32> + %879 = "ttir.relu"(%877, %878) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %880 = tensor.empty() : tensor<1x64xf32> + %881 = "ttir.relu"(%879, %880) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %882 = tensor.empty() : tensor<1x64xf32> + %883 = "ttir.relu"(%881, %882) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %884 = tensor.empty() : tensor<1x64xf32> + %885 = "ttir.relu"(%883, %884) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %886 = tensor.empty() : tensor<1x64xf32> + %887 = "ttir.relu"(%885, %886) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %888 = tensor.empty() : tensor<1x64xf32> + %889 = "ttir.relu"(%887, %888) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %890 = tensor.empty() : tensor<1x64xf32> + %891 = "ttir.relu"(%889, %890) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %892 = tensor.empty() : tensor<1x64xf32> + %893 = "ttir.relu"(%891, %892) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %894 = tensor.empty() : tensor<1x64xf32> + %895 = "ttir.relu"(%893, %894) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %896 = tensor.empty() : tensor<1x64xf32> + %897 = "ttir.relu"(%895, %896) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %898 = tensor.empty() : tensor<1x64xf32> + %899 = "ttir.relu"(%897, %898) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %900 = tensor.empty() : tensor<1x64xf32> + %901 = "ttir.relu"(%899, %900) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %902 = tensor.empty() : tensor<1x64xf32> + %903 = "ttir.relu"(%901, %902) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %904 = tensor.empty() : tensor<1x64xf32> + %905 = "ttir.relu"(%903, %904) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %906 = tensor.empty() : tensor<1x64xf32> + %907 = "ttir.relu"(%905, %906) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %908 = tensor.empty() : tensor<1x64xf32> + %909 = "ttir.relu"(%907, %908) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %910 = tensor.empty() : tensor<1x64xf32> + %911 = "ttir.relu"(%909, %910) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %912 = tensor.empty() : tensor<1x64xf32> + %913 = "ttir.relu"(%911, %912) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %914 = tensor.empty() : tensor<1x64xf32> + %915 = "ttir.relu"(%913, %914) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %916 = tensor.empty() : tensor<1x64xf32> + %917 = "ttir.relu"(%915, %916) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %918 = tensor.empty() : tensor<1x64xf32> + %919 = "ttir.relu"(%917, %918) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %920 = tensor.empty() : tensor<1x64xf32> + %921 = "ttir.relu"(%919, %920) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %922 = tensor.empty() : tensor<1x64xf32> + %923 = "ttir.relu"(%921, %922) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %924 = tensor.empty() : tensor<1x64xf32> + %925 = "ttir.relu"(%923, %924) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %926 = tensor.empty() : tensor<1x64xf32> + %927 = "ttir.relu"(%925, %926) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %928 = tensor.empty() : tensor<1x64xf32> + %929 = "ttir.relu"(%927, %928) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %930 = tensor.empty() : tensor<1x64xf32> + %931 = "ttir.relu"(%929, %930) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %932 = tensor.empty() : tensor<1x64xf32> + %933 = "ttir.relu"(%931, %932) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %934 = tensor.empty() : tensor<1x64xf32> + %935 = "ttir.relu"(%933, %934) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %936 = tensor.empty() : tensor<1x64xf32> + %937 = "ttir.relu"(%935, %936) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %938 = tensor.empty() : tensor<1x64xf32> + %939 = "ttir.relu"(%937, %938) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %940 = tensor.empty() : tensor<1x64xf32> + %941 = "ttir.relu"(%939, %940) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %942 = tensor.empty() : tensor<1x64xf32> + %943 = "ttir.relu"(%941, %942) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %944 = tensor.empty() : tensor<1x64xf32> + %945 = "ttir.relu"(%943, %944) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %946 = tensor.empty() : tensor<1x64xf32> + %947 = "ttir.relu"(%945, %946) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %948 = tensor.empty() : tensor<1x64xf32> + %949 = "ttir.relu"(%947, %948) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %950 = tensor.empty() : tensor<1x64xf32> + %951 = "ttir.relu"(%949, %950) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %952 = tensor.empty() : tensor<1x64xf32> + %953 = "ttir.relu"(%951, %952) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %954 = tensor.empty() : tensor<1x64xf32> + %955 = "ttir.relu"(%953, %954) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %956 = tensor.empty() : tensor<1x64xf32> + %957 = "ttir.relu"(%955, %956) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %958 = tensor.empty() : tensor<1x64xf32> + %959 = "ttir.relu"(%957, %958) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %960 = tensor.empty() : tensor<1x64xf32> + %961 = "ttir.relu"(%959, %960) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %962 = tensor.empty() : tensor<1x64xf32> + %963 = "ttir.relu"(%961, %962) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %964 = tensor.empty() : tensor<1x64xf32> + %965 = "ttir.relu"(%963, %964) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %966 = tensor.empty() : tensor<1x64xf32> + %967 = "ttir.relu"(%965, %966) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %968 = tensor.empty() : tensor<1x64xf32> + %969 = "ttir.relu"(%967, %968) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %970 = tensor.empty() : tensor<1x64xf32> + %971 = "ttir.relu"(%969, %970) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %972 = tensor.empty() : tensor<1x64xf32> + %973 = "ttir.relu"(%971, %972) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %974 = tensor.empty() : tensor<1x64xf32> + %975 = "ttir.relu"(%973, %974) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %976 = tensor.empty() : tensor<1x64xf32> + %977 = "ttir.relu"(%975, %976) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %978 = tensor.empty() : tensor<1x64xf32> + %979 = "ttir.relu"(%977, %978) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %980 = tensor.empty() : tensor<1x64xf32> + %981 = "ttir.relu"(%979, %980) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %982 = tensor.empty() : tensor<1x64xf32> + %983 = "ttir.relu"(%981, %982) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %984 = tensor.empty() : tensor<1x64xf32> + %985 = "ttir.relu"(%983, %984) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %986 = tensor.empty() : tensor<1x64xf32> + %987 = "ttir.relu"(%985, %986) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %988 = tensor.empty() : tensor<1x64xf32> + %989 = "ttir.relu"(%987, %988) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %990 = tensor.empty() : tensor<1x64xf32> + %991 = "ttir.relu"(%989, %990) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %992 = tensor.empty() : tensor<1x64xf32> + %993 = "ttir.relu"(%991, %992) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %994 = tensor.empty() : tensor<1x64xf32> + %995 = "ttir.relu"(%993, %994) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %996 = tensor.empty() : tensor<1x64xf32> + %997 = "ttir.relu"(%995, %996) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %998 = tensor.empty() : tensor<1x64xf32> + %999 = "ttir.relu"(%997, %998) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1000 = tensor.empty() : tensor<1x64xf32> + %1001 = "ttir.relu"(%999, %1000) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1002 = tensor.empty() : tensor<1x64xf32> + %1003 = "ttir.relu"(%1001, %1002) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1004 = tensor.empty() : tensor<1x64xf32> + %1005 = "ttir.relu"(%1003, %1004) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1006 = tensor.empty() : tensor<1x64xf32> + %1007 = "ttir.relu"(%1005, %1006) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1008 = tensor.empty() : tensor<1x64xf32> + %1009 = "ttir.relu"(%1007, %1008) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1010 = tensor.empty() : tensor<1x64xf32> + %1011 = "ttir.relu"(%1009, %1010) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1012 = tensor.empty() : tensor<1x64xf32> + %1013 = "ttir.relu"(%1011, %1012) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1014 = tensor.empty() : tensor<1x64xf32> + %1015 = "ttir.relu"(%1013, %1014) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1016 = tensor.empty() : tensor<1x64xf32> + %1017 = "ttir.relu"(%1015, %1016) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1018 = tensor.empty() : tensor<1x64xf32> + %1019 = "ttir.relu"(%1017, %1018) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1020 = tensor.empty() : tensor<1x64xf32> + %1021 = "ttir.relu"(%1019, %1020) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1022 = tensor.empty() : tensor<1x64xf32> + %1023 = "ttir.relu"(%1021, %1022) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1024 = tensor.empty() : tensor<1x64xf32> + %1025 = "ttir.relu"(%1023, %1024) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1026 = tensor.empty() : tensor<1x64xf32> + %1027 = "ttir.relu"(%1025, %1026) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1028 = tensor.empty() : tensor<1x64xf32> + %1029 = "ttir.relu"(%1027, %1028) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1030 = tensor.empty() : tensor<1x64xf32> + %1031 = "ttir.relu"(%1029, %1030) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1032 = tensor.empty() : tensor<1x64xf32> + %1033 = "ttir.relu"(%1031, %1032) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1034 = tensor.empty() : tensor<1x64xf32> + %1035 = "ttir.relu"(%1033, %1034) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1036 = tensor.empty() : tensor<1x64xf32> + %1037 = "ttir.relu"(%1035, %1036) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1038 = tensor.empty() : tensor<1x64xf32> + %1039 = "ttir.relu"(%1037, %1038) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1040 = tensor.empty() : tensor<1x64xf32> + %1041 = "ttir.relu"(%1039, %1040) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1042 = tensor.empty() : tensor<1x64xf32> + %1043 = "ttir.relu"(%1041, %1042) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1044 = tensor.empty() : tensor<1x64xf32> + %1045 = "ttir.relu"(%1043, %1044) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1046 = tensor.empty() : tensor<1x64xf32> + %1047 = "ttir.relu"(%1045, %1046) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1048 = tensor.empty() : tensor<1x64xf32> + %1049 = "ttir.relu"(%1047, %1048) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1050 = tensor.empty() : tensor<1x64xf32> + %1051 = "ttir.relu"(%1049, %1050) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1052 = tensor.empty() : tensor<1x64xf32> + %1053 = "ttir.relu"(%1051, %1052) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1054 = tensor.empty() : tensor<1x64xf32> + %1055 = "ttir.relu"(%1053, %1054) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1056 = tensor.empty() : tensor<1x64xf32> + %1057 = "ttir.relu"(%1055, %1056) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1058 = tensor.empty() : tensor<1x64xf32> + %1059 = "ttir.relu"(%1057, %1058) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1060 = tensor.empty() : tensor<1x64xf32> + %1061 = "ttir.relu"(%1059, %1060) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1062 = tensor.empty() : tensor<1x64xf32> + %1063 = "ttir.relu"(%1061, %1062) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1064 = tensor.empty() : tensor<1x64xf32> + %1065 = "ttir.relu"(%1063, %1064) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1066 = tensor.empty() : tensor<1x64xf32> + %1067 = "ttir.relu"(%1065, %1066) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1068 = tensor.empty() : tensor<1x64xf32> + %1069 = "ttir.relu"(%1067, %1068) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1070 = tensor.empty() : tensor<1x64xf32> + %1071 = "ttir.relu"(%1069, %1070) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1072 = tensor.empty() : tensor<1x64xf32> + %1073 = "ttir.relu"(%1071, %1072) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1074 = tensor.empty() : tensor<1x64xf32> + %1075 = "ttir.relu"(%1073, %1074) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1076 = tensor.empty() : tensor<1x64xf32> + %1077 = "ttir.relu"(%1075, %1076) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1078 = tensor.empty() : tensor<1x64xf32> + %1079 = "ttir.relu"(%1077, %1078) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1080 = tensor.empty() : tensor<1x64xf32> + %1081 = "ttir.relu"(%1079, %1080) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1082 = tensor.empty() : tensor<1x64xf32> + %1083 = "ttir.relu"(%1081, %1082) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1084 = tensor.empty() : tensor<1x64xf32> + %1085 = "ttir.relu"(%1083, %1084) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1086 = tensor.empty() : tensor<1x64xf32> + %1087 = "ttir.relu"(%1085, %1086) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1088 = tensor.empty() : tensor<1x64xf32> + %1089 = "ttir.relu"(%1087, %1088) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1090 = tensor.empty() : tensor<1x64xf32> + %1091 = "ttir.relu"(%1089, %1090) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1092 = tensor.empty() : tensor<1x64xf32> + %1093 = "ttir.relu"(%1091, %1092) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1094 = tensor.empty() : tensor<1x64xf32> + %1095 = "ttir.relu"(%1093, %1094) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1096 = tensor.empty() : tensor<1x64xf32> + %1097 = "ttir.relu"(%1095, %1096) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1098 = tensor.empty() : tensor<1x64xf32> + %1099 = "ttir.relu"(%1097, %1098) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1100 = tensor.empty() : tensor<1x64xf32> + %1101 = "ttir.relu"(%1099, %1100) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1102 = tensor.empty() : tensor<1x64xf32> + %1103 = "ttir.relu"(%1101, %1102) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1104 = tensor.empty() : tensor<1x64xf32> + %1105 = "ttir.relu"(%1103, %1104) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1106 = tensor.empty() : tensor<1x64xf32> + %1107 = "ttir.relu"(%1105, %1106) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1108 = tensor.empty() : tensor<1x64xf32> + %1109 = "ttir.relu"(%1107, %1108) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1110 = tensor.empty() : tensor<1x64xf32> + %1111 = "ttir.relu"(%1109, %1110) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1112 = tensor.empty() : tensor<1x64xf32> + %1113 = "ttir.relu"(%1111, %1112) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1114 = tensor.empty() : tensor<1x64xf32> + %1115 = "ttir.relu"(%1113, %1114) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1116 = tensor.empty() : tensor<1x64xf32> + %1117 = "ttir.relu"(%1115, %1116) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1118 = tensor.empty() : tensor<1x64xf32> + %1119 = "ttir.relu"(%1117, %1118) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1120 = tensor.empty() : tensor<1x64xf32> + %1121 = "ttir.relu"(%1119, %1120) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1122 = tensor.empty() : tensor<1x64xf32> + %1123 = "ttir.relu"(%1121, %1122) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1124 = tensor.empty() : tensor<1x64xf32> + %1125 = "ttir.relu"(%1123, %1124) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1126 = tensor.empty() : tensor<1x64xf32> + %1127 = "ttir.relu"(%1125, %1126) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1128 = tensor.empty() : tensor<1x64xf32> + %1129 = "ttir.relu"(%1127, %1128) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1130 = tensor.empty() : tensor<1x64xf32> + %1131 = "ttir.relu"(%1129, %1130) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1132 = tensor.empty() : tensor<1x64xf32> + %1133 = "ttir.relu"(%1131, %1132) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1134 = tensor.empty() : tensor<1x64xf32> + %1135 = "ttir.relu"(%1133, %1134) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1136 = tensor.empty() : tensor<1x64xf32> + %1137 = "ttir.relu"(%1135, %1136) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1138 = tensor.empty() : tensor<1x64xf32> + %1139 = "ttir.relu"(%1137, %1138) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1140 = tensor.empty() : tensor<1x64xf32> + %1141 = "ttir.relu"(%1139, %1140) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1142 = tensor.empty() : tensor<1x64xf32> + %1143 = "ttir.relu"(%1141, %1142) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1144 = tensor.empty() : tensor<1x64xf32> + %1145 = "ttir.relu"(%1143, %1144) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1146 = tensor.empty() : tensor<1x64xf32> + %1147 = "ttir.relu"(%1145, %1146) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1148 = tensor.empty() : tensor<1x64xf32> + %1149 = "ttir.relu"(%1147, %1148) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1150 = tensor.empty() : tensor<1x64xf32> + %1151 = "ttir.relu"(%1149, %1150) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1152 = tensor.empty() : tensor<1x64xf32> + %1153 = "ttir.relu"(%1151, %1152) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1154 = tensor.empty() : tensor<1x64xf32> + %1155 = "ttir.relu"(%1153, %1154) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1156 = tensor.empty() : tensor<1x64xf32> + %1157 = "ttir.relu"(%1155, %1156) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1158 = tensor.empty() : tensor<1x64xf32> + %1159 = "ttir.relu"(%1157, %1158) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1160 = tensor.empty() : tensor<1x64xf32> + %1161 = "ttir.relu"(%1159, %1160) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1162 = tensor.empty() : tensor<1x64xf32> + %1163 = "ttir.relu"(%1161, %1162) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1164 = tensor.empty() : tensor<1x64xf32> + %1165 = "ttir.relu"(%1163, %1164) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1166 = tensor.empty() : tensor<1x64xf32> + %1167 = "ttir.relu"(%1165, %1166) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1168 = tensor.empty() : tensor<1x64xf32> + %1169 = "ttir.relu"(%1167, %1168) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1170 = tensor.empty() : tensor<1x64xf32> + %1171 = "ttir.relu"(%1169, %1170) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1172 = tensor.empty() : tensor<1x64xf32> + %1173 = "ttir.relu"(%1171, %1172) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1174 = tensor.empty() : tensor<1x64xf32> + %1175 = "ttir.relu"(%1173, %1174) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1176 = tensor.empty() : tensor<1x64xf32> + %1177 = "ttir.relu"(%1175, %1176) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1178 = tensor.empty() : tensor<1x64xf32> + %1179 = "ttir.relu"(%1177, %1178) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1180 = tensor.empty() : tensor<1x64xf32> + %1181 = "ttir.relu"(%1179, %1180) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1182 = tensor.empty() : tensor<1x64xf32> + %1183 = "ttir.relu"(%1181, %1182) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1184 = tensor.empty() : tensor<1x64xf32> + %1185 = "ttir.relu"(%1183, %1184) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1186 = tensor.empty() : tensor<1x64xf32> + %1187 = "ttir.relu"(%1185, %1186) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1188 = tensor.empty() : tensor<1x64xf32> + %1189 = "ttir.relu"(%1187, %1188) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1190 = tensor.empty() : tensor<1x64xf32> + %1191 = "ttir.relu"(%1189, %1190) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1192 = tensor.empty() : tensor<1x64xf32> + %1193 = "ttir.relu"(%1191, %1192) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1194 = tensor.empty() : tensor<1x64xf32> + %1195 = "ttir.relu"(%1193, %1194) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1196 = tensor.empty() : tensor<1x64xf32> + %1197 = "ttir.relu"(%1195, %1196) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1198 = tensor.empty() : tensor<1x64xf32> + %1199 = "ttir.relu"(%1197, %1198) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1200 = tensor.empty() : tensor<1x64xf32> + %1201 = "ttir.relu"(%1199, %1200) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1202 = tensor.empty() : tensor<1x64xf32> + %1203 = "ttir.relu"(%1201, %1202) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1204 = tensor.empty() : tensor<1x64xf32> + %1205 = "ttir.relu"(%1203, %1204) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1206 = tensor.empty() : tensor<1x64xf32> + %1207 = "ttir.relu"(%1205, %1206) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1208 = tensor.empty() : tensor<1x64xf32> + %1209 = "ttir.relu"(%1207, %1208) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1210 = tensor.empty() : tensor<1x64xf32> + %1211 = "ttir.relu"(%1209, %1210) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1212 = tensor.empty() : tensor<1x64xf32> + %1213 = "ttir.relu"(%1211, %1212) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1214 = tensor.empty() : tensor<1x64xf32> + %1215 = "ttir.relu"(%1213, %1214) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1216 = tensor.empty() : tensor<1x64xf32> + %1217 = "ttir.relu"(%1215, %1216) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1218 = tensor.empty() : tensor<1x64xf32> + %1219 = "ttir.relu"(%1217, %1218) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1220 = tensor.empty() : tensor<1x64xf32> + %1221 = "ttir.relu"(%1219, %1220) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1222 = tensor.empty() : tensor<1x64xf32> + %1223 = "ttir.relu"(%1221, %1222) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1224 = tensor.empty() : tensor<1x64xf32> + %1225 = "ttir.relu"(%1223, %1224) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1226 = tensor.empty() : tensor<1x64xf32> + %1227 = "ttir.relu"(%1225, %1226) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1228 = tensor.empty() : tensor<1x64xf32> + %1229 = "ttir.relu"(%1227, %1228) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1230 = tensor.empty() : tensor<1x64xf32> + %1231 = "ttir.relu"(%1229, %1230) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1232 = tensor.empty() : tensor<1x64xf32> + %1233 = "ttir.relu"(%1231, %1232) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1234 = tensor.empty() : tensor<1x64xf32> + %1235 = "ttir.relu"(%1233, %1234) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1236 = tensor.empty() : tensor<1x64xf32> + %1237 = "ttir.relu"(%1235, %1236) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1238 = tensor.empty() : tensor<1x64xf32> + %1239 = "ttir.relu"(%1237, %1238) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1240 = tensor.empty() : tensor<1x64xf32> + %1241 = "ttir.relu"(%1239, %1240) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1242 = tensor.empty() : tensor<1x64xf32> + %1243 = "ttir.relu"(%1241, %1242) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1244 = tensor.empty() : tensor<1x64xf32> + %1245 = "ttir.relu"(%1243, %1244) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1246 = tensor.empty() : tensor<1x64xf32> + %1247 = "ttir.relu"(%1245, %1246) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1248 = tensor.empty() : tensor<1x64xf32> + %1249 = "ttir.relu"(%1247, %1248) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1250 = tensor.empty() : tensor<1x64xf32> + %1251 = "ttir.relu"(%1249, %1250) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1252 = tensor.empty() : tensor<1x64xf32> + %1253 = "ttir.relu"(%1251, %1252) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1254 = tensor.empty() : tensor<1x64xf32> + %1255 = "ttir.relu"(%1253, %1254) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1256 = tensor.empty() : tensor<1x64xf32> + %1257 = "ttir.relu"(%1255, %1256) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1258 = tensor.empty() : tensor<1x64xf32> + %1259 = "ttir.relu"(%1257, %1258) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1260 = tensor.empty() : tensor<1x64xf32> + %1261 = "ttir.relu"(%1259, %1260) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1262 = tensor.empty() : tensor<1x64xf32> + %1263 = "ttir.relu"(%1261, %1262) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1264 = tensor.empty() : tensor<1x64xf32> + %1265 = "ttir.relu"(%1263, %1264) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1266 = tensor.empty() : tensor<1x64xf32> + %1267 = "ttir.relu"(%1265, %1266) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1268 = tensor.empty() : tensor<1x64xf32> + %1269 = "ttir.relu"(%1267, %1268) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1270 = tensor.empty() : tensor<1x64xf32> + %1271 = "ttir.relu"(%1269, %1270) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1272 = tensor.empty() : tensor<1x64xf32> + %1273 = "ttir.relu"(%1271, %1272) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1274 = tensor.empty() : tensor<1x64xf32> + %1275 = "ttir.relu"(%1273, %1274) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1276 = tensor.empty() : tensor<1x64xf32> + %1277 = "ttir.relu"(%1275, %1276) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1278 = tensor.empty() : tensor<1x64xf32> + %1279 = "ttir.relu"(%1277, %1278) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1280 = tensor.empty() : tensor<1x64xf32> + %1281 = "ttir.relu"(%1279, %1280) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1282 = tensor.empty() : tensor<1x64xf32> + %1283 = "ttir.relu"(%1281, %1282) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1284 = tensor.empty() : tensor<1x64xf32> + %1285 = "ttir.relu"(%1283, %1284) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1286 = tensor.empty() : tensor<1x64xf32> + %1287 = "ttir.relu"(%1285, %1286) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1288 = tensor.empty() : tensor<1x64xf32> + %1289 = "ttir.relu"(%1287, %1288) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1290 = tensor.empty() : tensor<1x64xf32> + %1291 = "ttir.relu"(%1289, %1290) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1292 = tensor.empty() : tensor<1x64xf32> + %1293 = "ttir.relu"(%1291, %1292) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1294 = tensor.empty() : tensor<1x64xf32> + %1295 = "ttir.relu"(%1293, %1294) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1296 = tensor.empty() : tensor<1x64xf32> + %1297 = "ttir.relu"(%1295, %1296) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1298 = tensor.empty() : tensor<1x64xf32> + %1299 = "ttir.relu"(%1297, %1298) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1300 = tensor.empty() : tensor<1x64xf32> + %1301 = "ttir.relu"(%1299, %1300) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1302 = tensor.empty() : tensor<1x64xf32> + %1303 = "ttir.relu"(%1301, %1302) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1304 = tensor.empty() : tensor<1x64xf32> + %1305 = "ttir.relu"(%1303, %1304) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1306 = tensor.empty() : tensor<1x64xf32> + %1307 = "ttir.relu"(%1305, %1306) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1308 = tensor.empty() : tensor<1x64xf32> + %1309 = "ttir.relu"(%1307, %1308) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1310 = tensor.empty() : tensor<1x64xf32> + %1311 = "ttir.relu"(%1309, %1310) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1312 = tensor.empty() : tensor<1x64xf32> + %1313 = "ttir.relu"(%1311, %1312) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1314 = tensor.empty() : tensor<1x64xf32> + %1315 = "ttir.relu"(%1313, %1314) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1316 = tensor.empty() : tensor<1x64xf32> + %1317 = "ttir.relu"(%1315, %1316) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1318 = tensor.empty() : tensor<1x64xf32> + %1319 = "ttir.relu"(%1317, %1318) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1320 = tensor.empty() : tensor<1x64xf32> + %1321 = "ttir.relu"(%1319, %1320) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1322 = tensor.empty() : tensor<1x64xf32> + %1323 = "ttir.relu"(%1321, %1322) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1324 = tensor.empty() : tensor<1x64xf32> + %1325 = "ttir.relu"(%1323, %1324) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1326 = tensor.empty() : tensor<1x64xf32> + %1327 = "ttir.relu"(%1325, %1326) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1328 = tensor.empty() : tensor<1x64xf32> + %1329 = "ttir.relu"(%1327, %1328) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1330 = tensor.empty() : tensor<1x64xf32> + %1331 = "ttir.relu"(%1329, %1330) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1332 = tensor.empty() : tensor<1x64xf32> + %1333 = "ttir.relu"(%1331, %1332) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1334 = tensor.empty() : tensor<1x64xf32> + %1335 = "ttir.relu"(%1333, %1334) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1336 = tensor.empty() : tensor<1x64xf32> + %1337 = "ttir.relu"(%1335, %1336) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1338 = tensor.empty() : tensor<1x64xf32> + %1339 = "ttir.relu"(%1337, %1338) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1340 = tensor.empty() : tensor<1x64xf32> + %1341 = "ttir.relu"(%1339, %1340) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1342 = tensor.empty() : tensor<1x64xf32> + %1343 = "ttir.relu"(%1341, %1342) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1344 = tensor.empty() : tensor<1x64xf32> + %1345 = "ttir.relu"(%1343, %1344) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1346 = tensor.empty() : tensor<1x64xf32> + %1347 = "ttir.relu"(%1345, %1346) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1348 = tensor.empty() : tensor<1x64xf32> + %1349 = "ttir.relu"(%1347, %1348) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1350 = tensor.empty() : tensor<1x64xf32> + %1351 = "ttir.relu"(%1349, %1350) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1352 = tensor.empty() : tensor<1x64xf32> + %1353 = "ttir.relu"(%1351, %1352) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1354 = tensor.empty() : tensor<1x64xf32> + %1355 = "ttir.relu"(%1353, %1354) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1356 = tensor.empty() : tensor<1x64xf32> + %1357 = "ttir.relu"(%1355, %1356) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1358 = tensor.empty() : tensor<1x64xf32> + %1359 = "ttir.relu"(%1357, %1358) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1360 = tensor.empty() : tensor<1x64xf32> + %1361 = "ttir.relu"(%1359, %1360) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1362 = tensor.empty() : tensor<1x64xf32> + %1363 = "ttir.relu"(%1361, %1362) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1364 = tensor.empty() : tensor<1x64xf32> + %1365 = "ttir.relu"(%1363, %1364) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1366 = tensor.empty() : tensor<1x64xf32> + %1367 = "ttir.relu"(%1365, %1366) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1368 = tensor.empty() : tensor<1x64xf32> + %1369 = "ttir.relu"(%1367, %1368) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1370 = tensor.empty() : tensor<1x64xf32> + %1371 = "ttir.relu"(%1369, %1370) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1372 = tensor.empty() : tensor<1x64xf32> + %1373 = "ttir.relu"(%1371, %1372) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1374 = tensor.empty() : tensor<1x64xf32> + %1375 = "ttir.relu"(%1373, %1374) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1376 = tensor.empty() : tensor<1x64xf32> + %1377 = "ttir.relu"(%1375, %1376) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1378 = tensor.empty() : tensor<1x64xf32> + %1379 = "ttir.relu"(%1377, %1378) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1380 = tensor.empty() : tensor<1x64xf32> + %1381 = "ttir.relu"(%1379, %1380) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1382 = tensor.empty() : tensor<1x64xf32> + %1383 = "ttir.relu"(%1381, %1382) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1384 = tensor.empty() : tensor<1x64xf32> + %1385 = "ttir.relu"(%1383, %1384) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1386 = tensor.empty() : tensor<1x64xf32> + %1387 = "ttir.relu"(%1385, %1386) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1388 = tensor.empty() : tensor<1x64xf32> + %1389 = "ttir.relu"(%1387, %1388) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1390 = tensor.empty() : tensor<1x64xf32> + %1391 = "ttir.relu"(%1389, %1390) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1392 = tensor.empty() : tensor<1x64xf32> + %1393 = "ttir.relu"(%1391, %1392) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1394 = tensor.empty() : tensor<1x64xf32> + %1395 = "ttir.relu"(%1393, %1394) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1396 = tensor.empty() : tensor<1x64xf32> + %1397 = "ttir.relu"(%1395, %1396) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1398 = tensor.empty() : tensor<1x64xf32> + %1399 = "ttir.relu"(%1397, %1398) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1400 = tensor.empty() : tensor<1x64xf32> + %1401 = "ttir.relu"(%1399, %1400) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1402 = tensor.empty() : tensor<1x64xf32> + %1403 = "ttir.relu"(%1401, %1402) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1404 = tensor.empty() : tensor<1x64xf32> + %1405 = "ttir.relu"(%1403, %1404) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1406 = tensor.empty() : tensor<1x64xf32> + %1407 = "ttir.relu"(%1405, %1406) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1408 = tensor.empty() : tensor<1x64xf32> + %1409 = "ttir.relu"(%1407, %1408) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1410 = tensor.empty() : tensor<1x64xf32> + %1411 = "ttir.relu"(%1409, %1410) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1412 = tensor.empty() : tensor<1x64xf32> + %1413 = "ttir.relu"(%1411, %1412) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1414 = tensor.empty() : tensor<1x64xf32> + %1415 = "ttir.relu"(%1413, %1414) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1416 = tensor.empty() : tensor<1x64xf32> + %1417 = "ttir.relu"(%1415, %1416) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1418 = tensor.empty() : tensor<1x64xf32> + %1419 = "ttir.relu"(%1417, %1418) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1420 = tensor.empty() : tensor<1x64xf32> + %1421 = "ttir.relu"(%1419, %1420) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1422 = tensor.empty() : tensor<1x64xf32> + %1423 = "ttir.relu"(%1421, %1422) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1424 = tensor.empty() : tensor<1x64xf32> + %1425 = "ttir.relu"(%1423, %1424) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1426 = tensor.empty() : tensor<1x64xf32> + %1427 = "ttir.relu"(%1425, %1426) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1428 = tensor.empty() : tensor<1x64xf32> + %1429 = "ttir.relu"(%1427, %1428) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1430 = tensor.empty() : tensor<1x64xf32> + %1431 = "ttir.relu"(%1429, %1430) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1432 = tensor.empty() : tensor<1x64xf32> + %1433 = "ttir.relu"(%1431, %1432) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1434 = tensor.empty() : tensor<1x64xf32> + %1435 = "ttir.relu"(%1433, %1434) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1436 = tensor.empty() : tensor<1x64xf32> + %1437 = "ttir.relu"(%1435, %1436) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1438 = tensor.empty() : tensor<1x64xf32> + %1439 = "ttir.relu"(%1437, %1438) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1440 = tensor.empty() : tensor<1x64xf32> + %1441 = "ttir.relu"(%1439, %1440) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1442 = tensor.empty() : tensor<1x64xf32> + %1443 = "ttir.relu"(%1441, %1442) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1444 = tensor.empty() : tensor<1x64xf32> + %1445 = "ttir.relu"(%1443, %1444) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1446 = tensor.empty() : tensor<1x64xf32> + %1447 = "ttir.relu"(%1445, %1446) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1448 = tensor.empty() : tensor<1x64xf32> + %1449 = "ttir.relu"(%1447, %1448) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1450 = tensor.empty() : tensor<1x64xf32> + %1451 = "ttir.relu"(%1449, %1450) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1452 = tensor.empty() : tensor<1x64xf32> + %1453 = "ttir.relu"(%1451, %1452) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1454 = tensor.empty() : tensor<1x64xf32> + %1455 = "ttir.relu"(%1453, %1454) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1456 = tensor.empty() : tensor<1x64xf32> + %1457 = "ttir.relu"(%1455, %1456) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1458 = tensor.empty() : tensor<1x64xf32> + %1459 = "ttir.relu"(%1457, %1458) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1460 = tensor.empty() : tensor<1x64xf32> + %1461 = "ttir.relu"(%1459, %1460) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1462 = tensor.empty() : tensor<1x64xf32> + %1463 = "ttir.relu"(%1461, %1462) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1464 = tensor.empty() : tensor<1x64xf32> + %1465 = "ttir.relu"(%1463, %1464) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1466 = tensor.empty() : tensor<1x64xf32> + %1467 = "ttir.relu"(%1465, %1466) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1468 = tensor.empty() : tensor<1x64xf32> + %1469 = "ttir.relu"(%1467, %1468) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1470 = tensor.empty() : tensor<1x64xf32> + %1471 = "ttir.relu"(%1469, %1470) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1472 = tensor.empty() : tensor<1x64xf32> + %1473 = "ttir.relu"(%1471, %1472) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1474 = tensor.empty() : tensor<1x64xf32> + %1475 = "ttir.relu"(%1473, %1474) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1476 = tensor.empty() : tensor<1x64xf32> + %1477 = "ttir.relu"(%1475, %1476) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1478 = tensor.empty() : tensor<1x64xf32> + %1479 = "ttir.relu"(%1477, %1478) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1480 = tensor.empty() : tensor<1x64xf32> + %1481 = "ttir.relu"(%1479, %1480) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1482 = tensor.empty() : tensor<1x64xf32> + %1483 = "ttir.relu"(%1481, %1482) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1484 = tensor.empty() : tensor<1x64xf32> + %1485 = "ttir.relu"(%1483, %1484) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1486 = tensor.empty() : tensor<1x64xf32> + %1487 = "ttir.relu"(%1485, %1486) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1488 = tensor.empty() : tensor<1x64xf32> + %1489 = "ttir.relu"(%1487, %1488) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1490 = tensor.empty() : tensor<1x64xf32> + %1491 = "ttir.relu"(%1489, %1490) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1492 = tensor.empty() : tensor<1x64xf32> + %1493 = "ttir.relu"(%1491, %1492) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1494 = tensor.empty() : tensor<1x64xf32> + %1495 = "ttir.relu"(%1493, %1494) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1496 = tensor.empty() : tensor<1x64xf32> + %1497 = "ttir.relu"(%1495, %1496) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1498 = tensor.empty() : tensor<1x64xf32> + %1499 = "ttir.relu"(%1497, %1498) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1500 = tensor.empty() : tensor<1x64xf32> + %1501 = "ttir.relu"(%1499, %1500) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1502 = tensor.empty() : tensor<1x64xf32> + %1503 = "ttir.relu"(%1501, %1502) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1504 = tensor.empty() : tensor<1x64xf32> + %1505 = "ttir.relu"(%1503, %1504) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1506 = tensor.empty() : tensor<1x64xf32> + %1507 = "ttir.relu"(%1505, %1506) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1508 = tensor.empty() : tensor<1x64xf32> + %1509 = "ttir.relu"(%1507, %1508) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1510 = tensor.empty() : tensor<1x64xf32> + %1511 = "ttir.relu"(%1509, %1510) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1512 = tensor.empty() : tensor<1x64xf32> + %1513 = "ttir.relu"(%1511, %1512) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1514 = tensor.empty() : tensor<1x64xf32> + %1515 = "ttir.relu"(%1513, %1514) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1516 = tensor.empty() : tensor<1x64xf32> + %1517 = "ttir.relu"(%1515, %1516) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1518 = tensor.empty() : tensor<1x64xf32> + %1519 = "ttir.relu"(%1517, %1518) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1520 = tensor.empty() : tensor<1x64xf32> + %1521 = "ttir.relu"(%1519, %1520) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1522 = tensor.empty() : tensor<1x64xf32> + %1523 = "ttir.relu"(%1521, %1522) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1524 = tensor.empty() : tensor<1x64xf32> + %1525 = "ttir.relu"(%1523, %1524) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1526 = tensor.empty() : tensor<1x64xf32> + %1527 = "ttir.relu"(%1525, %1526) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1528 = tensor.empty() : tensor<1x64xf32> + %1529 = "ttir.relu"(%1527, %1528) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1530 = tensor.empty() : tensor<1x64xf32> + %1531 = "ttir.relu"(%1529, %1530) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1532 = tensor.empty() : tensor<1x64xf32> + %1533 = "ttir.relu"(%1531, %1532) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1534 = tensor.empty() : tensor<1x64xf32> + %1535 = "ttir.relu"(%1533, %1534) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1536 = tensor.empty() : tensor<1x64xf32> + %1537 = "ttir.relu"(%1535, %1536) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1538 = tensor.empty() : tensor<1x64xf32> + %1539 = "ttir.relu"(%1537, %1538) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1540 = tensor.empty() : tensor<1x64xf32> + %1541 = "ttir.relu"(%1539, %1540) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1542 = tensor.empty() : tensor<1x64xf32> + %1543 = "ttir.relu"(%1541, %1542) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1544 = tensor.empty() : tensor<1x64xf32> + %1545 = "ttir.relu"(%1543, %1544) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1546 = tensor.empty() : tensor<1x64xf32> + %1547 = "ttir.relu"(%1545, %1546) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1548 = tensor.empty() : tensor<1x64xf32> + %1549 = "ttir.relu"(%1547, %1548) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1550 = tensor.empty() : tensor<1x64xf32> + %1551 = "ttir.relu"(%1549, %1550) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1552 = tensor.empty() : tensor<1x64xf32> + %1553 = "ttir.relu"(%1551, %1552) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1554 = tensor.empty() : tensor<1x64xf32> + %1555 = "ttir.relu"(%1553, %1554) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1556 = tensor.empty() : tensor<1x64xf32> + %1557 = "ttir.relu"(%1555, %1556) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1558 = tensor.empty() : tensor<1x64xf32> + %1559 = "ttir.relu"(%1557, %1558) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1560 = tensor.empty() : tensor<1x64xf32> + %1561 = "ttir.relu"(%1559, %1560) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1562 = tensor.empty() : tensor<1x64xf32> + %1563 = "ttir.relu"(%1561, %1562) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1564 = tensor.empty() : tensor<1x64xf32> + %1565 = "ttir.relu"(%1563, %1564) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1566 = tensor.empty() : tensor<1x64xf32> + %1567 = "ttir.relu"(%1565, %1566) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1568 = tensor.empty() : tensor<1x64xf32> + %1569 = "ttir.relu"(%1567, %1568) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1570 = tensor.empty() : tensor<1x64xf32> + %1571 = "ttir.relu"(%1569, %1570) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1572 = tensor.empty() : tensor<1x64xf32> + %1573 = "ttir.relu"(%1571, %1572) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1574 = tensor.empty() : tensor<1x64xf32> + %1575 = "ttir.relu"(%1573, %1574) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1576 = tensor.empty() : tensor<1x64xf32> + %1577 = "ttir.relu"(%1575, %1576) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1578 = tensor.empty() : tensor<1x64xf32> + %1579 = "ttir.relu"(%1577, %1578) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1580 = tensor.empty() : tensor<1x64xf32> + %1581 = "ttir.relu"(%1579, %1580) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1582 = tensor.empty() : tensor<1x64xf32> + %1583 = "ttir.relu"(%1581, %1582) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1584 = tensor.empty() : tensor<1x64xf32> + %1585 = "ttir.relu"(%1583, %1584) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1586 = tensor.empty() : tensor<1x64xf32> + %1587 = "ttir.relu"(%1585, %1586) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1588 = tensor.empty() : tensor<1x64xf32> + %1589 = "ttir.relu"(%1587, %1588) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1590 = tensor.empty() : tensor<1x64xf32> + %1591 = "ttir.relu"(%1589, %1590) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1592 = tensor.empty() : tensor<1x64xf32> + %1593 = "ttir.relu"(%1591, %1592) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1594 = tensor.empty() : tensor<1x64xf32> + %1595 = "ttir.relu"(%1593, %1594) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1596 = tensor.empty() : tensor<1x64xf32> + %1597 = "ttir.relu"(%1595, %1596) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1598 = tensor.empty() : tensor<1x64xf32> + %1599 = "ttir.relu"(%1597, %1598) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1600 = tensor.empty() : tensor<1x64xf32> + %1601 = "ttir.relu"(%1599, %1600) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1602 = tensor.empty() : tensor<1x64xf32> + %1603 = "ttir.relu"(%1601, %1602) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1604 = tensor.empty() : tensor<1x64xf32> + %1605 = "ttir.relu"(%1603, %1604) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1606 = tensor.empty() : tensor<1x64xf32> + %1607 = "ttir.relu"(%1605, %1606) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1608 = tensor.empty() : tensor<1x64xf32> + %1609 = "ttir.relu"(%1607, %1608) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1610 = tensor.empty() : tensor<1x64xf32> + %1611 = "ttir.relu"(%1609, %1610) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1612 = tensor.empty() : tensor<1x64xf32> + %1613 = "ttir.relu"(%1611, %1612) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1614 = tensor.empty() : tensor<1x64xf32> + %1615 = "ttir.relu"(%1613, %1614) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1616 = tensor.empty() : tensor<1x64xf32> + %1617 = "ttir.relu"(%1615, %1616) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1618 = tensor.empty() : tensor<1x64xf32> + %1619 = "ttir.relu"(%1617, %1618) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1620 = tensor.empty() : tensor<1x64xf32> + %1621 = "ttir.relu"(%1619, %1620) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1622 = tensor.empty() : tensor<1x64xf32> + %1623 = "ttir.relu"(%1621, %1622) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1624 = tensor.empty() : tensor<1x64xf32> + %1625 = "ttir.relu"(%1623, %1624) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1626 = tensor.empty() : tensor<1x64xf32> + %1627 = "ttir.relu"(%1625, %1626) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1628 = tensor.empty() : tensor<1x64xf32> + %1629 = "ttir.relu"(%1627, %1628) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1630 = tensor.empty() : tensor<1x64xf32> + %1631 = "ttir.relu"(%1629, %1630) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1632 = tensor.empty() : tensor<1x64xf32> + %1633 = "ttir.relu"(%1631, %1632) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1634 = tensor.empty() : tensor<1x64xf32> + %1635 = "ttir.relu"(%1633, %1634) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1636 = tensor.empty() : tensor<1x64xf32> + %1637 = "ttir.relu"(%1635, %1636) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1638 = tensor.empty() : tensor<1x64xf32> + %1639 = "ttir.relu"(%1637, %1638) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1640 = tensor.empty() : tensor<1x64xf32> + %1641 = "ttir.relu"(%1639, %1640) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1642 = tensor.empty() : tensor<1x64xf32> + %1643 = "ttir.relu"(%1641, %1642) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1644 = tensor.empty() : tensor<1x64xf32> + %1645 = "ttir.relu"(%1643, %1644) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1646 = tensor.empty() : tensor<1x64xf32> + %1647 = "ttir.relu"(%1645, %1646) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1648 = tensor.empty() : tensor<1x64xf32> + %1649 = "ttir.relu"(%1647, %1648) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1650 = tensor.empty() : tensor<1x64xf32> + %1651 = "ttir.relu"(%1649, %1650) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1652 = tensor.empty() : tensor<1x64xf32> + %1653 = "ttir.relu"(%1651, %1652) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1654 = tensor.empty() : tensor<1x64xf32> + %1655 = "ttir.relu"(%1653, %1654) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1656 = tensor.empty() : tensor<1x64xf32> + %1657 = "ttir.relu"(%1655, %1656) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1658 = tensor.empty() : tensor<1x64xf32> + %1659 = "ttir.relu"(%1657, %1658) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1660 = tensor.empty() : tensor<1x64xf32> + %1661 = "ttir.relu"(%1659, %1660) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1662 = tensor.empty() : tensor<1x64xf32> + %1663 = "ttir.relu"(%1661, %1662) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1664 = tensor.empty() : tensor<1x64xf32> + %1665 = "ttir.relu"(%1663, %1664) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1666 = tensor.empty() : tensor<1x64xf32> + %1667 = "ttir.relu"(%1665, %1666) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1668 = tensor.empty() : tensor<1x64xf32> + %1669 = "ttir.relu"(%1667, %1668) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1670 = tensor.empty() : tensor<1x64xf32> + %1671 = "ttir.relu"(%1669, %1670) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1672 = tensor.empty() : tensor<1x64xf32> + %1673 = "ttir.relu"(%1671, %1672) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1674 = tensor.empty() : tensor<1x64xf32> + %1675 = "ttir.relu"(%1673, %1674) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1676 = tensor.empty() : tensor<1x64xf32> + %1677 = "ttir.relu"(%1675, %1676) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1678 = tensor.empty() : tensor<1x64xf32> + %1679 = "ttir.relu"(%1677, %1678) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1680 = tensor.empty() : tensor<1x64xf32> + %1681 = "ttir.relu"(%1679, %1680) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1682 = tensor.empty() : tensor<1x64xf32> + %1683 = "ttir.relu"(%1681, %1682) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1684 = tensor.empty() : tensor<1x64xf32> + %1685 = "ttir.relu"(%1683, %1684) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1686 = tensor.empty() : tensor<1x64xf32> + %1687 = "ttir.relu"(%1685, %1686) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1688 = tensor.empty() : tensor<1x64xf32> + %1689 = "ttir.relu"(%1687, %1688) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1690 = tensor.empty() : tensor<1x64xf32> + %1691 = "ttir.relu"(%1689, %1690) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1692 = tensor.empty() : tensor<1x64xf32> + %1693 = "ttir.relu"(%1691, %1692) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1694 = tensor.empty() : tensor<1x64xf32> + %1695 = "ttir.relu"(%1693, %1694) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1696 = tensor.empty() : tensor<1x64xf32> + %1697 = "ttir.relu"(%1695, %1696) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1698 = tensor.empty() : tensor<1x64xf32> + %1699 = "ttir.relu"(%1697, %1698) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1700 = tensor.empty() : tensor<1x64xf32> + %1701 = "ttir.relu"(%1699, %1700) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1702 = tensor.empty() : tensor<1x64xf32> + %1703 = "ttir.relu"(%1701, %1702) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1704 = tensor.empty() : tensor<1x64xf32> + %1705 = "ttir.relu"(%1703, %1704) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1706 = tensor.empty() : tensor<1x64xf32> + %1707 = "ttir.relu"(%1705, %1706) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1708 = tensor.empty() : tensor<1x64xf32> + %1709 = "ttir.relu"(%1707, %1708) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1710 = tensor.empty() : tensor<1x64xf32> + %1711 = "ttir.relu"(%1709, %1710) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1712 = tensor.empty() : tensor<1x64xf32> + %1713 = "ttir.relu"(%1711, %1712) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1714 = tensor.empty() : tensor<1x64xf32> + %1715 = "ttir.relu"(%1713, %1714) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1716 = tensor.empty() : tensor<1x64xf32> + %1717 = "ttir.relu"(%1715, %1716) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1718 = tensor.empty() : tensor<1x64xf32> + %1719 = "ttir.relu"(%1717, %1718) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1720 = tensor.empty() : tensor<1x64xf32> + %1721 = "ttir.relu"(%1719, %1720) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1722 = tensor.empty() : tensor<1x64xf32> + %1723 = "ttir.relu"(%1721, %1722) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1724 = tensor.empty() : tensor<1x64xf32> + %1725 = "ttir.relu"(%1723, %1724) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1726 = tensor.empty() : tensor<1x64xf32> + %1727 = "ttir.relu"(%1725, %1726) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1728 = tensor.empty() : tensor<1x64xf32> + %1729 = "ttir.relu"(%1727, %1728) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1730 = tensor.empty() : tensor<1x64xf32> + %1731 = "ttir.relu"(%1729, %1730) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1732 = tensor.empty() : tensor<1x64xf32> + %1733 = "ttir.relu"(%1731, %1732) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1734 = tensor.empty() : tensor<1x64xf32> + %1735 = "ttir.relu"(%1733, %1734) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1736 = tensor.empty() : tensor<1x64xf32> + %1737 = "ttir.relu"(%1735, %1736) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1738 = tensor.empty() : tensor<1x64xf32> + %1739 = "ttir.relu"(%1737, %1738) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1740 = tensor.empty() : tensor<1x64xf32> + %1741 = "ttir.relu"(%1739, %1740) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1742 = tensor.empty() : tensor<1x64xf32> + %1743 = "ttir.relu"(%1741, %1742) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1744 = tensor.empty() : tensor<1x64xf32> + %1745 = "ttir.relu"(%1743, %1744) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1746 = tensor.empty() : tensor<1x64xf32> + %1747 = "ttir.relu"(%1745, %1746) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1748 = tensor.empty() : tensor<1x64xf32> + %1749 = "ttir.relu"(%1747, %1748) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1750 = tensor.empty() : tensor<1x64xf32> + %1751 = "ttir.relu"(%1749, %1750) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1752 = tensor.empty() : tensor<1x64xf32> + %1753 = "ttir.relu"(%1751, %1752) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1754 = tensor.empty() : tensor<1x64xf32> + %1755 = "ttir.relu"(%1753, %1754) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1756 = tensor.empty() : tensor<1x64xf32> + %1757 = "ttir.relu"(%1755, %1756) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1758 = tensor.empty() : tensor<1x64xf32> + %1759 = "ttir.relu"(%1757, %1758) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1760 = tensor.empty() : tensor<1x64xf32> + %1761 = "ttir.relu"(%1759, %1760) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1762 = tensor.empty() : tensor<1x64xf32> + %1763 = "ttir.relu"(%1761, %1762) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1764 = tensor.empty() : tensor<1x64xf32> + %1765 = "ttir.relu"(%1763, %1764) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1766 = tensor.empty() : tensor<1x64xf32> + %1767 = "ttir.relu"(%1765, %1766) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1768 = tensor.empty() : tensor<1x64xf32> + %1769 = "ttir.relu"(%1767, %1768) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1770 = tensor.empty() : tensor<1x64xf32> + %1771 = "ttir.relu"(%1769, %1770) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1772 = tensor.empty() : tensor<1x64xf32> + %1773 = "ttir.relu"(%1771, %1772) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1774 = tensor.empty() : tensor<1x64xf32> + %1775 = "ttir.relu"(%1773, %1774) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1776 = tensor.empty() : tensor<1x64xf32> + %1777 = "ttir.relu"(%1775, %1776) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1778 = tensor.empty() : tensor<1x64xf32> + %1779 = "ttir.relu"(%1777, %1778) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1780 = tensor.empty() : tensor<1x64xf32> + %1781 = "ttir.relu"(%1779, %1780) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1782 = tensor.empty() : tensor<1x64xf32> + %1783 = "ttir.relu"(%1781, %1782) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1784 = tensor.empty() : tensor<1x64xf32> + %1785 = "ttir.relu"(%1783, %1784) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1786 = tensor.empty() : tensor<1x64xf32> + %1787 = "ttir.relu"(%1785, %1786) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1788 = tensor.empty() : tensor<1x64xf32> + %1789 = "ttir.relu"(%1787, %1788) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1790 = tensor.empty() : tensor<1x64xf32> + %1791 = "ttir.relu"(%1789, %1790) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1792 = tensor.empty() : tensor<1x64xf32> + %1793 = "ttir.relu"(%1791, %1792) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1794 = tensor.empty() : tensor<1x64xf32> + %1795 = "ttir.relu"(%1793, %1794) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1796 = tensor.empty() : tensor<1x64xf32> + %1797 = "ttir.relu"(%1795, %1796) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1798 = tensor.empty() : tensor<1x64xf32> + %1799 = "ttir.relu"(%1797, %1798) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1800 = tensor.empty() : tensor<1x64xf32> + %1801 = "ttir.relu"(%1799, %1800) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1802 = tensor.empty() : tensor<1x64xf32> + %1803 = "ttir.relu"(%1801, %1802) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1804 = tensor.empty() : tensor<1x64xf32> + %1805 = "ttir.relu"(%1803, %1804) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1806 = tensor.empty() : tensor<1x64xf32> + %1807 = "ttir.relu"(%1805, %1806) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1808 = tensor.empty() : tensor<1x64xf32> + %1809 = "ttir.relu"(%1807, %1808) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1810 = tensor.empty() : tensor<1x64xf32> + %1811 = "ttir.relu"(%1809, %1810) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1812 = tensor.empty() : tensor<1x64xf32> + %1813 = "ttir.relu"(%1811, %1812) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1814 = tensor.empty() : tensor<1x64xf32> + %1815 = "ttir.relu"(%1813, %1814) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1816 = tensor.empty() : tensor<1x64xf32> + %1817 = "ttir.relu"(%1815, %1816) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1818 = tensor.empty() : tensor<1x64xf32> + %1819 = "ttir.relu"(%1817, %1818) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1820 = tensor.empty() : tensor<1x64xf32> + %1821 = "ttir.relu"(%1819, %1820) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1822 = tensor.empty() : tensor<1x64xf32> + %1823 = "ttir.relu"(%1821, %1822) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1824 = tensor.empty() : tensor<1x64xf32> + %1825 = "ttir.relu"(%1823, %1824) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1826 = tensor.empty() : tensor<1x64xf32> + %1827 = "ttir.relu"(%1825, %1826) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1828 = tensor.empty() : tensor<1x64xf32> + %1829 = "ttir.relu"(%1827, %1828) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1830 = tensor.empty() : tensor<1x64xf32> + %1831 = "ttir.relu"(%1829, %1830) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1832 = tensor.empty() : tensor<1x64xf32> + %1833 = "ttir.relu"(%1831, %1832) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1834 = tensor.empty() : tensor<1x64xf32> + %1835 = "ttir.relu"(%1833, %1834) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1836 = tensor.empty() : tensor<1x64xf32> + %1837 = "ttir.relu"(%1835, %1836) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1838 = tensor.empty() : tensor<1x64xf32> + %1839 = "ttir.relu"(%1837, %1838) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1840 = tensor.empty() : tensor<1x64xf32> + %1841 = "ttir.relu"(%1839, %1840) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1842 = tensor.empty() : tensor<1x64xf32> + %1843 = "ttir.relu"(%1841, %1842) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1844 = tensor.empty() : tensor<1x64xf32> + %1845 = "ttir.relu"(%1843, %1844) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1846 = tensor.empty() : tensor<1x64xf32> + %1847 = "ttir.relu"(%1845, %1846) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1848 = tensor.empty() : tensor<1x64xf32> + %1849 = "ttir.relu"(%1847, %1848) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1850 = tensor.empty() : tensor<1x64xf32> + %1851 = "ttir.relu"(%1849, %1850) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1852 = tensor.empty() : tensor<1x64xf32> + %1853 = "ttir.relu"(%1851, %1852) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1854 = tensor.empty() : tensor<1x64xf32> + %1855 = "ttir.relu"(%1853, %1854) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1856 = tensor.empty() : tensor<1x64xf32> + %1857 = "ttir.relu"(%1855, %1856) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1858 = tensor.empty() : tensor<1x64xf32> + %1859 = "ttir.relu"(%1857, %1858) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1860 = tensor.empty() : tensor<1x64xf32> + %1861 = "ttir.relu"(%1859, %1860) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1862 = tensor.empty() : tensor<1x64xf32> + %1863 = "ttir.relu"(%1861, %1862) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1864 = tensor.empty() : tensor<1x64xf32> + %1865 = "ttir.relu"(%1863, %1864) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1866 = tensor.empty() : tensor<1x64xf32> + %1867 = "ttir.relu"(%1865, %1866) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1868 = tensor.empty() : tensor<1x64xf32> + %1869 = "ttir.relu"(%1867, %1868) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1870 = tensor.empty() : tensor<1x64xf32> + %1871 = "ttir.relu"(%1869, %1870) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1872 = tensor.empty() : tensor<1x64xf32> + %1873 = "ttir.relu"(%1871, %1872) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1874 = tensor.empty() : tensor<1x64xf32> + %1875 = "ttir.relu"(%1873, %1874) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1876 = tensor.empty() : tensor<1x64xf32> + %1877 = "ttir.relu"(%1875, %1876) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1878 = tensor.empty() : tensor<1x64xf32> + %1879 = "ttir.relu"(%1877, %1878) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1880 = tensor.empty() : tensor<1x64xf32> + %1881 = "ttir.relu"(%1879, %1880) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1882 = tensor.empty() : tensor<1x64xf32> + %1883 = "ttir.relu"(%1881, %1882) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1884 = tensor.empty() : tensor<1x64xf32> + %1885 = "ttir.relu"(%1883, %1884) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1886 = tensor.empty() : tensor<1x64xf32> + %1887 = "ttir.relu"(%1885, %1886) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1888 = tensor.empty() : tensor<1x64xf32> + %1889 = "ttir.relu"(%1887, %1888) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1890 = tensor.empty() : tensor<1x64xf32> + %1891 = "ttir.relu"(%1889, %1890) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1892 = tensor.empty() : tensor<1x64xf32> + %1893 = "ttir.relu"(%1891, %1892) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1894 = tensor.empty() : tensor<1x64xf32> + %1895 = "ttir.relu"(%1893, %1894) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1896 = tensor.empty() : tensor<1x64xf32> + %1897 = "ttir.relu"(%1895, %1896) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1898 = tensor.empty() : tensor<1x64xf32> + %1899 = "ttir.relu"(%1897, %1898) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1900 = tensor.empty() : tensor<1x64xf32> + %1901 = "ttir.relu"(%1899, %1900) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1902 = tensor.empty() : tensor<1x64xf32> + %1903 = "ttir.relu"(%1901, %1902) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1904 = tensor.empty() : tensor<1x64xf32> + %1905 = "ttir.relu"(%1903, %1904) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1906 = tensor.empty() : tensor<1x64xf32> + %1907 = "ttir.relu"(%1905, %1906) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1908 = tensor.empty() : tensor<1x64xf32> + %1909 = "ttir.relu"(%1907, %1908) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1910 = tensor.empty() : tensor<1x64xf32> + %1911 = "ttir.relu"(%1909, %1910) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1912 = tensor.empty() : tensor<1x64xf32> + %1913 = "ttir.relu"(%1911, %1912) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1914 = tensor.empty() : tensor<1x64xf32> + %1915 = "ttir.relu"(%1913, %1914) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1916 = tensor.empty() : tensor<1x64xf32> + %1917 = "ttir.relu"(%1915, %1916) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1918 = tensor.empty() : tensor<1x64xf32> + %1919 = "ttir.relu"(%1917, %1918) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1920 = tensor.empty() : tensor<1x64xf32> + %1921 = "ttir.relu"(%1919, %1920) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1922 = tensor.empty() : tensor<1x64xf32> + %1923 = "ttir.relu"(%1921, %1922) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1924 = tensor.empty() : tensor<1x64xf32> + %1925 = "ttir.relu"(%1923, %1924) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1926 = tensor.empty() : tensor<1x64xf32> + %1927 = "ttir.relu"(%1925, %1926) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1928 = tensor.empty() : tensor<1x64xf32> + %1929 = "ttir.relu"(%1927, %1928) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1930 = tensor.empty() : tensor<1x64xf32> + %1931 = "ttir.relu"(%1929, %1930) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1932 = tensor.empty() : tensor<1x64xf32> + %1933 = "ttir.relu"(%1931, %1932) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1934 = tensor.empty() : tensor<1x64xf32> + %1935 = "ttir.relu"(%1933, %1934) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1936 = tensor.empty() : tensor<1x64xf32> + %1937 = "ttir.relu"(%1935, %1936) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1938 = tensor.empty() : tensor<1x64xf32> + %1939 = "ttir.relu"(%1937, %1938) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1940 = tensor.empty() : tensor<1x64xf32> + %1941 = "ttir.relu"(%1939, %1940) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1942 = tensor.empty() : tensor<1x64xf32> + %1943 = "ttir.relu"(%1941, %1942) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1944 = tensor.empty() : tensor<1x64xf32> + %1945 = "ttir.relu"(%1943, %1944) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1946 = tensor.empty() : tensor<1x64xf32> + %1947 = "ttir.relu"(%1945, %1946) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1948 = tensor.empty() : tensor<1x64xf32> + %1949 = "ttir.relu"(%1947, %1948) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1950 = tensor.empty() : tensor<1x64xf32> + %1951 = "ttir.relu"(%1949, %1950) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1952 = tensor.empty() : tensor<1x64xf32> + %1953 = "ttir.relu"(%1951, %1952) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1954 = tensor.empty() : tensor<1x64xf32> + %1955 = "ttir.relu"(%1953, %1954) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1956 = tensor.empty() : tensor<1x64xf32> + %1957 = "ttir.relu"(%1955, %1956) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1958 = tensor.empty() : tensor<1x64xf32> + %1959 = "ttir.relu"(%1957, %1958) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1960 = tensor.empty() : tensor<1x64xf32> + %1961 = "ttir.relu"(%1959, %1960) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1962 = tensor.empty() : tensor<1x64xf32> + %1963 = "ttir.relu"(%1961, %1962) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1964 = tensor.empty() : tensor<1x64xf32> + %1965 = "ttir.relu"(%1963, %1964) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1966 = tensor.empty() : tensor<1x64xf32> + %1967 = "ttir.relu"(%1965, %1966) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1968 = tensor.empty() : tensor<1x64xf32> + %1969 = "ttir.relu"(%1967, %1968) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1970 = tensor.empty() : tensor<1x64xf32> + %1971 = "ttir.relu"(%1969, %1970) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1972 = tensor.empty() : tensor<1x64xf32> + %1973 = "ttir.relu"(%1971, %1972) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1974 = tensor.empty() : tensor<1x64xf32> + %1975 = "ttir.relu"(%1973, %1974) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1976 = tensor.empty() : tensor<1x64xf32> + %1977 = "ttir.relu"(%1975, %1976) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1978 = tensor.empty() : tensor<1x64xf32> + %1979 = "ttir.relu"(%1977, %1978) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1980 = tensor.empty() : tensor<1x64xf32> + %1981 = "ttir.relu"(%1979, %1980) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1982 = tensor.empty() : tensor<1x64xf32> + %1983 = "ttir.relu"(%1981, %1982) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1984 = tensor.empty() : tensor<1x64xf32> + %1985 = "ttir.relu"(%1983, %1984) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1986 = tensor.empty() : tensor<1x64xf32> + %1987 = "ttir.relu"(%1985, %1986) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1988 = tensor.empty() : tensor<1x64xf32> + %1989 = "ttir.relu"(%1987, %1988) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1990 = tensor.empty() : tensor<1x64xf32> + %1991 = "ttir.relu"(%1989, %1990) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1992 = tensor.empty() : tensor<1x64xf32> + %1993 = "ttir.relu"(%1991, %1992) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1994 = tensor.empty() : tensor<1x64xf32> + %1995 = "ttir.relu"(%1993, %1994) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1996 = tensor.empty() : tensor<1x64xf32> + %1997 = "ttir.relu"(%1995, %1996) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %1998 = tensor.empty() : tensor<1x64xf32> + %1999 = "ttir.relu"(%1997, %1998) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2000 = tensor.empty() : tensor<1x64xf32> + %2001 = "ttir.relu"(%1999, %2000) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2002 = tensor.empty() : tensor<1x64xf32> + %2003 = "ttir.relu"(%2001, %2002) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2004 = tensor.empty() : tensor<1x64xf32> + %2005 = "ttir.relu"(%2003, %2004) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2006 = tensor.empty() : tensor<1x64xf32> + %2007 = "ttir.relu"(%2005, %2006) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2008 = tensor.empty() : tensor<1x64xf32> + %2009 = "ttir.relu"(%2007, %2008) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2010 = tensor.empty() : tensor<1x64xf32> + %2011 = "ttir.relu"(%2009, %2010) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2012 = tensor.empty() : tensor<1x64xf32> + %2013 = "ttir.relu"(%2011, %2012) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2014 = tensor.empty() : tensor<1x64xf32> + %2015 = "ttir.relu"(%2013, %2014) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2016 = tensor.empty() : tensor<1x64xf32> + %2017 = "ttir.relu"(%2015, %2016) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2018 = tensor.empty() : tensor<1x64xf32> + %2019 = "ttir.relu"(%2017, %2018) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2020 = tensor.empty() : tensor<1x64xf32> + %2021 = "ttir.relu"(%2019, %2020) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2022 = tensor.empty() : tensor<1x64xf32> + %2023 = "ttir.relu"(%2021, %2022) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2024 = tensor.empty() : tensor<1x64xf32> + %2025 = "ttir.relu"(%2023, %2024) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2026 = tensor.empty() : tensor<1x64xf32> + %2027 = "ttir.relu"(%2025, %2026) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2028 = tensor.empty() : tensor<1x64xf32> + %2029 = "ttir.relu"(%2027, %2028) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2030 = tensor.empty() : tensor<1x64xf32> + %2031 = "ttir.relu"(%2029, %2030) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2032 = tensor.empty() : tensor<1x64xf32> + %2033 = "ttir.relu"(%2031, %2032) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2034 = tensor.empty() : tensor<1x64xf32> + %2035 = "ttir.relu"(%2033, %2034) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2036 = tensor.empty() : tensor<1x64xf32> + %2037 = "ttir.relu"(%2035, %2036) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2038 = tensor.empty() : tensor<1x64xf32> + %2039 = "ttir.relu"(%2037, %2038) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2040 = tensor.empty() : tensor<1x64xf32> + %2041 = "ttir.relu"(%2039, %2040) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2042 = tensor.empty() : tensor<1x64xf32> + %2043 = "ttir.relu"(%2041, %2042) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2044 = tensor.empty() : tensor<1x64xf32> + %2045 = "ttir.relu"(%2043, %2044) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2046 = tensor.empty() : tensor<1x64xf32> + %2047 = "ttir.relu"(%2045, %2046) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2048 = tensor.empty() : tensor<1x64xf32> + %2049 = "ttir.relu"(%2047, %2048) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2050 = tensor.empty() : tensor<1x64xf32> + %2051 = "ttir.relu"(%2049, %2050) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2052 = tensor.empty() : tensor<1x64xf32> + %2053 = "ttir.relu"(%2051, %2052) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2054 = tensor.empty() : tensor<1x64xf32> + %2055 = "ttir.relu"(%2053, %2054) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2056 = tensor.empty() : tensor<1x64xf32> + %2057 = "ttir.relu"(%2055, %2056) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2058 = tensor.empty() : tensor<1x64xf32> + %2059 = "ttir.relu"(%2057, %2058) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2060 = tensor.empty() : tensor<1x64xf32> + %2061 = "ttir.relu"(%2059, %2060) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2062 = tensor.empty() : tensor<1x64xf32> + %2063 = "ttir.relu"(%2061, %2062) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2064 = tensor.empty() : tensor<1x64xf32> + %2065 = "ttir.relu"(%2063, %2064) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2066 = tensor.empty() : tensor<1x64xf32> + %2067 = "ttir.relu"(%2065, %2066) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2068 = tensor.empty() : tensor<1x64xf32> + %2069 = "ttir.relu"(%2067, %2068) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2070 = tensor.empty() : tensor<1x64xf32> + %2071 = "ttir.relu"(%2069, %2070) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2072 = tensor.empty() : tensor<1x64xf32> + %2073 = "ttir.relu"(%2071, %2072) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2074 = tensor.empty() : tensor<1x64xf32> + %2075 = "ttir.relu"(%2073, %2074) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2076 = tensor.empty() : tensor<1x64xf32> + %2077 = "ttir.relu"(%2075, %2076) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2078 = tensor.empty() : tensor<1x64xf32> + %2079 = "ttir.relu"(%2077, %2078) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2080 = tensor.empty() : tensor<1x64xf32> + %2081 = "ttir.relu"(%2079, %2080) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2082 = tensor.empty() : tensor<1x64xf32> + %2083 = "ttir.relu"(%2081, %2082) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2084 = tensor.empty() : tensor<1x64xf32> + %2085 = "ttir.relu"(%2083, %2084) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2086 = tensor.empty() : tensor<1x64xf32> + %2087 = "ttir.relu"(%2085, %2086) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2088 = tensor.empty() : tensor<1x64xf32> + %2089 = "ttir.relu"(%2087, %2088) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2090 = tensor.empty() : tensor<1x64xf32> + %2091 = "ttir.relu"(%2089, %2090) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2092 = tensor.empty() : tensor<1x64xf32> + %2093 = "ttir.relu"(%2091, %2092) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2094 = tensor.empty() : tensor<1x64xf32> + %2095 = "ttir.relu"(%2093, %2094) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2096 = tensor.empty() : tensor<1x64xf32> + %2097 = "ttir.relu"(%2095, %2096) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2098 = tensor.empty() : tensor<1x64xf32> + %2099 = "ttir.relu"(%2097, %2098) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2100 = tensor.empty() : tensor<1x64xf32> + %2101 = "ttir.relu"(%2099, %2100) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2102 = tensor.empty() : tensor<1x64xf32> + %2103 = "ttir.relu"(%2101, %2102) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2104 = tensor.empty() : tensor<1x64xf32> + %2105 = "ttir.relu"(%2103, %2104) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2106 = tensor.empty() : tensor<1x64xf32> + %2107 = "ttir.relu"(%2105, %2106) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2108 = tensor.empty() : tensor<1x64xf32> + %2109 = "ttir.relu"(%2107, %2108) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2110 = tensor.empty() : tensor<1x64xf32> + %2111 = "ttir.relu"(%2109, %2110) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2112 = tensor.empty() : tensor<1x64xf32> + %2113 = "ttir.relu"(%2111, %2112) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2114 = tensor.empty() : tensor<1x64xf32> + %2115 = "ttir.relu"(%2113, %2114) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2116 = tensor.empty() : tensor<1x64xf32> + %2117 = "ttir.relu"(%2115, %2116) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2118 = tensor.empty() : tensor<1x64xf32> + %2119 = "ttir.relu"(%2117, %2118) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2120 = tensor.empty() : tensor<1x64xf32> + %2121 = "ttir.relu"(%2119, %2120) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2122 = tensor.empty() : tensor<1x64xf32> + %2123 = "ttir.relu"(%2121, %2122) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2124 = tensor.empty() : tensor<1x64xf32> + %2125 = "ttir.relu"(%2123, %2124) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2126 = tensor.empty() : tensor<1x64xf32> + %2127 = "ttir.relu"(%2125, %2126) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2128 = tensor.empty() : tensor<1x64xf32> + %2129 = "ttir.relu"(%2127, %2128) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2130 = tensor.empty() : tensor<1x64xf32> + %2131 = "ttir.relu"(%2129, %2130) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2132 = tensor.empty() : tensor<1x64xf32> + %2133 = "ttir.relu"(%2131, %2132) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2134 = tensor.empty() : tensor<1x64xf32> + %2135 = "ttir.relu"(%2133, %2134) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2136 = tensor.empty() : tensor<1x64xf32> + %2137 = "ttir.relu"(%2135, %2136) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2138 = tensor.empty() : tensor<1x64xf32> + %2139 = "ttir.relu"(%2137, %2138) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2140 = tensor.empty() : tensor<1x64xf32> + %2141 = "ttir.relu"(%2139, %2140) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2142 = tensor.empty() : tensor<1x64xf32> + %2143 = "ttir.relu"(%2141, %2142) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2144 = tensor.empty() : tensor<1x64xf32> + %2145 = "ttir.relu"(%2143, %2144) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2146 = tensor.empty() : tensor<1x64xf32> + %2147 = "ttir.relu"(%2145, %2146) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2148 = tensor.empty() : tensor<1x64xf32> + %2149 = "ttir.relu"(%2147, %2148) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2150 = tensor.empty() : tensor<1x64xf32> + %2151 = "ttir.relu"(%2149, %2150) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2152 = tensor.empty() : tensor<1x64xf32> + %2153 = "ttir.relu"(%2151, %2152) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2154 = tensor.empty() : tensor<1x64xf32> + %2155 = "ttir.relu"(%2153, %2154) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2156 = tensor.empty() : tensor<1x64xf32> + %2157 = "ttir.relu"(%2155, %2156) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2158 = tensor.empty() : tensor<1x64xf32> + %2159 = "ttir.relu"(%2157, %2158) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2160 = tensor.empty() : tensor<1x64xf32> + %2161 = "ttir.relu"(%2159, %2160) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2162 = tensor.empty() : tensor<1x64xf32> + %2163 = "ttir.relu"(%2161, %2162) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2164 = tensor.empty() : tensor<1x64xf32> + %2165 = "ttir.relu"(%2163, %2164) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2166 = tensor.empty() : tensor<1x64xf32> + %2167 = "ttir.relu"(%2165, %2166) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2168 = tensor.empty() : tensor<1x64xf32> + %2169 = "ttir.relu"(%2167, %2168) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2170 = tensor.empty() : tensor<1x64xf32> + %2171 = "ttir.relu"(%2169, %2170) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2172 = tensor.empty() : tensor<1x64xf32> + %2173 = "ttir.relu"(%2171, %2172) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2174 = tensor.empty() : tensor<1x64xf32> + %2175 = "ttir.relu"(%2173, %2174) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2176 = tensor.empty() : tensor<1x64xf32> + %2177 = "ttir.relu"(%2175, %2176) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2178 = tensor.empty() : tensor<1x64xf32> + %2179 = "ttir.relu"(%2177, %2178) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2180 = tensor.empty() : tensor<1x64xf32> + %2181 = "ttir.relu"(%2179, %2180) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2182 = tensor.empty() : tensor<1x64xf32> + %2183 = "ttir.relu"(%2181, %2182) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2184 = tensor.empty() : tensor<1x64xf32> + %2185 = "ttir.relu"(%2183, %2184) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2186 = tensor.empty() : tensor<1x64xf32> + %2187 = "ttir.relu"(%2185, %2186) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2188 = tensor.empty() : tensor<1x64xf32> + %2189 = "ttir.relu"(%2187, %2188) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2190 = tensor.empty() : tensor<1x64xf32> + %2191 = "ttir.relu"(%2189, %2190) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2192 = tensor.empty() : tensor<1x64xf32> + %2193 = "ttir.relu"(%2191, %2192) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2194 = tensor.empty() : tensor<1x64xf32> + %2195 = "ttir.relu"(%2193, %2194) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2196 = tensor.empty() : tensor<1x64xf32> + %2197 = "ttir.relu"(%2195, %2196) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2198 = tensor.empty() : tensor<1x64xf32> + %2199 = "ttir.relu"(%2197, %2198) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2200 = tensor.empty() : tensor<1x64xf32> + %2201 = "ttir.relu"(%2199, %2200) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2202 = tensor.empty() : tensor<1x64xf32> + %2203 = "ttir.relu"(%2201, %2202) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2204 = tensor.empty() : tensor<1x64xf32> + %2205 = "ttir.relu"(%2203, %2204) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2206 = tensor.empty() : tensor<1x64xf32> + %2207 = "ttir.relu"(%2205, %2206) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2208 = tensor.empty() : tensor<1x64xf32> + %2209 = "ttir.relu"(%2207, %2208) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2210 = tensor.empty() : tensor<1x64xf32> + %2211 = "ttir.relu"(%2209, %2210) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2212 = tensor.empty() : tensor<1x64xf32> + %2213 = "ttir.relu"(%2211, %2212) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2214 = tensor.empty() : tensor<1x64xf32> + %2215 = "ttir.relu"(%2213, %2214) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2216 = tensor.empty() : tensor<1x64xf32> + %2217 = "ttir.relu"(%2215, %2216) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2218 = tensor.empty() : tensor<1x64xf32> + %2219 = "ttir.relu"(%2217, %2218) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2220 = tensor.empty() : tensor<1x64xf32> + %2221 = "ttir.relu"(%2219, %2220) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2222 = tensor.empty() : tensor<1x64xf32> + %2223 = "ttir.relu"(%2221, %2222) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2224 = tensor.empty() : tensor<1x64xf32> + %2225 = "ttir.relu"(%2223, %2224) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2226 = tensor.empty() : tensor<1x64xf32> + %2227 = "ttir.relu"(%2225, %2226) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2228 = tensor.empty() : tensor<1x64xf32> + %2229 = "ttir.relu"(%2227, %2228) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2230 = tensor.empty() : tensor<1x64xf32> + %2231 = "ttir.relu"(%2229, %2230) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2232 = tensor.empty() : tensor<1x64xf32> + %2233 = "ttir.relu"(%2231, %2232) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2234 = tensor.empty() : tensor<1x64xf32> + %2235 = "ttir.relu"(%2233, %2234) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2236 = tensor.empty() : tensor<1x64xf32> + %2237 = "ttir.relu"(%2235, %2236) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2238 = tensor.empty() : tensor<1x64xf32> + %2239 = "ttir.relu"(%2237, %2238) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2240 = tensor.empty() : tensor<1x64xf32> + %2241 = "ttir.relu"(%2239, %2240) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2242 = tensor.empty() : tensor<1x64xf32> + %2243 = "ttir.relu"(%2241, %2242) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2244 = tensor.empty() : tensor<1x64xf32> + %2245 = "ttir.relu"(%2243, %2244) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2246 = tensor.empty() : tensor<1x64xf32> + %2247 = "ttir.relu"(%2245, %2246) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2248 = tensor.empty() : tensor<1x64xf32> + %2249 = "ttir.relu"(%2247, %2248) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2250 = tensor.empty() : tensor<1x64xf32> + %2251 = "ttir.relu"(%2249, %2250) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2252 = tensor.empty() : tensor<1x64xf32> + %2253 = "ttir.relu"(%2251, %2252) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2254 = tensor.empty() : tensor<1x64xf32> + %2255 = "ttir.relu"(%2253, %2254) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2256 = tensor.empty() : tensor<1x64xf32> + %2257 = "ttir.relu"(%2255, %2256) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2258 = tensor.empty() : tensor<1x64xf32> + %2259 = "ttir.relu"(%2257, %2258) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2260 = tensor.empty() : tensor<1x64xf32> + %2261 = "ttir.relu"(%2259, %2260) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2262 = tensor.empty() : tensor<1x64xf32> + %2263 = "ttir.relu"(%2261, %2262) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2264 = tensor.empty() : tensor<1x64xf32> + %2265 = "ttir.relu"(%2263, %2264) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2266 = tensor.empty() : tensor<1x64xf32> + %2267 = "ttir.relu"(%2265, %2266) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2268 = tensor.empty() : tensor<1x64xf32> + %2269 = "ttir.relu"(%2267, %2268) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2270 = tensor.empty() : tensor<1x64xf32> + %2271 = "ttir.relu"(%2269, %2270) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2272 = tensor.empty() : tensor<1x64xf32> + %2273 = "ttir.relu"(%2271, %2272) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2274 = tensor.empty() : tensor<1x64xf32> + %2275 = "ttir.relu"(%2273, %2274) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2276 = tensor.empty() : tensor<1x64xf32> + %2277 = "ttir.relu"(%2275, %2276) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2278 = tensor.empty() : tensor<1x64xf32> + %2279 = "ttir.relu"(%2277, %2278) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2280 = tensor.empty() : tensor<1x64xf32> + %2281 = "ttir.relu"(%2279, %2280) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2282 = tensor.empty() : tensor<1x64xf32> + %2283 = "ttir.relu"(%2281, %2282) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2284 = tensor.empty() : tensor<1x64xf32> + %2285 = "ttir.relu"(%2283, %2284) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2286 = tensor.empty() : tensor<1x64xf32> + %2287 = "ttir.relu"(%2285, %2286) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2288 = tensor.empty() : tensor<1x64xf32> + %2289 = "ttir.relu"(%2287, %2288) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2290 = tensor.empty() : tensor<1x64xf32> + %2291 = "ttir.relu"(%2289, %2290) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2292 = tensor.empty() : tensor<1x64xf32> + %2293 = "ttir.relu"(%2291, %2292) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2294 = tensor.empty() : tensor<1x64xf32> + %2295 = "ttir.relu"(%2293, %2294) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2296 = tensor.empty() : tensor<1x64xf32> + %2297 = "ttir.relu"(%2295, %2296) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2298 = tensor.empty() : tensor<1x64xf32> + %2299 = "ttir.relu"(%2297, %2298) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2300 = tensor.empty() : tensor<1x64xf32> + %2301 = "ttir.relu"(%2299, %2300) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2302 = tensor.empty() : tensor<1x64xf32> + %2303 = "ttir.relu"(%2301, %2302) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2304 = tensor.empty() : tensor<1x64xf32> + %2305 = "ttir.relu"(%2303, %2304) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2306 = tensor.empty() : tensor<1x64xf32> + %2307 = "ttir.relu"(%2305, %2306) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2308 = tensor.empty() : tensor<1x64xf32> + %2309 = "ttir.relu"(%2307, %2308) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2310 = tensor.empty() : tensor<1x64xf32> + %2311 = "ttir.relu"(%2309, %2310) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2312 = tensor.empty() : tensor<1x64xf32> + %2313 = "ttir.relu"(%2311, %2312) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2314 = tensor.empty() : tensor<1x64xf32> + %2315 = "ttir.relu"(%2313, %2314) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2316 = tensor.empty() : tensor<1x64xf32> + %2317 = "ttir.relu"(%2315, %2316) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2318 = tensor.empty() : tensor<1x64xf32> + %2319 = "ttir.relu"(%2317, %2318) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2320 = tensor.empty() : tensor<1x64xf32> + %2321 = "ttir.relu"(%2319, %2320) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2322 = tensor.empty() : tensor<1x64xf32> + %2323 = "ttir.relu"(%2321, %2322) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2324 = tensor.empty() : tensor<1x64xf32> + %2325 = "ttir.relu"(%2323, %2324) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2326 = tensor.empty() : tensor<1x64xf32> + %2327 = "ttir.relu"(%2325, %2326) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2328 = tensor.empty() : tensor<1x64xf32> + %2329 = "ttir.relu"(%2327, %2328) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2330 = tensor.empty() : tensor<1x64xf32> + %2331 = "ttir.relu"(%2329, %2330) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2332 = tensor.empty() : tensor<1x64xf32> + %2333 = "ttir.relu"(%2331, %2332) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2334 = tensor.empty() : tensor<1x64xf32> + %2335 = "ttir.relu"(%2333, %2334) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2336 = tensor.empty() : tensor<1x64xf32> + %2337 = "ttir.relu"(%2335, %2336) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2338 = tensor.empty() : tensor<1x64xf32> + %2339 = "ttir.relu"(%2337, %2338) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2340 = tensor.empty() : tensor<1x64xf32> + %2341 = "ttir.relu"(%2339, %2340) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2342 = tensor.empty() : tensor<1x64xf32> + %2343 = "ttir.relu"(%2341, %2342) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2344 = tensor.empty() : tensor<1x64xf32> + %2345 = "ttir.relu"(%2343, %2344) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2346 = tensor.empty() : tensor<1x64xf32> + %2347 = "ttir.relu"(%2345, %2346) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2348 = tensor.empty() : tensor<1x64xf32> + %2349 = "ttir.relu"(%2347, %2348) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2350 = tensor.empty() : tensor<1x64xf32> + %2351 = "ttir.relu"(%2349, %2350) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2352 = tensor.empty() : tensor<1x64xf32> + %2353 = "ttir.relu"(%2351, %2352) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2354 = tensor.empty() : tensor<1x64xf32> + %2355 = "ttir.relu"(%2353, %2354) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2356 = tensor.empty() : tensor<1x64xf32> + %2357 = "ttir.relu"(%2355, %2356) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2358 = tensor.empty() : tensor<1x64xf32> + %2359 = "ttir.relu"(%2357, %2358) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2360 = tensor.empty() : tensor<1x64xf32> + %2361 = "ttir.relu"(%2359, %2360) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2362 = tensor.empty() : tensor<1x64xf32> + %2363 = "ttir.relu"(%2361, %2362) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2364 = tensor.empty() : tensor<1x64xf32> + %2365 = "ttir.relu"(%2363, %2364) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2366 = tensor.empty() : tensor<1x64xf32> + %2367 = "ttir.relu"(%2365, %2366) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2368 = tensor.empty() : tensor<1x64xf32> + %2369 = "ttir.relu"(%2367, %2368) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2370 = tensor.empty() : tensor<1x64xf32> + %2371 = "ttir.relu"(%2369, %2370) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2372 = tensor.empty() : tensor<1x64xf32> + %2373 = "ttir.relu"(%2371, %2372) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2374 = tensor.empty() : tensor<1x64xf32> + %2375 = "ttir.relu"(%2373, %2374) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2376 = tensor.empty() : tensor<1x64xf32> + %2377 = "ttir.relu"(%2375, %2376) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2378 = tensor.empty() : tensor<1x64xf32> + %2379 = "ttir.relu"(%2377, %2378) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2380 = tensor.empty() : tensor<1x64xf32> + %2381 = "ttir.relu"(%2379, %2380) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2382 = tensor.empty() : tensor<1x64xf32> + %2383 = "ttir.relu"(%2381, %2382) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2384 = tensor.empty() : tensor<1x64xf32> + %2385 = "ttir.relu"(%2383, %2384) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2386 = tensor.empty() : tensor<1x64xf32> + %2387 = "ttir.relu"(%2385, %2386) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2388 = tensor.empty() : tensor<1x64xf32> + %2389 = "ttir.relu"(%2387, %2388) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2390 = tensor.empty() : tensor<1x64xf32> + %2391 = "ttir.relu"(%2389, %2390) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2392 = tensor.empty() : tensor<1x64xf32> + %2393 = "ttir.relu"(%2391, %2392) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2394 = tensor.empty() : tensor<1x64xf32> + %2395 = "ttir.relu"(%2393, %2394) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2396 = tensor.empty() : tensor<1x64xf32> + %2397 = "ttir.relu"(%2395, %2396) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2398 = tensor.empty() : tensor<1x64xf32> + %2399 = "ttir.relu"(%2397, %2398) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2400 = tensor.empty() : tensor<1x64xf32> + %2401 = "ttir.relu"(%2399, %2400) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2402 = tensor.empty() : tensor<1x64xf32> + %2403 = "ttir.relu"(%2401, %2402) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2404 = tensor.empty() : tensor<1x64xf32> + %2405 = "ttir.relu"(%2403, %2404) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2406 = tensor.empty() : tensor<1x64xf32> + %2407 = "ttir.relu"(%2405, %2406) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2408 = tensor.empty() : tensor<1x64xf32> + %2409 = "ttir.relu"(%2407, %2408) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2410 = tensor.empty() : tensor<1x64xf32> + %2411 = "ttir.relu"(%2409, %2410) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2412 = tensor.empty() : tensor<1x64xf32> + %2413 = "ttir.relu"(%2411, %2412) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2414 = tensor.empty() : tensor<1x64xf32> + %2415 = "ttir.relu"(%2413, %2414) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2416 = tensor.empty() : tensor<1x64xf32> + %2417 = "ttir.relu"(%2415, %2416) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2418 = tensor.empty() : tensor<1x64xf32> + %2419 = "ttir.relu"(%2417, %2418) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2420 = tensor.empty() : tensor<1x64xf32> + %2421 = "ttir.relu"(%2419, %2420) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2422 = tensor.empty() : tensor<1x64xf32> + %2423 = "ttir.relu"(%2421, %2422) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2424 = tensor.empty() : tensor<1x64xf32> + %2425 = "ttir.relu"(%2423, %2424) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2426 = tensor.empty() : tensor<1x64xf32> + %2427 = "ttir.relu"(%2425, %2426) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2428 = tensor.empty() : tensor<1x64xf32> + %2429 = "ttir.relu"(%2427, %2428) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2430 = tensor.empty() : tensor<1x64xf32> + %2431 = "ttir.relu"(%2429, %2430) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2432 = tensor.empty() : tensor<1x64xf32> + %2433 = "ttir.relu"(%2431, %2432) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2434 = tensor.empty() : tensor<1x64xf32> + %2435 = "ttir.relu"(%2433, %2434) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2436 = tensor.empty() : tensor<1x64xf32> + %2437 = "ttir.relu"(%2435, %2436) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2438 = tensor.empty() : tensor<1x64xf32> + %2439 = "ttir.relu"(%2437, %2438) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2440 = tensor.empty() : tensor<1x64xf32> + %2441 = "ttir.relu"(%2439, %2440) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2442 = tensor.empty() : tensor<1x64xf32> + %2443 = "ttir.relu"(%2441, %2442) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2444 = tensor.empty() : tensor<1x64xf32> + %2445 = "ttir.relu"(%2443, %2444) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2446 = tensor.empty() : tensor<1x64xf32> + %2447 = "ttir.relu"(%2445, %2446) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2448 = tensor.empty() : tensor<1x64xf32> + %2449 = "ttir.relu"(%2447, %2448) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2450 = tensor.empty() : tensor<1x64xf32> + %2451 = "ttir.relu"(%2449, %2450) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2452 = tensor.empty() : tensor<1x64xf32> + %2453 = "ttir.relu"(%2451, %2452) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2454 = tensor.empty() : tensor<1x64xf32> + %2455 = "ttir.relu"(%2453, %2454) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2456 = tensor.empty() : tensor<1x64xf32> + %2457 = "ttir.relu"(%2455, %2456) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2458 = tensor.empty() : tensor<1x64xf32> + %2459 = "ttir.relu"(%2457, %2458) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2460 = tensor.empty() : tensor<1x64xf32> + %2461 = "ttir.relu"(%2459, %2460) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2462 = tensor.empty() : tensor<1x64xf32> + %2463 = "ttir.relu"(%2461, %2462) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2464 = tensor.empty() : tensor<1x64xf32> + %2465 = "ttir.relu"(%2463, %2464) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2466 = tensor.empty() : tensor<1x64xf32> + %2467 = "ttir.relu"(%2465, %2466) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2468 = tensor.empty() : tensor<1x64xf32> + %2469 = "ttir.relu"(%2467, %2468) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2470 = tensor.empty() : tensor<1x64xf32> + %2471 = "ttir.relu"(%2469, %2470) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2472 = tensor.empty() : tensor<1x64xf32> + %2473 = "ttir.relu"(%2471, %2472) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2474 = tensor.empty() : tensor<1x64xf32> + %2475 = "ttir.relu"(%2473, %2474) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2476 = tensor.empty() : tensor<1x64xf32> + %2477 = "ttir.relu"(%2475, %2476) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2478 = tensor.empty() : tensor<1x64xf32> + %2479 = "ttir.relu"(%2477, %2478) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2480 = tensor.empty() : tensor<1x64xf32> + %2481 = "ttir.relu"(%2479, %2480) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2482 = tensor.empty() : tensor<1x64xf32> + %2483 = "ttir.relu"(%2481, %2482) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2484 = tensor.empty() : tensor<1x64xf32> + %2485 = "ttir.relu"(%2483, %2484) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2486 = tensor.empty() : tensor<1x64xf32> + %2487 = "ttir.relu"(%2485, %2486) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2488 = tensor.empty() : tensor<1x64xf32> + %2489 = "ttir.relu"(%2487, %2488) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2490 = tensor.empty() : tensor<1x64xf32> + %2491 = "ttir.relu"(%2489, %2490) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2492 = tensor.empty() : tensor<1x64xf32> + %2493 = "ttir.relu"(%2491, %2492) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2494 = tensor.empty() : tensor<1x64xf32> + %2495 = "ttir.relu"(%2493, %2494) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2496 = tensor.empty() : tensor<1x64xf32> + %2497 = "ttir.relu"(%2495, %2496) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2498 = tensor.empty() : tensor<1x64xf32> + %2499 = "ttir.relu"(%2497, %2498) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2500 = tensor.empty() : tensor<1x64xf32> + %2501 = "ttir.relu"(%2499, %2500) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2502 = tensor.empty() : tensor<1x64xf32> + %2503 = "ttir.relu"(%2501, %2502) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2504 = tensor.empty() : tensor<1x64xf32> + %2505 = "ttir.relu"(%2503, %2504) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2506 = tensor.empty() : tensor<1x64xf32> + %2507 = "ttir.relu"(%2505, %2506) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2508 = tensor.empty() : tensor<1x64xf32> + %2509 = "ttir.relu"(%2507, %2508) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2510 = tensor.empty() : tensor<1x64xf32> + %2511 = "ttir.relu"(%2509, %2510) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2512 = tensor.empty() : tensor<1x64xf32> + %2513 = "ttir.relu"(%2511, %2512) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2514 = tensor.empty() : tensor<1x64xf32> + %2515 = "ttir.relu"(%2513, %2514) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2516 = tensor.empty() : tensor<1x64xf32> + %2517 = "ttir.relu"(%2515, %2516) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2518 = tensor.empty() : tensor<1x64xf32> + %2519 = "ttir.relu"(%2517, %2518) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2520 = tensor.empty() : tensor<1x64xf32> + %2521 = "ttir.relu"(%2519, %2520) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2522 = tensor.empty() : tensor<1x64xf32> + %2523 = "ttir.relu"(%2521, %2522) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2524 = tensor.empty() : tensor<1x64xf32> + %2525 = "ttir.relu"(%2523, %2524) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2526 = tensor.empty() : tensor<1x64xf32> + %2527 = "ttir.relu"(%2525, %2526) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2528 = tensor.empty() : tensor<1x64xf32> + %2529 = "ttir.relu"(%2527, %2528) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2530 = tensor.empty() : tensor<1x64xf32> + %2531 = "ttir.relu"(%2529, %2530) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2532 = tensor.empty() : tensor<1x64xf32> + %2533 = "ttir.relu"(%2531, %2532) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2534 = tensor.empty() : tensor<1x64xf32> + %2535 = "ttir.relu"(%2533, %2534) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2536 = tensor.empty() : tensor<1x64xf32> + %2537 = "ttir.relu"(%2535, %2536) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2538 = tensor.empty() : tensor<1x64xf32> + %2539 = "ttir.relu"(%2537, %2538) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2540 = tensor.empty() : tensor<1x64xf32> + %2541 = "ttir.relu"(%2539, %2540) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2542 = tensor.empty() : tensor<1x64xf32> + %2543 = "ttir.relu"(%2541, %2542) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2544 = tensor.empty() : tensor<1x64xf32> + %2545 = "ttir.relu"(%2543, %2544) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2546 = tensor.empty() : tensor<1x64xf32> + %2547 = "ttir.relu"(%2545, %2546) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2548 = tensor.empty() : tensor<1x64xf32> + %2549 = "ttir.relu"(%2547, %2548) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2550 = tensor.empty() : tensor<1x64xf32> + %2551 = "ttir.relu"(%2549, %2550) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2552 = tensor.empty() : tensor<1x64xf32> + %2553 = "ttir.relu"(%2551, %2552) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2554 = tensor.empty() : tensor<1x64xf32> + %2555 = "ttir.relu"(%2553, %2554) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2556 = tensor.empty() : tensor<1x64xf32> + %2557 = "ttir.relu"(%2555, %2556) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2558 = tensor.empty() : tensor<1x64xf32> + %2559 = "ttir.relu"(%2557, %2558) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2560 = tensor.empty() : tensor<1x64xf32> + %2561 = "ttir.relu"(%2559, %2560) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2562 = tensor.empty() : tensor<1x64xf32> + %2563 = "ttir.relu"(%2561, %2562) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2564 = tensor.empty() : tensor<1x64xf32> + %2565 = "ttir.relu"(%2563, %2564) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2566 = tensor.empty() : tensor<1x64xf32> + %2567 = "ttir.relu"(%2565, %2566) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2568 = tensor.empty() : tensor<1x64xf32> + %2569 = "ttir.relu"(%2567, %2568) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2570 = tensor.empty() : tensor<1x64xf32> + %2571 = "ttir.relu"(%2569, %2570) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2572 = tensor.empty() : tensor<1x64xf32> + %2573 = "ttir.relu"(%2571, %2572) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2574 = tensor.empty() : tensor<1x64xf32> + %2575 = "ttir.relu"(%2573, %2574) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2576 = tensor.empty() : tensor<1x64xf32> + %2577 = "ttir.relu"(%2575, %2576) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2578 = tensor.empty() : tensor<1x64xf32> + %2579 = "ttir.relu"(%2577, %2578) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2580 = tensor.empty() : tensor<1x64xf32> + %2581 = "ttir.relu"(%2579, %2580) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2582 = tensor.empty() : tensor<1x64xf32> + %2583 = "ttir.relu"(%2581, %2582) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2584 = tensor.empty() : tensor<1x64xf32> + %2585 = "ttir.relu"(%2583, %2584) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2586 = tensor.empty() : tensor<1x64xf32> + %2587 = "ttir.relu"(%2585, %2586) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2588 = tensor.empty() : tensor<1x64xf32> + %2589 = "ttir.relu"(%2587, %2588) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2590 = tensor.empty() : tensor<1x64xf32> + %2591 = "ttir.relu"(%2589, %2590) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2592 = tensor.empty() : tensor<1x64xf32> + %2593 = "ttir.relu"(%2591, %2592) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2594 = tensor.empty() : tensor<1x64xf32> + %2595 = "ttir.relu"(%2593, %2594) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2596 = tensor.empty() : tensor<1x64xf32> + %2597 = "ttir.relu"(%2595, %2596) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2598 = tensor.empty() : tensor<1x64xf32> + %2599 = "ttir.relu"(%2597, %2598) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2600 = tensor.empty() : tensor<1x64xf32> + %2601 = "ttir.relu"(%2599, %2600) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2602 = tensor.empty() : tensor<1x64xf32> + %2603 = "ttir.relu"(%2601, %2602) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2604 = tensor.empty() : tensor<1x64xf32> + %2605 = "ttir.relu"(%2603, %2604) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2606 = tensor.empty() : tensor<1x64xf32> + %2607 = "ttir.relu"(%2605, %2606) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2608 = tensor.empty() : tensor<1x64xf32> + %2609 = "ttir.relu"(%2607, %2608) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2610 = tensor.empty() : tensor<1x64xf32> + %2611 = "ttir.relu"(%2609, %2610) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2612 = tensor.empty() : tensor<1x64xf32> + %2613 = "ttir.relu"(%2611, %2612) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2614 = tensor.empty() : tensor<1x64xf32> + %2615 = "ttir.relu"(%2613, %2614) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2616 = tensor.empty() : tensor<1x64xf32> + %2617 = "ttir.relu"(%2615, %2616) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2618 = tensor.empty() : tensor<1x64xf32> + %2619 = "ttir.relu"(%2617, %2618) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2620 = tensor.empty() : tensor<1x64xf32> + %2621 = "ttir.relu"(%2619, %2620) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2622 = tensor.empty() : tensor<1x64xf32> + %2623 = "ttir.relu"(%2621, %2622) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2624 = tensor.empty() : tensor<1x64xf32> + %2625 = "ttir.relu"(%2623, %2624) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2626 = tensor.empty() : tensor<1x64xf32> + %2627 = "ttir.relu"(%2625, %2626) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2628 = tensor.empty() : tensor<1x64xf32> + %2629 = "ttir.relu"(%2627, %2628) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2630 = tensor.empty() : tensor<1x64xf32> + %2631 = "ttir.relu"(%2629, %2630) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2632 = tensor.empty() : tensor<1x64xf32> + %2633 = "ttir.relu"(%2631, %2632) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2634 = tensor.empty() : tensor<1x64xf32> + %2635 = "ttir.relu"(%2633, %2634) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2636 = tensor.empty() : tensor<1x64xf32> + %2637 = "ttir.relu"(%2635, %2636) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2638 = tensor.empty() : tensor<1x64xf32> + %2639 = "ttir.relu"(%2637, %2638) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2640 = tensor.empty() : tensor<1x64xf32> + %2641 = "ttir.relu"(%2639, %2640) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2642 = tensor.empty() : tensor<1x64xf32> + %2643 = "ttir.relu"(%2641, %2642) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2644 = tensor.empty() : tensor<1x64xf32> + %2645 = "ttir.relu"(%2643, %2644) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2646 = tensor.empty() : tensor<1x64xf32> + %2647 = "ttir.relu"(%2645, %2646) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2648 = tensor.empty() : tensor<1x64xf32> + %2649 = "ttir.relu"(%2647, %2648) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2650 = tensor.empty() : tensor<1x64xf32> + %2651 = "ttir.relu"(%2649, %2650) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2652 = tensor.empty() : tensor<1x64xf32> + %2653 = "ttir.relu"(%2651, %2652) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2654 = tensor.empty() : tensor<1x64xf32> + %2655 = "ttir.relu"(%2653, %2654) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2656 = tensor.empty() : tensor<1x64xf32> + %2657 = "ttir.relu"(%2655, %2656) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2658 = tensor.empty() : tensor<1x64xf32> + %2659 = "ttir.relu"(%2657, %2658) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2660 = tensor.empty() : tensor<1x64xf32> + %2661 = "ttir.relu"(%2659, %2660) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2662 = tensor.empty() : tensor<1x64xf32> + %2663 = "ttir.relu"(%2661, %2662) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2664 = tensor.empty() : tensor<1x64xf32> + %2665 = "ttir.relu"(%2663, %2664) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2666 = tensor.empty() : tensor<1x64xf32> + %2667 = "ttir.relu"(%2665, %2666) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2668 = tensor.empty() : tensor<1x64xf32> + %2669 = "ttir.relu"(%2667, %2668) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2670 = tensor.empty() : tensor<1x64xf32> + %2671 = "ttir.relu"(%2669, %2670) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2672 = tensor.empty() : tensor<1x64xf32> + %2673 = "ttir.relu"(%2671, %2672) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2674 = tensor.empty() : tensor<1x64xf32> + %2675 = "ttir.relu"(%2673, %2674) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2676 = tensor.empty() : tensor<1x64xf32> + %2677 = "ttir.relu"(%2675, %2676) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2678 = tensor.empty() : tensor<1x64xf32> + %2679 = "ttir.relu"(%2677, %2678) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2680 = tensor.empty() : tensor<1x64xf32> + %2681 = "ttir.relu"(%2679, %2680) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2682 = tensor.empty() : tensor<1x64xf32> + %2683 = "ttir.relu"(%2681, %2682) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2684 = tensor.empty() : tensor<1x64xf32> + %2685 = "ttir.relu"(%2683, %2684) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2686 = tensor.empty() : tensor<1x64xf32> + %2687 = "ttir.relu"(%2685, %2686) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2688 = tensor.empty() : tensor<1x64xf32> + %2689 = "ttir.relu"(%2687, %2688) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2690 = tensor.empty() : tensor<1x64xf32> + %2691 = "ttir.relu"(%2689, %2690) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2692 = tensor.empty() : tensor<1x64xf32> + %2693 = "ttir.relu"(%2691, %2692) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2694 = tensor.empty() : tensor<1x64xf32> + %2695 = "ttir.relu"(%2693, %2694) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2696 = tensor.empty() : tensor<1x64xf32> + %2697 = "ttir.relu"(%2695, %2696) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2698 = tensor.empty() : tensor<1x64xf32> + %2699 = "ttir.relu"(%2697, %2698) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2700 = tensor.empty() : tensor<1x64xf32> + %2701 = "ttir.relu"(%2699, %2700) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2702 = tensor.empty() : tensor<1x64xf32> + %2703 = "ttir.relu"(%2701, %2702) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2704 = tensor.empty() : tensor<1x64xf32> + %2705 = "ttir.relu"(%2703, %2704) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2706 = tensor.empty() : tensor<1x64xf32> + %2707 = "ttir.relu"(%2705, %2706) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2708 = tensor.empty() : tensor<1x64xf32> + %2709 = "ttir.relu"(%2707, %2708) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2710 = tensor.empty() : tensor<1x64xf32> + %2711 = "ttir.relu"(%2709, %2710) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2712 = tensor.empty() : tensor<1x64xf32> + %2713 = "ttir.relu"(%2711, %2712) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2714 = tensor.empty() : tensor<1x64xf32> + %2715 = "ttir.relu"(%2713, %2714) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2716 = tensor.empty() : tensor<1x64xf32> + %2717 = "ttir.relu"(%2715, %2716) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2718 = tensor.empty() : tensor<1x64xf32> + %2719 = "ttir.relu"(%2717, %2718) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2720 = tensor.empty() : tensor<1x64xf32> + %2721 = "ttir.relu"(%2719, %2720) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2722 = tensor.empty() : tensor<1x64xf32> + %2723 = "ttir.relu"(%2721, %2722) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2724 = tensor.empty() : tensor<1x64xf32> + %2725 = "ttir.relu"(%2723, %2724) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2726 = tensor.empty() : tensor<1x64xf32> + %2727 = "ttir.relu"(%2725, %2726) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2728 = tensor.empty() : tensor<1x64xf32> + %2729 = "ttir.relu"(%2727, %2728) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2730 = tensor.empty() : tensor<1x64xf32> + %2731 = "ttir.relu"(%2729, %2730) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2732 = tensor.empty() : tensor<1x64xf32> + %2733 = "ttir.relu"(%2731, %2732) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2734 = tensor.empty() : tensor<1x64xf32> + %2735 = "ttir.relu"(%2733, %2734) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2736 = tensor.empty() : tensor<1x64xf32> + %2737 = "ttir.relu"(%2735, %2736) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2738 = tensor.empty() : tensor<1x64xf32> + %2739 = "ttir.relu"(%2737, %2738) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2740 = tensor.empty() : tensor<1x64xf32> + %2741 = "ttir.relu"(%2739, %2740) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2742 = tensor.empty() : tensor<1x64xf32> + %2743 = "ttir.relu"(%2741, %2742) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2744 = tensor.empty() : tensor<1x64xf32> + %2745 = "ttir.relu"(%2743, %2744) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2746 = tensor.empty() : tensor<1x64xf32> + %2747 = "ttir.relu"(%2745, %2746) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2748 = tensor.empty() : tensor<1x64xf32> + %2749 = "ttir.relu"(%2747, %2748) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2750 = tensor.empty() : tensor<1x64xf32> + %2751 = "ttir.relu"(%2749, %2750) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2752 = tensor.empty() : tensor<1x64xf32> + %2753 = "ttir.relu"(%2751, %2752) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2754 = tensor.empty() : tensor<1x64xf32> + %2755 = "ttir.relu"(%2753, %2754) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2756 = tensor.empty() : tensor<1x64xf32> + %2757 = "ttir.relu"(%2755, %2756) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2758 = tensor.empty() : tensor<1x64xf32> + %2759 = "ttir.relu"(%2757, %2758) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2760 = tensor.empty() : tensor<1x64xf32> + %2761 = "ttir.relu"(%2759, %2760) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2762 = tensor.empty() : tensor<1x64xf32> + %2763 = "ttir.relu"(%2761, %2762) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2764 = tensor.empty() : tensor<1x64xf32> + %2765 = "ttir.relu"(%2763, %2764) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2766 = tensor.empty() : tensor<1x64xf32> + %2767 = "ttir.relu"(%2765, %2766) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2768 = tensor.empty() : tensor<1x64xf32> + %2769 = "ttir.relu"(%2767, %2768) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2770 = tensor.empty() : tensor<1x64xf32> + %2771 = "ttir.relu"(%2769, %2770) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2772 = tensor.empty() : tensor<1x64xf32> + %2773 = "ttir.relu"(%2771, %2772) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2774 = tensor.empty() : tensor<1x64xf32> + %2775 = "ttir.relu"(%2773, %2774) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2776 = tensor.empty() : tensor<1x64xf32> + %2777 = "ttir.relu"(%2775, %2776) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2778 = tensor.empty() : tensor<1x64xf32> + %2779 = "ttir.relu"(%2777, %2778) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2780 = tensor.empty() : tensor<1x64xf32> + %2781 = "ttir.relu"(%2779, %2780) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2782 = tensor.empty() : tensor<1x64xf32> + %2783 = "ttir.relu"(%2781, %2782) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2784 = tensor.empty() : tensor<1x64xf32> + %2785 = "ttir.relu"(%2783, %2784) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2786 = tensor.empty() : tensor<1x64xf32> + %2787 = "ttir.relu"(%2785, %2786) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2788 = tensor.empty() : tensor<1x64xf32> + %2789 = "ttir.relu"(%2787, %2788) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2790 = tensor.empty() : tensor<1x64xf32> + %2791 = "ttir.relu"(%2789, %2790) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2792 = tensor.empty() : tensor<1x64xf32> + %2793 = "ttir.relu"(%2791, %2792) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2794 = tensor.empty() : tensor<1x64xf32> + %2795 = "ttir.relu"(%2793, %2794) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2796 = tensor.empty() : tensor<1x64xf32> + %2797 = "ttir.relu"(%2795, %2796) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2798 = tensor.empty() : tensor<1x64xf32> + %2799 = "ttir.relu"(%2797, %2798) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2800 = tensor.empty() : tensor<1x64xf32> + %2801 = "ttir.relu"(%2799, %2800) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2802 = tensor.empty() : tensor<1x64xf32> + %2803 = "ttir.relu"(%2801, %2802) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2804 = tensor.empty() : tensor<1x64xf32> + %2805 = "ttir.relu"(%2803, %2804) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2806 = tensor.empty() : tensor<1x64xf32> + %2807 = "ttir.relu"(%2805, %2806) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2808 = tensor.empty() : tensor<1x64xf32> + %2809 = "ttir.relu"(%2807, %2808) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2810 = tensor.empty() : tensor<1x64xf32> + %2811 = "ttir.relu"(%2809, %2810) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2812 = tensor.empty() : tensor<1x64xf32> + %2813 = "ttir.relu"(%2811, %2812) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2814 = tensor.empty() : tensor<1x64xf32> + %2815 = "ttir.relu"(%2813, %2814) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2816 = tensor.empty() : tensor<1x64xf32> + %2817 = "ttir.relu"(%2815, %2816) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2818 = tensor.empty() : tensor<1x64xf32> + %2819 = "ttir.relu"(%2817, %2818) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2820 = tensor.empty() : tensor<1x64xf32> + %2821 = "ttir.relu"(%2819, %2820) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2822 = tensor.empty() : tensor<1x64xf32> + %2823 = "ttir.relu"(%2821, %2822) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2824 = tensor.empty() : tensor<1x64xf32> + %2825 = "ttir.relu"(%2823, %2824) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2826 = tensor.empty() : tensor<1x64xf32> + %2827 = "ttir.relu"(%2825, %2826) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2828 = tensor.empty() : tensor<1x64xf32> + %2829 = "ttir.relu"(%2827, %2828) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2830 = tensor.empty() : tensor<1x64xf32> + %2831 = "ttir.relu"(%2829, %2830) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2832 = tensor.empty() : tensor<1x64xf32> + %2833 = "ttir.relu"(%2831, %2832) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2834 = tensor.empty() : tensor<1x64xf32> + %2835 = "ttir.relu"(%2833, %2834) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2836 = tensor.empty() : tensor<1x64xf32> + %2837 = "ttir.relu"(%2835, %2836) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2838 = tensor.empty() : tensor<1x64xf32> + %2839 = "ttir.relu"(%2837, %2838) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2840 = tensor.empty() : tensor<1x64xf32> + %2841 = "ttir.relu"(%2839, %2840) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2842 = tensor.empty() : tensor<1x64xf32> + %2843 = "ttir.relu"(%2841, %2842) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2844 = tensor.empty() : tensor<1x64xf32> + %2845 = "ttir.relu"(%2843, %2844) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2846 = tensor.empty() : tensor<1x64xf32> + %2847 = "ttir.relu"(%2845, %2846) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2848 = tensor.empty() : tensor<1x64xf32> + %2849 = "ttir.relu"(%2847, %2848) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2850 = tensor.empty() : tensor<1x64xf32> + %2851 = "ttir.relu"(%2849, %2850) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2852 = tensor.empty() : tensor<1x64xf32> + %2853 = "ttir.relu"(%2851, %2852) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2854 = tensor.empty() : tensor<1x64xf32> + %2855 = "ttir.relu"(%2853, %2854) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2856 = tensor.empty() : tensor<1x64xf32> + %2857 = "ttir.relu"(%2855, %2856) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2858 = tensor.empty() : tensor<1x64xf32> + %2859 = "ttir.relu"(%2857, %2858) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2860 = tensor.empty() : tensor<1x64xf32> + %2861 = "ttir.relu"(%2859, %2860) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2862 = tensor.empty() : tensor<1x64xf32> + %2863 = "ttir.relu"(%2861, %2862) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2864 = tensor.empty() : tensor<1x64xf32> + %2865 = "ttir.relu"(%2863, %2864) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2866 = tensor.empty() : tensor<1x64xf32> + %2867 = "ttir.relu"(%2865, %2866) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2868 = tensor.empty() : tensor<1x64xf32> + %2869 = "ttir.relu"(%2867, %2868) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2870 = tensor.empty() : tensor<1x64xf32> + %2871 = "ttir.relu"(%2869, %2870) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2872 = tensor.empty() : tensor<1x64xf32> + %2873 = "ttir.relu"(%2871, %2872) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2874 = tensor.empty() : tensor<1x64xf32> + %2875 = "ttir.relu"(%2873, %2874) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2876 = tensor.empty() : tensor<1x64xf32> + %2877 = "ttir.relu"(%2875, %2876) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2878 = tensor.empty() : tensor<1x64xf32> + %2879 = "ttir.relu"(%2877, %2878) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2880 = tensor.empty() : tensor<1x64xf32> + %2881 = "ttir.relu"(%2879, %2880) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2882 = tensor.empty() : tensor<1x64xf32> + %2883 = "ttir.relu"(%2881, %2882) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2884 = tensor.empty() : tensor<1x64xf32> + %2885 = "ttir.relu"(%2883, %2884) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2886 = tensor.empty() : tensor<1x64xf32> + %2887 = "ttir.relu"(%2885, %2886) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2888 = tensor.empty() : tensor<1x64xf32> + %2889 = "ttir.relu"(%2887, %2888) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2890 = tensor.empty() : tensor<1x64xf32> + %2891 = "ttir.relu"(%2889, %2890) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2892 = tensor.empty() : tensor<1x64xf32> + %2893 = "ttir.relu"(%2891, %2892) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2894 = tensor.empty() : tensor<1x64xf32> + %2895 = "ttir.relu"(%2893, %2894) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2896 = tensor.empty() : tensor<1x64xf32> + %2897 = "ttir.relu"(%2895, %2896) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2898 = tensor.empty() : tensor<1x64xf32> + %2899 = "ttir.relu"(%2897, %2898) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2900 = tensor.empty() : tensor<1x64xf32> + %2901 = "ttir.relu"(%2899, %2900) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2902 = tensor.empty() : tensor<1x64xf32> + %2903 = "ttir.relu"(%2901, %2902) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2904 = tensor.empty() : tensor<1x64xf32> + %2905 = "ttir.relu"(%2903, %2904) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2906 = tensor.empty() : tensor<1x64xf32> + %2907 = "ttir.relu"(%2905, %2906) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2908 = tensor.empty() : tensor<1x64xf32> + %2909 = "ttir.relu"(%2907, %2908) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2910 = tensor.empty() : tensor<1x64xf32> + %2911 = "ttir.relu"(%2909, %2910) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2912 = tensor.empty() : tensor<1x64xf32> + %2913 = "ttir.relu"(%2911, %2912) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2914 = tensor.empty() : tensor<1x64xf32> + %2915 = "ttir.relu"(%2913, %2914) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2916 = tensor.empty() : tensor<1x64xf32> + %2917 = "ttir.relu"(%2915, %2916) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2918 = tensor.empty() : tensor<1x64xf32> + %2919 = "ttir.relu"(%2917, %2918) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2920 = tensor.empty() : tensor<1x64xf32> + %2921 = "ttir.relu"(%2919, %2920) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2922 = tensor.empty() : tensor<1x64xf32> + %2923 = "ttir.relu"(%2921, %2922) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2924 = tensor.empty() : tensor<1x64xf32> + %2925 = "ttir.relu"(%2923, %2924) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2926 = tensor.empty() : tensor<1x64xf32> + %2927 = "ttir.relu"(%2925, %2926) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2928 = tensor.empty() : tensor<1x64xf32> + %2929 = "ttir.relu"(%2927, %2928) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2930 = tensor.empty() : tensor<1x64xf32> + %2931 = "ttir.relu"(%2929, %2930) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2932 = tensor.empty() : tensor<1x64xf32> + %2933 = "ttir.relu"(%2931, %2932) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2934 = tensor.empty() : tensor<1x64xf32> + %2935 = "ttir.relu"(%2933, %2934) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2936 = tensor.empty() : tensor<1x64xf32> + %2937 = "ttir.relu"(%2935, %2936) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2938 = tensor.empty() : tensor<1x64xf32> + %2939 = "ttir.relu"(%2937, %2938) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2940 = tensor.empty() : tensor<1x64xf32> + %2941 = "ttir.relu"(%2939, %2940) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2942 = tensor.empty() : tensor<1x64xf32> + %2943 = "ttir.relu"(%2941, %2942) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2944 = tensor.empty() : tensor<1x64xf32> + %2945 = "ttir.relu"(%2943, %2944) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2946 = tensor.empty() : tensor<1x64xf32> + %2947 = "ttir.relu"(%2945, %2946) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2948 = tensor.empty() : tensor<1x64xf32> + %2949 = "ttir.relu"(%2947, %2948) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2950 = tensor.empty() : tensor<1x64xf32> + %2951 = "ttir.relu"(%2949, %2950) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2952 = tensor.empty() : tensor<1x64xf32> + %2953 = "ttir.relu"(%2951, %2952) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2954 = tensor.empty() : tensor<1x64xf32> + %2955 = "ttir.relu"(%2953, %2954) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2956 = tensor.empty() : tensor<1x64xf32> + %2957 = "ttir.relu"(%2955, %2956) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2958 = tensor.empty() : tensor<1x64xf32> + %2959 = "ttir.relu"(%2957, %2958) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2960 = tensor.empty() : tensor<1x64xf32> + %2961 = "ttir.relu"(%2959, %2960) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2962 = tensor.empty() : tensor<1x64xf32> + %2963 = "ttir.relu"(%2961, %2962) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2964 = tensor.empty() : tensor<1x64xf32> + %2965 = "ttir.relu"(%2963, %2964) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2966 = tensor.empty() : tensor<1x64xf32> + %2967 = "ttir.relu"(%2965, %2966) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2968 = tensor.empty() : tensor<1x64xf32> + %2969 = "ttir.relu"(%2967, %2968) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2970 = tensor.empty() : tensor<1x64xf32> + %2971 = "ttir.relu"(%2969, %2970) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2972 = tensor.empty() : tensor<1x64xf32> + %2973 = "ttir.relu"(%2971, %2972) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2974 = tensor.empty() : tensor<1x64xf32> + %2975 = "ttir.relu"(%2973, %2974) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2976 = tensor.empty() : tensor<1x64xf32> + %2977 = "ttir.relu"(%2975, %2976) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2978 = tensor.empty() : tensor<1x64xf32> + %2979 = "ttir.relu"(%2977, %2978) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2980 = tensor.empty() : tensor<1x64xf32> + %2981 = "ttir.relu"(%2979, %2980) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2982 = tensor.empty() : tensor<1x64xf32> + %2983 = "ttir.relu"(%2981, %2982) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2984 = tensor.empty() : tensor<1x64xf32> + %2985 = "ttir.relu"(%2983, %2984) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2986 = tensor.empty() : tensor<1x64xf32> + %2987 = "ttir.relu"(%2985, %2986) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2988 = tensor.empty() : tensor<1x64xf32> + %2989 = "ttir.relu"(%2987, %2988) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2990 = tensor.empty() : tensor<1x64xf32> + %2991 = "ttir.relu"(%2989, %2990) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2992 = tensor.empty() : tensor<1x64xf32> + %2993 = "ttir.relu"(%2991, %2992) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2994 = tensor.empty() : tensor<1x64xf32> + %2995 = "ttir.relu"(%2993, %2994) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2996 = tensor.empty() : tensor<1x64xf32> + %2997 = "ttir.relu"(%2995, %2996) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2998 = tensor.empty() : tensor<1x64xf32> + %2999 = "ttir.relu"(%2997, %2998) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3000 = tensor.empty() : tensor<1x64xf32> + %3001 = "ttir.relu"(%2999, %3000) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3002 = tensor.empty() : tensor<1x64xf32> + %3003 = "ttir.relu"(%3001, %3002) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3004 = tensor.empty() : tensor<1x64xf32> + %3005 = "ttir.relu"(%3003, %3004) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3006 = tensor.empty() : tensor<1x64xf32> + %3007 = "ttir.relu"(%3005, %3006) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3008 = tensor.empty() : tensor<1x64xf32> + %3009 = "ttir.relu"(%3007, %3008) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3010 = tensor.empty() : tensor<1x64xf32> + %3011 = "ttir.relu"(%3009, %3010) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3012 = tensor.empty() : tensor<1x64xf32> + %3013 = "ttir.relu"(%3011, %3012) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3014 = tensor.empty() : tensor<1x64xf32> + %3015 = "ttir.relu"(%3013, %3014) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3016 = tensor.empty() : tensor<1x64xf32> + %3017 = "ttir.relu"(%3015, %3016) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3018 = tensor.empty() : tensor<1x64xf32> + %3019 = "ttir.relu"(%3017, %3018) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3020 = tensor.empty() : tensor<1x64xf32> + %3021 = "ttir.relu"(%3019, %3020) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3022 = tensor.empty() : tensor<1x64xf32> + %3023 = "ttir.relu"(%3021, %3022) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3024 = tensor.empty() : tensor<1x64xf32> + %3025 = "ttir.relu"(%3023, %3024) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3026 = tensor.empty() : tensor<1x64xf32> + %3027 = "ttir.relu"(%3025, %3026) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3028 = tensor.empty() : tensor<1x64xf32> + %3029 = "ttir.relu"(%3027, %3028) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3030 = tensor.empty() : tensor<1x64xf32> + %3031 = "ttir.relu"(%3029, %3030) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3032 = tensor.empty() : tensor<1x64xf32> + %3033 = "ttir.relu"(%3031, %3032) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3034 = tensor.empty() : tensor<1x64xf32> + %3035 = "ttir.relu"(%3033, %3034) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3036 = tensor.empty() : tensor<1x64xf32> + %3037 = "ttir.relu"(%3035, %3036) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3038 = tensor.empty() : tensor<1x64xf32> + %3039 = "ttir.relu"(%3037, %3038) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3040 = tensor.empty() : tensor<1x64xf32> + %3041 = "ttir.relu"(%3039, %3040) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3042 = tensor.empty() : tensor<1x64xf32> + %3043 = "ttir.relu"(%3041, %3042) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3044 = tensor.empty() : tensor<1x64xf32> + %3045 = "ttir.relu"(%3043, %3044) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3046 = tensor.empty() : tensor<1x64xf32> + %3047 = "ttir.relu"(%3045, %3046) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3048 = tensor.empty() : tensor<1x64xf32> + %3049 = "ttir.relu"(%3047, %3048) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3050 = tensor.empty() : tensor<1x64xf32> + %3051 = "ttir.relu"(%3049, %3050) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3052 = tensor.empty() : tensor<1x64xf32> + %3053 = "ttir.relu"(%3051, %3052) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3054 = tensor.empty() : tensor<1x64xf32> + %3055 = "ttir.relu"(%3053, %3054) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3056 = tensor.empty() : tensor<1x64xf32> + %3057 = "ttir.relu"(%3055, %3056) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3058 = tensor.empty() : tensor<1x64xf32> + %3059 = "ttir.relu"(%3057, %3058) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3060 = tensor.empty() : tensor<1x64xf32> + %3061 = "ttir.relu"(%3059, %3060) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3062 = tensor.empty() : tensor<1x64xf32> + %3063 = "ttir.relu"(%3061, %3062) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3064 = tensor.empty() : tensor<1x64xf32> + %3065 = "ttir.relu"(%3063, %3064) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3066 = tensor.empty() : tensor<1x64xf32> + %3067 = "ttir.relu"(%3065, %3066) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3068 = tensor.empty() : tensor<1x64xf32> + %3069 = "ttir.relu"(%3067, %3068) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3070 = tensor.empty() : tensor<1x64xf32> + %3071 = "ttir.relu"(%3069, %3070) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3072 = tensor.empty() : tensor<1x64xf32> + %3073 = "ttir.relu"(%3071, %3072) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3074 = tensor.empty() : tensor<1x64xf32> + %3075 = "ttir.relu"(%3073, %3074) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3076 = tensor.empty() : tensor<1x64xf32> + %3077 = "ttir.relu"(%3075, %3076) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3078 = tensor.empty() : tensor<1x64xf32> + %3079 = "ttir.relu"(%3077, %3078) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3080 = tensor.empty() : tensor<1x64xf32> + %3081 = "ttir.relu"(%3079, %3080) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3082 = tensor.empty() : tensor<1x64xf32> + %3083 = "ttir.relu"(%3081, %3082) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3084 = tensor.empty() : tensor<1x64xf32> + %3085 = "ttir.relu"(%3083, %3084) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3086 = tensor.empty() : tensor<1x64xf32> + %3087 = "ttir.relu"(%3085, %3086) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3088 = tensor.empty() : tensor<1x64xf32> + %3089 = "ttir.relu"(%3087, %3088) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3090 = tensor.empty() : tensor<1x64xf32> + %3091 = "ttir.relu"(%3089, %3090) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3092 = tensor.empty() : tensor<1x64xf32> + %3093 = "ttir.relu"(%3091, %3092) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3094 = tensor.empty() : tensor<1x64xf32> + %3095 = "ttir.relu"(%3093, %3094) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3096 = tensor.empty() : tensor<1x64xf32> + %3097 = "ttir.relu"(%3095, %3096) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3098 = tensor.empty() : tensor<1x64xf32> + %3099 = "ttir.relu"(%3097, %3098) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3100 = tensor.empty() : tensor<1x64xf32> + %3101 = "ttir.relu"(%3099, %3100) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3102 = tensor.empty() : tensor<1x64xf32> + %3103 = "ttir.relu"(%3101, %3102) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3104 = tensor.empty() : tensor<1x64xf32> + %3105 = "ttir.relu"(%3103, %3104) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3106 = tensor.empty() : tensor<1x64xf32> + %3107 = "ttir.relu"(%3105, %3106) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3108 = tensor.empty() : tensor<1x64xf32> + %3109 = "ttir.relu"(%3107, %3108) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3110 = tensor.empty() : tensor<1x64xf32> + %3111 = "ttir.relu"(%3109, %3110) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3112 = tensor.empty() : tensor<1x64xf32> + %3113 = "ttir.relu"(%3111, %3112) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3114 = tensor.empty() : tensor<1x64xf32> + %3115 = "ttir.relu"(%3113, %3114) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3116 = tensor.empty() : tensor<1x64xf32> + %3117 = "ttir.relu"(%3115, %3116) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3118 = tensor.empty() : tensor<1x64xf32> + %3119 = "ttir.relu"(%3117, %3118) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3120 = tensor.empty() : tensor<1x64xf32> + %3121 = "ttir.relu"(%3119, %3120) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3122 = tensor.empty() : tensor<1x64xf32> + %3123 = "ttir.relu"(%3121, %3122) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3124 = tensor.empty() : tensor<1x64xf32> + %3125 = "ttir.relu"(%3123, %3124) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3126 = tensor.empty() : tensor<1x64xf32> + %3127 = "ttir.relu"(%3125, %3126) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3128 = tensor.empty() : tensor<1x64xf32> + %3129 = "ttir.relu"(%3127, %3128) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3130 = tensor.empty() : tensor<1x64xf32> + %3131 = "ttir.relu"(%3129, %3130) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3132 = tensor.empty() : tensor<1x64xf32> + %3133 = "ttir.relu"(%3131, %3132) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3134 = tensor.empty() : tensor<1x64xf32> + %3135 = "ttir.relu"(%3133, %3134) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3136 = tensor.empty() : tensor<1x64xf32> + %3137 = "ttir.relu"(%3135, %3136) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3138 = tensor.empty() : tensor<1x64xf32> + %3139 = "ttir.relu"(%3137, %3138) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3140 = tensor.empty() : tensor<1x64xf32> + %3141 = "ttir.relu"(%3139, %3140) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3142 = tensor.empty() : tensor<1x64xf32> + %3143 = "ttir.relu"(%3141, %3142) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3144 = tensor.empty() : tensor<1x64xf32> + %3145 = "ttir.relu"(%3143, %3144) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3146 = tensor.empty() : tensor<1x64xf32> + %3147 = "ttir.relu"(%3145, %3146) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3148 = tensor.empty() : tensor<1x64xf32> + %3149 = "ttir.relu"(%3147, %3148) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3150 = tensor.empty() : tensor<1x64xf32> + %3151 = "ttir.relu"(%3149, %3150) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3152 = tensor.empty() : tensor<1x64xf32> + %3153 = "ttir.relu"(%3151, %3152) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3154 = tensor.empty() : tensor<1x64xf32> + %3155 = "ttir.relu"(%3153, %3154) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3156 = tensor.empty() : tensor<1x64xf32> + %3157 = "ttir.relu"(%3155, %3156) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3158 = tensor.empty() : tensor<1x64xf32> + %3159 = "ttir.relu"(%3157, %3158) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3160 = tensor.empty() : tensor<1x64xf32> + %3161 = "ttir.relu"(%3159, %3160) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3162 = tensor.empty() : tensor<1x64xf32> + %3163 = "ttir.relu"(%3161, %3162) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3164 = tensor.empty() : tensor<1x64xf32> + %3165 = "ttir.relu"(%3163, %3164) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3166 = tensor.empty() : tensor<1x64xf32> + %3167 = "ttir.relu"(%3165, %3166) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3168 = tensor.empty() : tensor<1x64xf32> + %3169 = "ttir.relu"(%3167, %3168) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3170 = tensor.empty() : tensor<1x64xf32> + %3171 = "ttir.relu"(%3169, %3170) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3172 = tensor.empty() : tensor<1x64xf32> + %3173 = "ttir.relu"(%3171, %3172) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3174 = tensor.empty() : tensor<1x64xf32> + %3175 = "ttir.relu"(%3173, %3174) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3176 = tensor.empty() : tensor<1x64xf32> + %3177 = "ttir.relu"(%3175, %3176) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3178 = tensor.empty() : tensor<1x64xf32> + %3179 = "ttir.relu"(%3177, %3178) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3180 = tensor.empty() : tensor<1x64xf32> + %3181 = "ttir.relu"(%3179, %3180) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3182 = tensor.empty() : tensor<1x64xf32> + %3183 = "ttir.relu"(%3181, %3182) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3184 = tensor.empty() : tensor<1x64xf32> + %3185 = "ttir.relu"(%3183, %3184) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3186 = tensor.empty() : tensor<1x64xf32> + %3187 = "ttir.relu"(%3185, %3186) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3188 = tensor.empty() : tensor<1x64xf32> + %3189 = "ttir.relu"(%3187, %3188) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3190 = tensor.empty() : tensor<1x64xf32> + %3191 = "ttir.relu"(%3189, %3190) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3192 = tensor.empty() : tensor<1x64xf32> + %3193 = "ttir.relu"(%3191, %3192) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3194 = tensor.empty() : tensor<1x64xf32> + %3195 = "ttir.relu"(%3193, %3194) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3196 = tensor.empty() : tensor<1x64xf32> + %3197 = "ttir.relu"(%3195, %3196) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3198 = tensor.empty() : tensor<1x64xf32> + %3199 = "ttir.relu"(%3197, %3198) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3200 = tensor.empty() : tensor<1x64xf32> + %3201 = "ttir.relu"(%3199, %3200) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3202 = tensor.empty() : tensor<1x64xf32> + %3203 = "ttir.relu"(%3201, %3202) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3204 = tensor.empty() : tensor<1x64xf32> + %3205 = "ttir.relu"(%3203, %3204) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3206 = tensor.empty() : tensor<1x64xf32> + %3207 = "ttir.relu"(%3205, %3206) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3208 = tensor.empty() : tensor<1x64xf32> + %3209 = "ttir.relu"(%3207, %3208) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3210 = tensor.empty() : tensor<1x64xf32> + %3211 = "ttir.relu"(%3209, %3210) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3212 = tensor.empty() : tensor<1x64xf32> + %3213 = "ttir.relu"(%3211, %3212) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3214 = tensor.empty() : tensor<1x64xf32> + %3215 = "ttir.relu"(%3213, %3214) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3216 = tensor.empty() : tensor<1x64xf32> + %3217 = "ttir.relu"(%3215, %3216) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3218 = tensor.empty() : tensor<1x64xf32> + %3219 = "ttir.relu"(%3217, %3218) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3220 = tensor.empty() : tensor<1x64xf32> + %3221 = "ttir.relu"(%3219, %3220) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3222 = tensor.empty() : tensor<1x64xf32> + %3223 = "ttir.relu"(%3221, %3222) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3224 = tensor.empty() : tensor<1x64xf32> + %3225 = "ttir.relu"(%3223, %3224) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3226 = tensor.empty() : tensor<1x64xf32> + %3227 = "ttir.relu"(%3225, %3226) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3228 = tensor.empty() : tensor<1x64xf32> + %3229 = "ttir.relu"(%3227, %3228) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3230 = tensor.empty() : tensor<1x64xf32> + %3231 = "ttir.relu"(%3229, %3230) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3232 = tensor.empty() : tensor<1x64xf32> + %3233 = "ttir.relu"(%3231, %3232) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3234 = tensor.empty() : tensor<1x64xf32> + %3235 = "ttir.relu"(%3233, %3234) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3236 = tensor.empty() : tensor<1x64xf32> + %3237 = "ttir.relu"(%3235, %3236) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3238 = tensor.empty() : tensor<1x64xf32> + %3239 = "ttir.relu"(%3237, %3238) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3240 = tensor.empty() : tensor<1x64xf32> + %3241 = "ttir.relu"(%3239, %3240) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3242 = tensor.empty() : tensor<1x64xf32> + %3243 = "ttir.relu"(%3241, %3242) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3244 = tensor.empty() : tensor<1x64xf32> + %3245 = "ttir.relu"(%3243, %3244) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3246 = tensor.empty() : tensor<1x64xf32> + %3247 = "ttir.relu"(%3245, %3246) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3248 = tensor.empty() : tensor<1x64xf32> + %3249 = "ttir.relu"(%3247, %3248) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3250 = tensor.empty() : tensor<1x64xf32> + %3251 = "ttir.relu"(%3249, %3250) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3252 = tensor.empty() : tensor<1x64xf32> + %3253 = "ttir.relu"(%3251, %3252) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3254 = tensor.empty() : tensor<1x64xf32> + %3255 = "ttir.relu"(%3253, %3254) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3256 = tensor.empty() : tensor<1x64xf32> + %3257 = "ttir.relu"(%3255, %3256) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3258 = tensor.empty() : tensor<1x64xf32> + %3259 = "ttir.relu"(%3257, %3258) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3260 = tensor.empty() : tensor<1x64xf32> + %3261 = "ttir.relu"(%3259, %3260) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3262 = tensor.empty() : tensor<1x64xf32> + %3263 = "ttir.relu"(%3261, %3262) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3264 = tensor.empty() : tensor<1x64xf32> + %3265 = "ttir.relu"(%3263, %3264) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3266 = tensor.empty() : tensor<1x64xf32> + %3267 = "ttir.relu"(%3265, %3266) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3268 = tensor.empty() : tensor<1x64xf32> + %3269 = "ttir.relu"(%3267, %3268) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3270 = tensor.empty() : tensor<1x64xf32> + %3271 = "ttir.relu"(%3269, %3270) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3272 = tensor.empty() : tensor<1x64xf32> + %3273 = "ttir.relu"(%3271, %3272) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3274 = tensor.empty() : tensor<1x64xf32> + %3275 = "ttir.relu"(%3273, %3274) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3276 = tensor.empty() : tensor<1x64xf32> + %3277 = "ttir.relu"(%3275, %3276) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3278 = tensor.empty() : tensor<1x64xf32> + %3279 = "ttir.relu"(%3277, %3278) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3280 = tensor.empty() : tensor<1x64xf32> + %3281 = "ttir.relu"(%3279, %3280) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3282 = tensor.empty() : tensor<1x64xf32> + %3283 = "ttir.relu"(%3281, %3282) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3284 = tensor.empty() : tensor<1x64xf32> + %3285 = "ttir.relu"(%3283, %3284) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3286 = tensor.empty() : tensor<1x64xf32> + %3287 = "ttir.relu"(%3285, %3286) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3288 = tensor.empty() : tensor<1x64xf32> + %3289 = "ttir.relu"(%3287, %3288) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3290 = tensor.empty() : tensor<1x64xf32> + %3291 = "ttir.relu"(%3289, %3290) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3292 = tensor.empty() : tensor<1x64xf32> + %3293 = "ttir.relu"(%3291, %3292) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3294 = tensor.empty() : tensor<1x64xf32> + %3295 = "ttir.relu"(%3293, %3294) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3296 = tensor.empty() : tensor<1x64xf32> + %3297 = "ttir.relu"(%3295, %3296) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3298 = tensor.empty() : tensor<1x64xf32> + %3299 = "ttir.relu"(%3297, %3298) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3300 = tensor.empty() : tensor<1x64xf32> + %3301 = "ttir.relu"(%3299, %3300) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3302 = tensor.empty() : tensor<1x64xf32> + %3303 = "ttir.relu"(%3301, %3302) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3304 = tensor.empty() : tensor<1x64xf32> + %3305 = "ttir.relu"(%3303, %3304) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3306 = tensor.empty() : tensor<1x64xf32> + %3307 = "ttir.relu"(%3305, %3306) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3308 = tensor.empty() : tensor<1x64xf32> + %3309 = "ttir.relu"(%3307, %3308) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3310 = tensor.empty() : tensor<1x64xf32> + %3311 = "ttir.relu"(%3309, %3310) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3312 = tensor.empty() : tensor<1x64xf32> + %3313 = "ttir.relu"(%3311, %3312) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3314 = tensor.empty() : tensor<1x64xf32> + %3315 = "ttir.relu"(%3313, %3314) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3316 = tensor.empty() : tensor<1x64xf32> + %3317 = "ttir.relu"(%3315, %3316) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3318 = tensor.empty() : tensor<1x64xf32> + %3319 = "ttir.relu"(%3317, %3318) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3320 = tensor.empty() : tensor<1x64xf32> + %3321 = "ttir.relu"(%3319, %3320) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3322 = tensor.empty() : tensor<1x64xf32> + %3323 = "ttir.relu"(%3321, %3322) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3324 = tensor.empty() : tensor<1x64xf32> + %3325 = "ttir.relu"(%3323, %3324) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3326 = tensor.empty() : tensor<1x64xf32> + %3327 = "ttir.relu"(%3325, %3326) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3328 = tensor.empty() : tensor<1x64xf32> + %3329 = "ttir.relu"(%3327, %3328) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3330 = tensor.empty() : tensor<1x64xf32> + %3331 = "ttir.relu"(%3329, %3330) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3332 = tensor.empty() : tensor<1x64xf32> + %3333 = "ttir.relu"(%3331, %3332) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3334 = tensor.empty() : tensor<1x64xf32> + %3335 = "ttir.relu"(%3333, %3334) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3336 = tensor.empty() : tensor<1x64xf32> + %3337 = "ttir.relu"(%3335, %3336) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3338 = tensor.empty() : tensor<1x64xf32> + %3339 = "ttir.relu"(%3337, %3338) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3340 = tensor.empty() : tensor<1x64xf32> + %3341 = "ttir.relu"(%3339, %3340) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3342 = tensor.empty() : tensor<1x64xf32> + %3343 = "ttir.relu"(%3341, %3342) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3344 = tensor.empty() : tensor<1x64xf32> + %3345 = "ttir.relu"(%3343, %3344) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3346 = tensor.empty() : tensor<1x64xf32> + %3347 = "ttir.relu"(%3345, %3346) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3348 = tensor.empty() : tensor<1x64xf32> + %3349 = "ttir.relu"(%3347, %3348) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3350 = tensor.empty() : tensor<1x64xf32> + %3351 = "ttir.relu"(%3349, %3350) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3352 = tensor.empty() : tensor<1x64xf32> + %3353 = "ttir.relu"(%3351, %3352) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3354 = tensor.empty() : tensor<1x64xf32> + %3355 = "ttir.relu"(%3353, %3354) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3356 = tensor.empty() : tensor<1x64xf32> + %3357 = "ttir.relu"(%3355, %3356) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3358 = tensor.empty() : tensor<1x64xf32> + %3359 = "ttir.relu"(%3357, %3358) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3360 = tensor.empty() : tensor<1x64xf32> + %3361 = "ttir.relu"(%3359, %3360) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3362 = tensor.empty() : tensor<1x64xf32> + %3363 = "ttir.relu"(%3361, %3362) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3364 = tensor.empty() : tensor<1x64xf32> + %3365 = "ttir.relu"(%3363, %3364) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3366 = tensor.empty() : tensor<1x64xf32> + %3367 = "ttir.relu"(%3365, %3366) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3368 = tensor.empty() : tensor<1x64xf32> + %3369 = "ttir.relu"(%3367, %3368) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3370 = tensor.empty() : tensor<1x64xf32> + %3371 = "ttir.relu"(%3369, %3370) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3372 = tensor.empty() : tensor<1x64xf32> + %3373 = "ttir.relu"(%3371, %3372) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3374 = tensor.empty() : tensor<1x64xf32> + %3375 = "ttir.relu"(%3373, %3374) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3376 = tensor.empty() : tensor<1x64xf32> + %3377 = "ttir.relu"(%3375, %3376) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3378 = tensor.empty() : tensor<1x64xf32> + %3379 = "ttir.relu"(%3377, %3378) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3380 = tensor.empty() : tensor<1x64xf32> + %3381 = "ttir.relu"(%3379, %3380) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3382 = tensor.empty() : tensor<1x64xf32> + %3383 = "ttir.relu"(%3381, %3382) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3384 = tensor.empty() : tensor<1x64xf32> + %3385 = "ttir.relu"(%3383, %3384) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3386 = tensor.empty() : tensor<1x64xf32> + %3387 = "ttir.relu"(%3385, %3386) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3388 = tensor.empty() : tensor<1x64xf32> + %3389 = "ttir.relu"(%3387, %3388) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3390 = tensor.empty() : tensor<1x64xf32> + %3391 = "ttir.relu"(%3389, %3390) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3392 = tensor.empty() : tensor<1x64xf32> + %3393 = "ttir.relu"(%3391, %3392) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3394 = tensor.empty() : tensor<1x64xf32> + %3395 = "ttir.relu"(%3393, %3394) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3396 = tensor.empty() : tensor<1x64xf32> + %3397 = "ttir.relu"(%3395, %3396) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3398 = tensor.empty() : tensor<1x64xf32> + %3399 = "ttir.relu"(%3397, %3398) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3400 = tensor.empty() : tensor<1x64xf32> + %3401 = "ttir.relu"(%3399, %3400) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3402 = tensor.empty() : tensor<1x64xf32> + %3403 = "ttir.relu"(%3401, %3402) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3404 = tensor.empty() : tensor<1x64xf32> + %3405 = "ttir.relu"(%3403, %3404) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3406 = tensor.empty() : tensor<1x64xf32> + %3407 = "ttir.relu"(%3405, %3406) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3408 = tensor.empty() : tensor<1x64xf32> + %3409 = "ttir.relu"(%3407, %3408) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3410 = tensor.empty() : tensor<1x64xf32> + %3411 = "ttir.relu"(%3409, %3410) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3412 = tensor.empty() : tensor<1x64xf32> + %3413 = "ttir.relu"(%3411, %3412) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3414 = tensor.empty() : tensor<1x64xf32> + %3415 = "ttir.relu"(%3413, %3414) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3416 = tensor.empty() : tensor<1x64xf32> + %3417 = "ttir.relu"(%3415, %3416) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3418 = tensor.empty() : tensor<1x64xf32> + %3419 = "ttir.relu"(%3417, %3418) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3420 = tensor.empty() : tensor<1x64xf32> + %3421 = "ttir.relu"(%3419, %3420) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3422 = tensor.empty() : tensor<1x64xf32> + %3423 = "ttir.relu"(%3421, %3422) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3424 = tensor.empty() : tensor<1x64xf32> + %3425 = "ttir.relu"(%3423, %3424) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3426 = tensor.empty() : tensor<1x64xf32> + %3427 = "ttir.relu"(%3425, %3426) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3428 = tensor.empty() : tensor<1x64xf32> + %3429 = "ttir.relu"(%3427, %3428) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3430 = tensor.empty() : tensor<1x64xf32> + %3431 = "ttir.relu"(%3429, %3430) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3432 = tensor.empty() : tensor<1x64xf32> + %3433 = "ttir.relu"(%3431, %3432) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3434 = tensor.empty() : tensor<1x64xf32> + %3435 = "ttir.relu"(%3433, %3434) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3436 = tensor.empty() : tensor<1x64xf32> + %3437 = "ttir.relu"(%3435, %3436) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3438 = tensor.empty() : tensor<1x64xf32> + %3439 = "ttir.relu"(%3437, %3438) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3440 = tensor.empty() : tensor<1x64xf32> + %3441 = "ttir.relu"(%3439, %3440) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3442 = tensor.empty() : tensor<1x64xf32> + %3443 = "ttir.relu"(%3441, %3442) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3444 = tensor.empty() : tensor<1x64xf32> + %3445 = "ttir.relu"(%3443, %3444) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3446 = tensor.empty() : tensor<1x64xf32> + %3447 = "ttir.relu"(%3445, %3446) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3448 = tensor.empty() : tensor<1x64xf32> + %3449 = "ttir.relu"(%3447, %3448) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3450 = tensor.empty() : tensor<1x64xf32> + %3451 = "ttir.relu"(%3449, %3450) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3452 = tensor.empty() : tensor<1x64xf32> + %3453 = "ttir.relu"(%3451, %3452) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3454 = tensor.empty() : tensor<1x64xf32> + %3455 = "ttir.relu"(%3453, %3454) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3456 = tensor.empty() : tensor<1x64xf32> + %3457 = "ttir.relu"(%3455, %3456) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3458 = tensor.empty() : tensor<1x64xf32> + %3459 = "ttir.relu"(%3457, %3458) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3460 = tensor.empty() : tensor<1x64xf32> + %3461 = "ttir.relu"(%3459, %3460) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3462 = tensor.empty() : tensor<1x64xf32> + %3463 = "ttir.relu"(%3461, %3462) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3464 = tensor.empty() : tensor<1x64xf32> + %3465 = "ttir.relu"(%3463, %3464) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3466 = tensor.empty() : tensor<1x64xf32> + %3467 = "ttir.relu"(%3465, %3466) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3468 = tensor.empty() : tensor<1x64xf32> + %3469 = "ttir.relu"(%3467, %3468) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3470 = tensor.empty() : tensor<1x64xf32> + %3471 = "ttir.relu"(%3469, %3470) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3472 = tensor.empty() : tensor<1x64xf32> + %3473 = "ttir.relu"(%3471, %3472) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3474 = tensor.empty() : tensor<1x64xf32> + %3475 = "ttir.relu"(%3473, %3474) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3476 = tensor.empty() : tensor<1x64xf32> + %3477 = "ttir.relu"(%3475, %3476) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3478 = tensor.empty() : tensor<1x64xf32> + %3479 = "ttir.relu"(%3477, %3478) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3480 = tensor.empty() : tensor<1x64xf32> + %3481 = "ttir.relu"(%3479, %3480) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3482 = tensor.empty() : tensor<1x64xf32> + %3483 = "ttir.relu"(%3481, %3482) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3484 = tensor.empty() : tensor<1x64xf32> + %3485 = "ttir.relu"(%3483, %3484) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3486 = tensor.empty() : tensor<1x64xf32> + %3487 = "ttir.relu"(%3485, %3486) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3488 = tensor.empty() : tensor<1x64xf32> + %3489 = "ttir.relu"(%3487, %3488) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3490 = tensor.empty() : tensor<1x64xf32> + %3491 = "ttir.relu"(%3489, %3490) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3492 = tensor.empty() : tensor<1x64xf32> + %3493 = "ttir.relu"(%3491, %3492) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3494 = tensor.empty() : tensor<1x64xf32> + %3495 = "ttir.relu"(%3493, %3494) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3496 = tensor.empty() : tensor<1x64xf32> + %3497 = "ttir.relu"(%3495, %3496) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3498 = tensor.empty() : tensor<1x64xf32> + %3499 = "ttir.relu"(%3497, %3498) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3500 = tensor.empty() : tensor<1x64xf32> + %3501 = "ttir.relu"(%3499, %3500) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3502 = tensor.empty() : tensor<1x64xf32> + %3503 = "ttir.relu"(%3501, %3502) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3504 = tensor.empty() : tensor<1x64xf32> + %3505 = "ttir.relu"(%3503, %3504) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3506 = tensor.empty() : tensor<1x64xf32> + %3507 = "ttir.relu"(%3505, %3506) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3508 = tensor.empty() : tensor<1x64xf32> + %3509 = "ttir.relu"(%3507, %3508) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3510 = tensor.empty() : tensor<1x64xf32> + %3511 = "ttir.relu"(%3509, %3510) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3512 = tensor.empty() : tensor<1x64xf32> + %3513 = "ttir.relu"(%3511, %3512) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3514 = tensor.empty() : tensor<1x64xf32> + %3515 = "ttir.relu"(%3513, %3514) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3516 = tensor.empty() : tensor<1x64xf32> + %3517 = "ttir.relu"(%3515, %3516) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3518 = tensor.empty() : tensor<1x64xf32> + %3519 = "ttir.relu"(%3517, %3518) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3520 = tensor.empty() : tensor<1x64xf32> + %3521 = "ttir.relu"(%3519, %3520) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3522 = tensor.empty() : tensor<1x64xf32> + %3523 = "ttir.relu"(%3521, %3522) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3524 = tensor.empty() : tensor<1x64xf32> + %3525 = "ttir.relu"(%3523, %3524) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3526 = tensor.empty() : tensor<1x64xf32> + %3527 = "ttir.relu"(%3525, %3526) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3528 = tensor.empty() : tensor<1x64xf32> + %3529 = "ttir.relu"(%3527, %3528) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3530 = tensor.empty() : tensor<1x64xf32> + %3531 = "ttir.relu"(%3529, %3530) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3532 = tensor.empty() : tensor<1x64xf32> + %3533 = "ttir.relu"(%3531, %3532) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3534 = tensor.empty() : tensor<1x64xf32> + %3535 = "ttir.relu"(%3533, %3534) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3536 = tensor.empty() : tensor<1x64xf32> + %3537 = "ttir.relu"(%3535, %3536) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3538 = tensor.empty() : tensor<1x64xf32> + %3539 = "ttir.relu"(%3537, %3538) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3540 = tensor.empty() : tensor<1x64xf32> + %3541 = "ttir.relu"(%3539, %3540) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3542 = tensor.empty() : tensor<1x64xf32> + %3543 = "ttir.relu"(%3541, %3542) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3544 = tensor.empty() : tensor<1x64xf32> + %3545 = "ttir.relu"(%3543, %3544) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3546 = tensor.empty() : tensor<1x64xf32> + %3547 = "ttir.relu"(%3545, %3546) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3548 = tensor.empty() : tensor<1x64xf32> + %3549 = "ttir.relu"(%3547, %3548) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3550 = tensor.empty() : tensor<1x64xf32> + %3551 = "ttir.relu"(%3549, %3550) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3552 = tensor.empty() : tensor<1x64xf32> + %3553 = "ttir.relu"(%3551, %3552) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3554 = tensor.empty() : tensor<1x64xf32> + %3555 = "ttir.relu"(%3553, %3554) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3556 = tensor.empty() : tensor<1x64xf32> + %3557 = "ttir.relu"(%3555, %3556) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3558 = tensor.empty() : tensor<1x64xf32> + %3559 = "ttir.relu"(%3557, %3558) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3560 = tensor.empty() : tensor<1x64xf32> + %3561 = "ttir.relu"(%3559, %3560) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3562 = tensor.empty() : tensor<1x64xf32> + %3563 = "ttir.relu"(%3561, %3562) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3564 = tensor.empty() : tensor<1x64xf32> + %3565 = "ttir.relu"(%3563, %3564) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3566 = tensor.empty() : tensor<1x64xf32> + %3567 = "ttir.relu"(%3565, %3566) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3568 = tensor.empty() : tensor<1x64xf32> + %3569 = "ttir.relu"(%3567, %3568) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3570 = tensor.empty() : tensor<1x64xf32> + %3571 = "ttir.relu"(%3569, %3570) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3572 = tensor.empty() : tensor<1x64xf32> + %3573 = "ttir.relu"(%3571, %3572) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3574 = tensor.empty() : tensor<1x64xf32> + %3575 = "ttir.relu"(%3573, %3574) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3576 = tensor.empty() : tensor<1x64xf32> + %3577 = "ttir.relu"(%3575, %3576) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3578 = tensor.empty() : tensor<1x64xf32> + %3579 = "ttir.relu"(%3577, %3578) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3580 = tensor.empty() : tensor<1x64xf32> + %3581 = "ttir.relu"(%3579, %3580) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3582 = tensor.empty() : tensor<1x64xf32> + %3583 = "ttir.relu"(%3581, %3582) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3584 = tensor.empty() : tensor<1x64xf32> + %3585 = "ttir.relu"(%3583, %3584) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3586 = tensor.empty() : tensor<1x64xf32> + %3587 = "ttir.relu"(%3585, %3586) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3588 = tensor.empty() : tensor<1x64xf32> + %3589 = "ttir.relu"(%3587, %3588) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3590 = tensor.empty() : tensor<1x64xf32> + %3591 = "ttir.relu"(%3589, %3590) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3592 = tensor.empty() : tensor<1x64xf32> + %3593 = "ttir.relu"(%3591, %3592) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3594 = tensor.empty() : tensor<1x64xf32> + %3595 = "ttir.relu"(%3593, %3594) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3596 = tensor.empty() : tensor<1x64xf32> + %3597 = "ttir.relu"(%3595, %3596) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3598 = tensor.empty() : tensor<1x64xf32> + %3599 = "ttir.relu"(%3597, %3598) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3600 = tensor.empty() : tensor<1x64xf32> + %3601 = "ttir.relu"(%3599, %3600) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3602 = tensor.empty() : tensor<1x64xf32> + %3603 = "ttir.relu"(%3601, %3602) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3604 = tensor.empty() : tensor<1x64xf32> + %3605 = "ttir.relu"(%3603, %3604) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3606 = tensor.empty() : tensor<1x64xf32> + %3607 = "ttir.relu"(%3605, %3606) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3608 = tensor.empty() : tensor<1x64xf32> + %3609 = "ttir.relu"(%3607, %3608) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3610 = tensor.empty() : tensor<1x64xf32> + %3611 = "ttir.relu"(%3609, %3610) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3612 = tensor.empty() : tensor<1x64xf32> + %3613 = "ttir.relu"(%3611, %3612) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3614 = tensor.empty() : tensor<1x64xf32> + %3615 = "ttir.relu"(%3613, %3614) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3616 = tensor.empty() : tensor<1x64xf32> + %3617 = "ttir.relu"(%3615, %3616) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3618 = tensor.empty() : tensor<1x64xf32> + %3619 = "ttir.relu"(%3617, %3618) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3620 = tensor.empty() : tensor<1x64xf32> + %3621 = "ttir.relu"(%3619, %3620) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3622 = tensor.empty() : tensor<1x64xf32> + %3623 = "ttir.relu"(%3621, %3622) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3624 = tensor.empty() : tensor<1x64xf32> + %3625 = "ttir.relu"(%3623, %3624) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3626 = tensor.empty() : tensor<1x64xf32> + %3627 = "ttir.relu"(%3625, %3626) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3628 = tensor.empty() : tensor<1x64xf32> + %3629 = "ttir.relu"(%3627, %3628) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3630 = tensor.empty() : tensor<1x64xf32> + %3631 = "ttir.relu"(%3629, %3630) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3632 = tensor.empty() : tensor<1x64xf32> + %3633 = "ttir.relu"(%3631, %3632) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3634 = tensor.empty() : tensor<1x64xf32> + %3635 = "ttir.relu"(%3633, %3634) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3636 = tensor.empty() : tensor<1x64xf32> + %3637 = "ttir.relu"(%3635, %3636) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3638 = tensor.empty() : tensor<1x64xf32> + %3639 = "ttir.relu"(%3637, %3638) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3640 = tensor.empty() : tensor<1x64xf32> + %3641 = "ttir.relu"(%3639, %3640) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3642 = tensor.empty() : tensor<1x64xf32> + %3643 = "ttir.relu"(%3641, %3642) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3644 = tensor.empty() : tensor<1x64xf32> + %3645 = "ttir.relu"(%3643, %3644) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3646 = tensor.empty() : tensor<1x64xf32> + %3647 = "ttir.relu"(%3645, %3646) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3648 = tensor.empty() : tensor<1x64xf32> + %3649 = "ttir.relu"(%3647, %3648) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3650 = tensor.empty() : tensor<1x64xf32> + %3651 = "ttir.relu"(%3649, %3650) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3652 = tensor.empty() : tensor<1x64xf32> + %3653 = "ttir.relu"(%3651, %3652) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3654 = tensor.empty() : tensor<1x64xf32> + %3655 = "ttir.relu"(%3653, %3654) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3656 = tensor.empty() : tensor<1x64xf32> + %3657 = "ttir.relu"(%3655, %3656) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3658 = tensor.empty() : tensor<1x64xf32> + %3659 = "ttir.relu"(%3657, %3658) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3660 = tensor.empty() : tensor<1x64xf32> + %3661 = "ttir.relu"(%3659, %3660) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3662 = tensor.empty() : tensor<1x64xf32> + %3663 = "ttir.relu"(%3661, %3662) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3664 = tensor.empty() : tensor<1x64xf32> + %3665 = "ttir.relu"(%3663, %3664) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3666 = tensor.empty() : tensor<1x64xf32> + %3667 = "ttir.relu"(%3665, %3666) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3668 = tensor.empty() : tensor<1x64xf32> + %3669 = "ttir.relu"(%3667, %3668) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3670 = tensor.empty() : tensor<1x64xf32> + %3671 = "ttir.relu"(%3669, %3670) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3672 = tensor.empty() : tensor<1x64xf32> + %3673 = "ttir.relu"(%3671, %3672) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3674 = tensor.empty() : tensor<1x64xf32> + %3675 = "ttir.relu"(%3673, %3674) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3676 = tensor.empty() : tensor<1x64xf32> + %3677 = "ttir.relu"(%3675, %3676) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3678 = tensor.empty() : tensor<1x64xf32> + %3679 = "ttir.relu"(%3677, %3678) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3680 = tensor.empty() : tensor<1x64xf32> + %3681 = "ttir.relu"(%3679, %3680) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3682 = tensor.empty() : tensor<1x64xf32> + %3683 = "ttir.relu"(%3681, %3682) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3684 = tensor.empty() : tensor<1x64xf32> + %3685 = "ttir.relu"(%3683, %3684) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3686 = tensor.empty() : tensor<1x64xf32> + %3687 = "ttir.relu"(%3685, %3686) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3688 = tensor.empty() : tensor<1x64xf32> + %3689 = "ttir.relu"(%3687, %3688) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3690 = tensor.empty() : tensor<1x64xf32> + %3691 = "ttir.relu"(%3689, %3690) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3692 = tensor.empty() : tensor<1x64xf32> + %3693 = "ttir.relu"(%3691, %3692) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3694 = tensor.empty() : tensor<1x64xf32> + %3695 = "ttir.relu"(%3693, %3694) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3696 = tensor.empty() : tensor<1x64xf32> + %3697 = "ttir.relu"(%3695, %3696) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3698 = tensor.empty() : tensor<1x64xf32> + %3699 = "ttir.relu"(%3697, %3698) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3700 = tensor.empty() : tensor<1x64xf32> + %3701 = "ttir.relu"(%3699, %3700) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3702 = tensor.empty() : tensor<1x64xf32> + %3703 = "ttir.relu"(%3701, %3702) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3704 = tensor.empty() : tensor<1x64xf32> + %3705 = "ttir.relu"(%3703, %3704) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3706 = tensor.empty() : tensor<1x64xf32> + %3707 = "ttir.relu"(%3705, %3706) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3708 = tensor.empty() : tensor<1x64xf32> + %3709 = "ttir.relu"(%3707, %3708) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3710 = tensor.empty() : tensor<1x64xf32> + %3711 = "ttir.relu"(%3709, %3710) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3712 = tensor.empty() : tensor<1x64xf32> + %3713 = "ttir.relu"(%3711, %3712) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3714 = tensor.empty() : tensor<1x64xf32> + %3715 = "ttir.relu"(%3713, %3714) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3716 = tensor.empty() : tensor<1x64xf32> + %3717 = "ttir.relu"(%3715, %3716) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3718 = tensor.empty() : tensor<1x64xf32> + %3719 = "ttir.relu"(%3717, %3718) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3720 = tensor.empty() : tensor<1x64xf32> + %3721 = "ttir.relu"(%3719, %3720) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3722 = tensor.empty() : tensor<1x64xf32> + %3723 = "ttir.relu"(%3721, %3722) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3724 = tensor.empty() : tensor<1x64xf32> + %3725 = "ttir.relu"(%3723, %3724) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3726 = tensor.empty() : tensor<1x64xf32> + %3727 = "ttir.relu"(%3725, %3726) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3728 = tensor.empty() : tensor<1x64xf32> + %3729 = "ttir.relu"(%3727, %3728) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3730 = tensor.empty() : tensor<1x64xf32> + %3731 = "ttir.relu"(%3729, %3730) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3732 = tensor.empty() : tensor<1x64xf32> + %3733 = "ttir.relu"(%3731, %3732) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3734 = tensor.empty() : tensor<1x64xf32> + %3735 = "ttir.relu"(%3733, %3734) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3736 = tensor.empty() : tensor<1x64xf32> + %3737 = "ttir.relu"(%3735, %3736) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3738 = tensor.empty() : tensor<1x64xf32> + %3739 = "ttir.relu"(%3737, %3738) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3740 = tensor.empty() : tensor<1x64xf32> + %3741 = "ttir.relu"(%3739, %3740) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3742 = tensor.empty() : tensor<1x64xf32> + %3743 = "ttir.relu"(%3741, %3742) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3744 = tensor.empty() : tensor<1x64xf32> + %3745 = "ttir.relu"(%3743, %3744) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3746 = tensor.empty() : tensor<1x64xf32> + %3747 = "ttir.relu"(%3745, %3746) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3748 = tensor.empty() : tensor<1x64xf32> + %3749 = "ttir.relu"(%3747, %3748) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3750 = tensor.empty() : tensor<1x64xf32> + %3751 = "ttir.relu"(%3749, %3750) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3752 = tensor.empty() : tensor<1x64xf32> + %3753 = "ttir.relu"(%3751, %3752) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3754 = tensor.empty() : tensor<1x64xf32> + %3755 = "ttir.relu"(%3753, %3754) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3756 = tensor.empty() : tensor<1x64xf32> + %3757 = "ttir.relu"(%3755, %3756) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3758 = tensor.empty() : tensor<1x64xf32> + %3759 = "ttir.relu"(%3757, %3758) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3760 = tensor.empty() : tensor<1x64xf32> + %3761 = "ttir.relu"(%3759, %3760) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3762 = tensor.empty() : tensor<1x64xf32> + %3763 = "ttir.relu"(%3761, %3762) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3764 = tensor.empty() : tensor<1x64xf32> + %3765 = "ttir.relu"(%3763, %3764) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3766 = tensor.empty() : tensor<1x64xf32> + %3767 = "ttir.relu"(%3765, %3766) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3768 = tensor.empty() : tensor<1x64xf32> + %3769 = "ttir.relu"(%3767, %3768) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3770 = tensor.empty() : tensor<1x64xf32> + %3771 = "ttir.relu"(%3769, %3770) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3772 = tensor.empty() : tensor<1x64xf32> + %3773 = "ttir.relu"(%3771, %3772) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3774 = tensor.empty() : tensor<1x64xf32> + %3775 = "ttir.relu"(%3773, %3774) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3776 = tensor.empty() : tensor<1x64xf32> + %3777 = "ttir.relu"(%3775, %3776) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3778 = tensor.empty() : tensor<1x64xf32> + %3779 = "ttir.relu"(%3777, %3778) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3780 = tensor.empty() : tensor<1x64xf32> + %3781 = "ttir.relu"(%3779, %3780) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3782 = tensor.empty() : tensor<1x64xf32> + %3783 = "ttir.relu"(%3781, %3782) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3784 = tensor.empty() : tensor<1x64xf32> + %3785 = "ttir.relu"(%3783, %3784) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3786 = tensor.empty() : tensor<1x64xf32> + %3787 = "ttir.relu"(%3785, %3786) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3788 = tensor.empty() : tensor<1x64xf32> + %3789 = "ttir.relu"(%3787, %3788) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3790 = tensor.empty() : tensor<1x64xf32> + %3791 = "ttir.relu"(%3789, %3790) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3792 = tensor.empty() : tensor<1x64xf32> + %3793 = "ttir.relu"(%3791, %3792) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3794 = tensor.empty() : tensor<1x64xf32> + %3795 = "ttir.relu"(%3793, %3794) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3796 = tensor.empty() : tensor<1x64xf32> + %3797 = "ttir.relu"(%3795, %3796) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3798 = tensor.empty() : tensor<1x64xf32> + %3799 = "ttir.relu"(%3797, %3798) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3800 = tensor.empty() : tensor<1x64xf32> + %3801 = "ttir.relu"(%3799, %3800) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3802 = tensor.empty() : tensor<1x64xf32> + %3803 = "ttir.relu"(%3801, %3802) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3804 = tensor.empty() : tensor<1x64xf32> + %3805 = "ttir.relu"(%3803, %3804) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3806 = tensor.empty() : tensor<1x64xf32> + %3807 = "ttir.relu"(%3805, %3806) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3808 = tensor.empty() : tensor<1x64xf32> + %3809 = "ttir.relu"(%3807, %3808) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3810 = tensor.empty() : tensor<1x64xf32> + %3811 = "ttir.relu"(%3809, %3810) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3812 = tensor.empty() : tensor<1x64xf32> + %3813 = "ttir.relu"(%3811, %3812) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3814 = tensor.empty() : tensor<1x64xf32> + %3815 = "ttir.relu"(%3813, %3814) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3816 = tensor.empty() : tensor<1x64xf32> + %3817 = "ttir.relu"(%3815, %3816) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3818 = tensor.empty() : tensor<1x64xf32> + %3819 = "ttir.relu"(%3817, %3818) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3820 = tensor.empty() : tensor<1x64xf32> + %3821 = "ttir.relu"(%3819, %3820) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3822 = tensor.empty() : tensor<1x64xf32> + %3823 = "ttir.relu"(%3821, %3822) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3824 = tensor.empty() : tensor<1x64xf32> + %3825 = "ttir.relu"(%3823, %3824) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3826 = tensor.empty() : tensor<1x64xf32> + %3827 = "ttir.relu"(%3825, %3826) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3828 = tensor.empty() : tensor<1x64xf32> + %3829 = "ttir.relu"(%3827, %3828) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3830 = tensor.empty() : tensor<1x64xf32> + %3831 = "ttir.relu"(%3829, %3830) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3832 = tensor.empty() : tensor<1x64xf32> + %3833 = "ttir.relu"(%3831, %3832) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3834 = tensor.empty() : tensor<1x64xf32> + %3835 = "ttir.relu"(%3833, %3834) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3836 = tensor.empty() : tensor<1x64xf32> + %3837 = "ttir.relu"(%3835, %3836) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3838 = tensor.empty() : tensor<1x64xf32> + %3839 = "ttir.relu"(%3837, %3838) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3840 = tensor.empty() : tensor<1x64xf32> + %3841 = "ttir.relu"(%3839, %3840) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3842 = tensor.empty() : tensor<1x64xf32> + %3843 = "ttir.relu"(%3841, %3842) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3844 = tensor.empty() : tensor<1x64xf32> + %3845 = "ttir.relu"(%3843, %3844) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3846 = tensor.empty() : tensor<1x64xf32> + %3847 = "ttir.relu"(%3845, %3846) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3848 = tensor.empty() : tensor<1x64xf32> + %3849 = "ttir.relu"(%3847, %3848) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3850 = tensor.empty() : tensor<1x64xf32> + %3851 = "ttir.relu"(%3849, %3850) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3852 = tensor.empty() : tensor<1x64xf32> + %3853 = "ttir.relu"(%3851, %3852) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3854 = tensor.empty() : tensor<1x64xf32> + %3855 = "ttir.relu"(%3853, %3854) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3856 = tensor.empty() : tensor<1x64xf32> + %3857 = "ttir.relu"(%3855, %3856) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3858 = tensor.empty() : tensor<1x64xf32> + %3859 = "ttir.relu"(%3857, %3858) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3860 = tensor.empty() : tensor<1x64xf32> + %3861 = "ttir.relu"(%3859, %3860) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3862 = tensor.empty() : tensor<1x64xf32> + %3863 = "ttir.relu"(%3861, %3862) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3864 = tensor.empty() : tensor<1x64xf32> + %3865 = "ttir.relu"(%3863, %3864) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3866 = tensor.empty() : tensor<1x64xf32> + %3867 = "ttir.relu"(%3865, %3866) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3868 = tensor.empty() : tensor<1x64xf32> + %3869 = "ttir.relu"(%3867, %3868) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3870 = tensor.empty() : tensor<1x64xf32> + %3871 = "ttir.relu"(%3869, %3870) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3872 = tensor.empty() : tensor<1x64xf32> + %3873 = "ttir.relu"(%3871, %3872) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3874 = tensor.empty() : tensor<1x64xf32> + %3875 = "ttir.relu"(%3873, %3874) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3876 = tensor.empty() : tensor<1x64xf32> + %3877 = "ttir.relu"(%3875, %3876) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3878 = tensor.empty() : tensor<1x64xf32> + %3879 = "ttir.relu"(%3877, %3878) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3880 = tensor.empty() : tensor<1x64xf32> + %3881 = "ttir.relu"(%3879, %3880) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3882 = tensor.empty() : tensor<1x64xf32> + %3883 = "ttir.relu"(%3881, %3882) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3884 = tensor.empty() : tensor<1x64xf32> + %3885 = "ttir.relu"(%3883, %3884) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3886 = tensor.empty() : tensor<1x64xf32> + %3887 = "ttir.relu"(%3885, %3886) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3888 = tensor.empty() : tensor<1x64xf32> + %3889 = "ttir.relu"(%3887, %3888) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3890 = tensor.empty() : tensor<1x64xf32> + %3891 = "ttir.relu"(%3889, %3890) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3892 = tensor.empty() : tensor<1x64xf32> + %3893 = "ttir.relu"(%3891, %3892) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3894 = tensor.empty() : tensor<1x64xf32> + %3895 = "ttir.relu"(%3893, %3894) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3896 = tensor.empty() : tensor<1x64xf32> + %3897 = "ttir.relu"(%3895, %3896) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3898 = tensor.empty() : tensor<1x64xf32> + %3899 = "ttir.relu"(%3897, %3898) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3900 = tensor.empty() : tensor<1x64xf32> + %3901 = "ttir.relu"(%3899, %3900) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3902 = tensor.empty() : tensor<1x64xf32> + %3903 = "ttir.relu"(%3901, %3902) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3904 = tensor.empty() : tensor<1x64xf32> + %3905 = "ttir.relu"(%3903, %3904) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3906 = tensor.empty() : tensor<1x64xf32> + %3907 = "ttir.relu"(%3905, %3906) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3908 = tensor.empty() : tensor<1x64xf32> + %3909 = "ttir.relu"(%3907, %3908) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3910 = tensor.empty() : tensor<1x64xf32> + %3911 = "ttir.relu"(%3909, %3910) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3912 = tensor.empty() : tensor<1x64xf32> + %3913 = "ttir.relu"(%3911, %3912) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3914 = tensor.empty() : tensor<1x64xf32> + %3915 = "ttir.relu"(%3913, %3914) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3916 = tensor.empty() : tensor<1x64xf32> + %3917 = "ttir.relu"(%3915, %3916) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3918 = tensor.empty() : tensor<1x64xf32> + %3919 = "ttir.relu"(%3917, %3918) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3920 = tensor.empty() : tensor<1x64xf32> + %3921 = "ttir.relu"(%3919, %3920) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3922 = tensor.empty() : tensor<1x64xf32> + %3923 = "ttir.relu"(%3921, %3922) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3924 = tensor.empty() : tensor<1x64xf32> + %3925 = "ttir.relu"(%3923, %3924) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3926 = tensor.empty() : tensor<1x64xf32> + %3927 = "ttir.relu"(%3925, %3926) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3928 = tensor.empty() : tensor<1x64xf32> + %3929 = "ttir.relu"(%3927, %3928) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3930 = tensor.empty() : tensor<1x64xf32> + %3931 = "ttir.relu"(%3929, %3930) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3932 = tensor.empty() : tensor<1x64xf32> + %3933 = "ttir.relu"(%3931, %3932) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3934 = tensor.empty() : tensor<1x64xf32> + %3935 = "ttir.relu"(%3933, %3934) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3936 = tensor.empty() : tensor<1x64xf32> + %3937 = "ttir.relu"(%3935, %3936) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3938 = tensor.empty() : tensor<1x64xf32> + %3939 = "ttir.relu"(%3937, %3938) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3940 = tensor.empty() : tensor<1x64xf32> + %3941 = "ttir.relu"(%3939, %3940) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3942 = tensor.empty() : tensor<1x64xf32> + %3943 = "ttir.relu"(%3941, %3942) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3944 = tensor.empty() : tensor<1x64xf32> + %3945 = "ttir.relu"(%3943, %3944) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3946 = tensor.empty() : tensor<1x64xf32> + %3947 = "ttir.relu"(%3945, %3946) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3948 = tensor.empty() : tensor<1x64xf32> + %3949 = "ttir.relu"(%3947, %3948) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3950 = tensor.empty() : tensor<1x64xf32> + %3951 = "ttir.relu"(%3949, %3950) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3952 = tensor.empty() : tensor<1x64xf32> + %3953 = "ttir.relu"(%3951, %3952) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3954 = tensor.empty() : tensor<1x64xf32> + %3955 = "ttir.relu"(%3953, %3954) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3956 = tensor.empty() : tensor<1x64xf32> + %3957 = "ttir.relu"(%3955, %3956) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3958 = tensor.empty() : tensor<1x64xf32> + %3959 = "ttir.relu"(%3957, %3958) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3960 = tensor.empty() : tensor<1x64xf32> + %3961 = "ttir.relu"(%3959, %3960) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3962 = tensor.empty() : tensor<1x64xf32> + %3963 = "ttir.relu"(%3961, %3962) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3964 = tensor.empty() : tensor<1x64xf32> + %3965 = "ttir.relu"(%3963, %3964) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3966 = tensor.empty() : tensor<1x64xf32> + %3967 = "ttir.relu"(%3965, %3966) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3968 = tensor.empty() : tensor<1x64xf32> + %3969 = "ttir.relu"(%3967, %3968) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3970 = tensor.empty() : tensor<1x64xf32> + %3971 = "ttir.relu"(%3969, %3970) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3972 = tensor.empty() : tensor<1x64xf32> + %3973 = "ttir.relu"(%3971, %3972) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3974 = tensor.empty() : tensor<1x64xf32> + %3975 = "ttir.relu"(%3973, %3974) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3976 = tensor.empty() : tensor<1x64xf32> + %3977 = "ttir.relu"(%3975, %3976) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3978 = tensor.empty() : tensor<1x64xf32> + %3979 = "ttir.relu"(%3977, %3978) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3980 = tensor.empty() : tensor<1x64xf32> + %3981 = "ttir.relu"(%3979, %3980) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3982 = tensor.empty() : tensor<1x64xf32> + %3983 = "ttir.relu"(%3981, %3982) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3984 = tensor.empty() : tensor<1x64xf32> + %3985 = "ttir.relu"(%3983, %3984) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3986 = tensor.empty() : tensor<1x64xf32> + %3987 = "ttir.relu"(%3985, %3986) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3988 = tensor.empty() : tensor<1x64xf32> + %3989 = "ttir.relu"(%3987, %3988) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3990 = tensor.empty() : tensor<1x64xf32> + %3991 = "ttir.relu"(%3989, %3990) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3992 = tensor.empty() : tensor<1x64xf32> + %3993 = "ttir.relu"(%3991, %3992) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3994 = tensor.empty() : tensor<1x64xf32> + %3995 = "ttir.relu"(%3993, %3994) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3996 = tensor.empty() : tensor<1x64xf32> + %3997 = "ttir.relu"(%3995, %3996) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %3998 = tensor.empty() : tensor<1x64xf32> + %3999 = "ttir.relu"(%3997, %3998) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4000 = tensor.empty() : tensor<1x64xf32> + %4001 = "ttir.relu"(%3999, %4000) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4002 = tensor.empty() : tensor<1x64xf32> + %4003 = "ttir.relu"(%4001, %4002) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4004 = tensor.empty() : tensor<1x64xf32> + %4005 = "ttir.relu"(%4003, %4004) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4006 = tensor.empty() : tensor<1x64xf32> + %4007 = "ttir.relu"(%4005, %4006) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4008 = tensor.empty() : tensor<1x64xf32> + %4009 = "ttir.relu"(%4007, %4008) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4010 = tensor.empty() : tensor<1x64xf32> + %4011 = "ttir.relu"(%4009, %4010) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4012 = tensor.empty() : tensor<1x64xf32> + %4013 = "ttir.relu"(%4011, %4012) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4014 = tensor.empty() : tensor<1x64xf32> + %4015 = "ttir.relu"(%4013, %4014) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4016 = tensor.empty() : tensor<1x64xf32> + %4017 = "ttir.relu"(%4015, %4016) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4018 = tensor.empty() : tensor<1x64xf32> + %4019 = "ttir.relu"(%4017, %4018) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4020 = tensor.empty() : tensor<1x64xf32> + %4021 = "ttir.relu"(%4019, %4020) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4022 = tensor.empty() : tensor<1x64xf32> + %4023 = "ttir.relu"(%4021, %4022) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4024 = tensor.empty() : tensor<1x64xf32> + %4025 = "ttir.relu"(%4023, %4024) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4026 = tensor.empty() : tensor<1x64xf32> + %4027 = "ttir.relu"(%4025, %4026) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4028 = tensor.empty() : tensor<1x64xf32> + %4029 = "ttir.relu"(%4027, %4028) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4030 = tensor.empty() : tensor<1x64xf32> + %4031 = "ttir.relu"(%4029, %4030) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4032 = tensor.empty() : tensor<1x64xf32> + %4033 = "ttir.relu"(%4031, %4032) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4034 = tensor.empty() : tensor<1x64xf32> + %4035 = "ttir.relu"(%4033, %4034) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4036 = tensor.empty() : tensor<1x64xf32> + %4037 = "ttir.relu"(%4035, %4036) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4038 = tensor.empty() : tensor<1x64xf32> + %4039 = "ttir.relu"(%4037, %4038) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4040 = tensor.empty() : tensor<1x64xf32> + %4041 = "ttir.relu"(%4039, %4040) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4042 = tensor.empty() : tensor<1x64xf32> + %4043 = "ttir.relu"(%4041, %4042) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4044 = tensor.empty() : tensor<1x64xf32> + %4045 = "ttir.relu"(%4043, %4044) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4046 = tensor.empty() : tensor<1x64xf32> + %4047 = "ttir.relu"(%4045, %4046) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4048 = tensor.empty() : tensor<1x64xf32> + %4049 = "ttir.relu"(%4047, %4048) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4050 = tensor.empty() : tensor<1x64xf32> + %4051 = "ttir.relu"(%4049, %4050) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4052 = tensor.empty() : tensor<1x64xf32> + %4053 = "ttir.relu"(%4051, %4052) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4054 = tensor.empty() : tensor<1x64xf32> + %4055 = "ttir.relu"(%4053, %4054) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4056 = tensor.empty() : tensor<1x64xf32> + %4057 = "ttir.relu"(%4055, %4056) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4058 = tensor.empty() : tensor<1x64xf32> + %4059 = "ttir.relu"(%4057, %4058) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4060 = tensor.empty() : tensor<1x64xf32> + %4061 = "ttir.relu"(%4059, %4060) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4062 = tensor.empty() : tensor<1x64xf32> + %4063 = "ttir.relu"(%4061, %4062) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4064 = tensor.empty() : tensor<1x64xf32> + %4065 = "ttir.relu"(%4063, %4064) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4066 = tensor.empty() : tensor<1x64xf32> + %4067 = "ttir.relu"(%4065, %4066) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4068 = tensor.empty() : tensor<1x64xf32> + %4069 = "ttir.relu"(%4067, %4068) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4070 = tensor.empty() : tensor<1x64xf32> + %4071 = "ttir.relu"(%4069, %4070) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4072 = tensor.empty() : tensor<1x64xf32> + %4073 = "ttir.relu"(%4071, %4072) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4074 = tensor.empty() : tensor<1x64xf32> + %4075 = "ttir.relu"(%4073, %4074) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4076 = tensor.empty() : tensor<1x64xf32> + %4077 = "ttir.relu"(%4075, %4076) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4078 = tensor.empty() : tensor<1x64xf32> + %4079 = "ttir.relu"(%4077, %4078) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4080 = tensor.empty() : tensor<1x64xf32> + %4081 = "ttir.relu"(%4079, %4080) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4082 = tensor.empty() : tensor<1x64xf32> + %4083 = "ttir.relu"(%4081, %4082) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4084 = tensor.empty() : tensor<1x64xf32> + %4085 = "ttir.relu"(%4083, %4084) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4086 = tensor.empty() : tensor<1x64xf32> + %4087 = "ttir.relu"(%4085, %4086) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4088 = tensor.empty() : tensor<1x64xf32> + %4089 = "ttir.relu"(%4087, %4088) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4090 = tensor.empty() : tensor<1x64xf32> + %4091 = "ttir.relu"(%4089, %4090) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4092 = tensor.empty() : tensor<1x64xf32> + %4093 = "ttir.relu"(%4091, %4092) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4094 = tensor.empty() : tensor<1x64xf32> + %4095 = "ttir.relu"(%4093, %4094) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4096 = tensor.empty() : tensor<1x64xf32> + %4097 = "ttir.relu"(%4095, %4096) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4098 = tensor.empty() : tensor<1x64xf32> + %4099 = "ttir.relu"(%4097, %4098) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4100 = tensor.empty() : tensor<1x64xf32> + %4101 = "ttir.relu"(%4099, %4100) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4102 = tensor.empty() : tensor<1x64xf32> + %4103 = "ttir.relu"(%4101, %4102) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4104 = tensor.empty() : tensor<1x64xf32> + %4105 = "ttir.relu"(%4103, %4104) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4106 = tensor.empty() : tensor<1x64xf32> + %4107 = "ttir.relu"(%4105, %4106) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4108 = tensor.empty() : tensor<1x64xf32> + %4109 = "ttir.relu"(%4107, %4108) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4110 = tensor.empty() : tensor<1x64xf32> + %4111 = "ttir.relu"(%4109, %4110) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4112 = tensor.empty() : tensor<1x64xf32> + %4113 = "ttir.relu"(%4111, %4112) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4114 = tensor.empty() : tensor<1x64xf32> + %4115 = "ttir.relu"(%4113, %4114) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4116 = tensor.empty() : tensor<1x64xf32> + %4117 = "ttir.relu"(%4115, %4116) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4118 = tensor.empty() : tensor<1x64xf32> + %4119 = "ttir.relu"(%4117, %4118) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4120 = tensor.empty() : tensor<1x64xf32> + %4121 = "ttir.relu"(%4119, %4120) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4122 = tensor.empty() : tensor<1x64xf32> + %4123 = "ttir.relu"(%4121, %4122) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4124 = tensor.empty() : tensor<1x64xf32> + %4125 = "ttir.relu"(%4123, %4124) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4126 = tensor.empty() : tensor<1x64xf32> + %4127 = "ttir.relu"(%4125, %4126) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4128 = tensor.empty() : tensor<1x64xf32> + %4129 = "ttir.relu"(%4127, %4128) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4130 = tensor.empty() : tensor<1x64xf32> + %4131 = "ttir.relu"(%4129, %4130) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4132 = tensor.empty() : tensor<1x64xf32> + %4133 = "ttir.relu"(%4131, %4132) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4134 = tensor.empty() : tensor<1x64xf32> + %4135 = "ttir.relu"(%4133, %4134) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4136 = tensor.empty() : tensor<1x64xf32> + %4137 = "ttir.relu"(%4135, %4136) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4138 = tensor.empty() : tensor<1x64xf32> + %4139 = "ttir.relu"(%4137, %4138) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4140 = tensor.empty() : tensor<1x64xf32> + %4141 = "ttir.relu"(%4139, %4140) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4142 = tensor.empty() : tensor<1x64xf32> + %4143 = "ttir.relu"(%4141, %4142) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4144 = tensor.empty() : tensor<1x64xf32> + %4145 = "ttir.relu"(%4143, %4144) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4146 = tensor.empty() : tensor<1x64xf32> + %4147 = "ttir.relu"(%4145, %4146) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4148 = tensor.empty() : tensor<1x64xf32> + %4149 = "ttir.relu"(%4147, %4148) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4150 = tensor.empty() : tensor<1x64xf32> + %4151 = "ttir.relu"(%4149, %4150) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4152 = tensor.empty() : tensor<1x64xf32> + %4153 = "ttir.relu"(%4151, %4152) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4154 = tensor.empty() : tensor<1x64xf32> + %4155 = "ttir.relu"(%4153, %4154) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4156 = tensor.empty() : tensor<1x64xf32> + %4157 = "ttir.relu"(%4155, %4156) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4158 = tensor.empty() : tensor<1x64xf32> + %4159 = "ttir.relu"(%4157, %4158) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4160 = tensor.empty() : tensor<1x64xf32> + %4161 = "ttir.relu"(%4159, %4160) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4162 = tensor.empty() : tensor<1x64xf32> + %4163 = "ttir.relu"(%4161, %4162) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4164 = tensor.empty() : tensor<1x64xf32> + %4165 = "ttir.relu"(%4163, %4164) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4166 = tensor.empty() : tensor<1x64xf32> + %4167 = "ttir.relu"(%4165, %4166) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4168 = tensor.empty() : tensor<1x64xf32> + %4169 = "ttir.relu"(%4167, %4168) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4170 = tensor.empty() : tensor<1x64xf32> + %4171 = "ttir.relu"(%4169, %4170) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4172 = tensor.empty() : tensor<1x64xf32> + %4173 = "ttir.relu"(%4171, %4172) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4174 = tensor.empty() : tensor<1x64xf32> + %4175 = "ttir.relu"(%4173, %4174) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4176 = tensor.empty() : tensor<1x64xf32> + %4177 = "ttir.relu"(%4175, %4176) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4178 = tensor.empty() : tensor<1x64xf32> + %4179 = "ttir.relu"(%4177, %4178) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4180 = tensor.empty() : tensor<1x64xf32> + %4181 = "ttir.relu"(%4179, %4180) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4182 = tensor.empty() : tensor<1x64xf32> + %4183 = "ttir.relu"(%4181, %4182) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4184 = tensor.empty() : tensor<1x64xf32> + %4185 = "ttir.relu"(%4183, %4184) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4186 = tensor.empty() : tensor<1x64xf32> + %4187 = "ttir.relu"(%4185, %4186) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4188 = tensor.empty() : tensor<1x64xf32> + %4189 = "ttir.relu"(%4187, %4188) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4190 = tensor.empty() : tensor<1x64xf32> + %4191 = "ttir.relu"(%4189, %4190) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4192 = tensor.empty() : tensor<1x64xf32> + %4193 = "ttir.relu"(%4191, %4192) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4194 = tensor.empty() : tensor<1x64xf32> + %4195 = "ttir.relu"(%4193, %4194) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4196 = tensor.empty() : tensor<1x64xf32> + %4197 = "ttir.relu"(%4195, %4196) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4198 = tensor.empty() : tensor<1x64xf32> + %4199 = "ttir.relu"(%4197, %4198) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4200 = tensor.empty() : tensor<1x64xf32> + %4201 = "ttir.relu"(%4199, %4200) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4202 = tensor.empty() : tensor<1x64xf32> + %4203 = "ttir.relu"(%4201, %4202) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4204 = tensor.empty() : tensor<1x64xf32> + %4205 = "ttir.relu"(%4203, %4204) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4206 = tensor.empty() : tensor<1x64xf32> + %4207 = "ttir.relu"(%4205, %4206) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4208 = tensor.empty() : tensor<1x64xf32> + %4209 = "ttir.relu"(%4207, %4208) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4210 = tensor.empty() : tensor<1x64xf32> + %4211 = "ttir.relu"(%4209, %4210) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4212 = tensor.empty() : tensor<1x64xf32> + %4213 = "ttir.relu"(%4211, %4212) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4214 = tensor.empty() : tensor<1x64xf32> + %4215 = "ttir.relu"(%4213, %4214) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4216 = tensor.empty() : tensor<1x64xf32> + %4217 = "ttir.relu"(%4215, %4216) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4218 = tensor.empty() : tensor<1x64xf32> + %4219 = "ttir.relu"(%4217, %4218) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4220 = tensor.empty() : tensor<1x64xf32> + %4221 = "ttir.relu"(%4219, %4220) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4222 = tensor.empty() : tensor<1x64xf32> + %4223 = "ttir.relu"(%4221, %4222) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4224 = tensor.empty() : tensor<1x64xf32> + %4225 = "ttir.relu"(%4223, %4224) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4226 = tensor.empty() : tensor<1x64xf32> + %4227 = "ttir.relu"(%4225, %4226) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4228 = tensor.empty() : tensor<1x64xf32> + %4229 = "ttir.relu"(%4227, %4228) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4230 = tensor.empty() : tensor<1x64xf32> + %4231 = "ttir.relu"(%4229, %4230) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4232 = tensor.empty() : tensor<1x64xf32> + %4233 = "ttir.relu"(%4231, %4232) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4234 = tensor.empty() : tensor<1x64xf32> + %4235 = "ttir.relu"(%4233, %4234) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4236 = tensor.empty() : tensor<1x64xf32> + %4237 = "ttir.relu"(%4235, %4236) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4238 = tensor.empty() : tensor<1x64xf32> + %4239 = "ttir.relu"(%4237, %4238) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4240 = tensor.empty() : tensor<1x64xf32> + %4241 = "ttir.relu"(%4239, %4240) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4242 = tensor.empty() : tensor<1x64xf32> + %4243 = "ttir.relu"(%4241, %4242) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4244 = tensor.empty() : tensor<1x64xf32> + %4245 = "ttir.relu"(%4243, %4244) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4246 = tensor.empty() : tensor<1x64xf32> + %4247 = "ttir.relu"(%4245, %4246) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4248 = tensor.empty() : tensor<1x64xf32> + %4249 = "ttir.relu"(%4247, %4248) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4250 = tensor.empty() : tensor<1x64xf32> + %4251 = "ttir.relu"(%4249, %4250) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4252 = tensor.empty() : tensor<1x64xf32> + %4253 = "ttir.relu"(%4251, %4252) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4254 = tensor.empty() : tensor<1x64xf32> + %4255 = "ttir.relu"(%4253, %4254) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4256 = tensor.empty() : tensor<1x64xf32> + %4257 = "ttir.relu"(%4255, %4256) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4258 = tensor.empty() : tensor<1x64xf32> + %4259 = "ttir.relu"(%4257, %4258) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4260 = tensor.empty() : tensor<1x64xf32> + %4261 = "ttir.relu"(%4259, %4260) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4262 = tensor.empty() : tensor<1x64xf32> + %4263 = "ttir.relu"(%4261, %4262) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4264 = tensor.empty() : tensor<1x64xf32> + %4265 = "ttir.relu"(%4263, %4264) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4266 = tensor.empty() : tensor<1x64xf32> + %4267 = "ttir.relu"(%4265, %4266) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4268 = tensor.empty() : tensor<1x64xf32> + %4269 = "ttir.relu"(%4267, %4268) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4270 = tensor.empty() : tensor<1x64xf32> + %4271 = "ttir.relu"(%4269, %4270) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4272 = tensor.empty() : tensor<1x64xf32> + %4273 = "ttir.relu"(%4271, %4272) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4274 = tensor.empty() : tensor<1x64xf32> + %4275 = "ttir.relu"(%4273, %4274) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4276 = tensor.empty() : tensor<1x64xf32> + %4277 = "ttir.relu"(%4275, %4276) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4278 = tensor.empty() : tensor<1x64xf32> + %4279 = "ttir.relu"(%4277, %4278) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4280 = tensor.empty() : tensor<1x64xf32> + %4281 = "ttir.relu"(%4279, %4280) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4282 = tensor.empty() : tensor<1x64xf32> + %4283 = "ttir.relu"(%4281, %4282) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4284 = tensor.empty() : tensor<1x64xf32> + %4285 = "ttir.relu"(%4283, %4284) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4286 = tensor.empty() : tensor<1x64xf32> + %4287 = "ttir.relu"(%4285, %4286) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4288 = tensor.empty() : tensor<1x64xf32> + %4289 = "ttir.relu"(%4287, %4288) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4290 = tensor.empty() : tensor<1x64xf32> + %4291 = "ttir.relu"(%4289, %4290) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4292 = tensor.empty() : tensor<1x64xf32> + %4293 = "ttir.relu"(%4291, %4292) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4294 = tensor.empty() : tensor<1x64xf32> + %4295 = "ttir.relu"(%4293, %4294) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4296 = tensor.empty() : tensor<1x64xf32> + %4297 = "ttir.relu"(%4295, %4296) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4298 = tensor.empty() : tensor<1x64xf32> + %4299 = "ttir.relu"(%4297, %4298) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4300 = tensor.empty() : tensor<1x64xf32> + %4301 = "ttir.relu"(%4299, %4300) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4302 = tensor.empty() : tensor<1x64xf32> + %4303 = "ttir.relu"(%4301, %4302) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4304 = tensor.empty() : tensor<1x64xf32> + %4305 = "ttir.relu"(%4303, %4304) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4306 = tensor.empty() : tensor<1x64xf32> + %4307 = "ttir.relu"(%4305, %4306) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4308 = tensor.empty() : tensor<1x64xf32> + %4309 = "ttir.relu"(%4307, %4308) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4310 = tensor.empty() : tensor<1x64xf32> + %4311 = "ttir.relu"(%4309, %4310) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4312 = tensor.empty() : tensor<1x64xf32> + %4313 = "ttir.relu"(%4311, %4312) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4314 = tensor.empty() : tensor<1x64xf32> + %4315 = "ttir.relu"(%4313, %4314) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4316 = tensor.empty() : tensor<1x64xf32> + %4317 = "ttir.relu"(%4315, %4316) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4318 = tensor.empty() : tensor<1x64xf32> + %4319 = "ttir.relu"(%4317, %4318) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4320 = tensor.empty() : tensor<1x64xf32> + %4321 = "ttir.relu"(%4319, %4320) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4322 = tensor.empty() : tensor<1x64xf32> + %4323 = "ttir.relu"(%4321, %4322) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4324 = tensor.empty() : tensor<1x64xf32> + %4325 = "ttir.relu"(%4323, %4324) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4326 = tensor.empty() : tensor<1x64xf32> + %4327 = "ttir.relu"(%4325, %4326) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4328 = tensor.empty() : tensor<1x64xf32> + %4329 = "ttir.relu"(%4327, %4328) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4330 = tensor.empty() : tensor<1x64xf32> + %4331 = "ttir.relu"(%4329, %4330) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4332 = tensor.empty() : tensor<1x64xf32> + %4333 = "ttir.relu"(%4331, %4332) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4334 = tensor.empty() : tensor<1x64xf32> + %4335 = "ttir.relu"(%4333, %4334) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4336 = tensor.empty() : tensor<1x64xf32> + %4337 = "ttir.relu"(%4335, %4336) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4338 = tensor.empty() : tensor<1x64xf32> + %4339 = "ttir.relu"(%4337, %4338) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4340 = tensor.empty() : tensor<1x64xf32> + %4341 = "ttir.relu"(%4339, %4340) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4342 = tensor.empty() : tensor<1x64xf32> + %4343 = "ttir.relu"(%4341, %4342) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4344 = tensor.empty() : tensor<1x64xf32> + %4345 = "ttir.relu"(%4343, %4344) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4346 = tensor.empty() : tensor<1x64xf32> + %4347 = "ttir.relu"(%4345, %4346) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4348 = tensor.empty() : tensor<1x64xf32> + %4349 = "ttir.relu"(%4347, %4348) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4350 = tensor.empty() : tensor<1x64xf32> + %4351 = "ttir.relu"(%4349, %4350) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4352 = tensor.empty() : tensor<1x64xf32> + %4353 = "ttir.relu"(%4351, %4352) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4354 = tensor.empty() : tensor<1x64xf32> + %4355 = "ttir.relu"(%4353, %4354) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4356 = tensor.empty() : tensor<1x64xf32> + %4357 = "ttir.relu"(%4355, %4356) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4358 = tensor.empty() : tensor<1x64xf32> + %4359 = "ttir.relu"(%4357, %4358) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4360 = tensor.empty() : tensor<1x64xf32> + %4361 = "ttir.relu"(%4359, %4360) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4362 = tensor.empty() : tensor<1x64xf32> + %4363 = "ttir.relu"(%4361, %4362) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4364 = tensor.empty() : tensor<1x64xf32> + %4365 = "ttir.relu"(%4363, %4364) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4366 = tensor.empty() : tensor<1x64xf32> + %4367 = "ttir.relu"(%4365, %4366) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4368 = tensor.empty() : tensor<1x64xf32> + %4369 = "ttir.relu"(%4367, %4368) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4370 = tensor.empty() : tensor<1x64xf32> + %4371 = "ttir.relu"(%4369, %4370) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4372 = tensor.empty() : tensor<1x64xf32> + %4373 = "ttir.relu"(%4371, %4372) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4374 = tensor.empty() : tensor<1x64xf32> + %4375 = "ttir.relu"(%4373, %4374) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4376 = tensor.empty() : tensor<1x64xf32> + %4377 = "ttir.relu"(%4375, %4376) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4378 = tensor.empty() : tensor<1x64xf32> + %4379 = "ttir.relu"(%4377, %4378) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4380 = tensor.empty() : tensor<1x64xf32> + %4381 = "ttir.relu"(%4379, %4380) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4382 = tensor.empty() : tensor<1x64xf32> + %4383 = "ttir.relu"(%4381, %4382) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4384 = tensor.empty() : tensor<1x64xf32> + %4385 = "ttir.relu"(%4383, %4384) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4386 = tensor.empty() : tensor<1x64xf32> + %4387 = "ttir.relu"(%4385, %4386) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4388 = tensor.empty() : tensor<1x64xf32> + %4389 = "ttir.relu"(%4387, %4388) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4390 = tensor.empty() : tensor<1x64xf32> + %4391 = "ttir.relu"(%4389, %4390) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4392 = tensor.empty() : tensor<1x64xf32> + %4393 = "ttir.relu"(%4391, %4392) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4394 = tensor.empty() : tensor<1x64xf32> + %4395 = "ttir.relu"(%4393, %4394) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4396 = tensor.empty() : tensor<1x64xf32> + %4397 = "ttir.relu"(%4395, %4396) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4398 = tensor.empty() : tensor<1x64xf32> + %4399 = "ttir.relu"(%4397, %4398) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4400 = tensor.empty() : tensor<1x64xf32> + %4401 = "ttir.relu"(%4399, %4400) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4402 = tensor.empty() : tensor<1x64xf32> + %4403 = "ttir.relu"(%4401, %4402) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4404 = tensor.empty() : tensor<1x64xf32> + %4405 = "ttir.relu"(%4403, %4404) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4406 = tensor.empty() : tensor<1x64xf32> + %4407 = "ttir.relu"(%4405, %4406) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4408 = tensor.empty() : tensor<1x64xf32> + %4409 = "ttir.relu"(%4407, %4408) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4410 = tensor.empty() : tensor<1x64xf32> + %4411 = "ttir.relu"(%4409, %4410) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4412 = tensor.empty() : tensor<1x64xf32> + %4413 = "ttir.relu"(%4411, %4412) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4414 = tensor.empty() : tensor<1x64xf32> + %4415 = "ttir.relu"(%4413, %4414) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4416 = tensor.empty() : tensor<1x64xf32> + %4417 = "ttir.relu"(%4415, %4416) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4418 = tensor.empty() : tensor<1x64xf32> + %4419 = "ttir.relu"(%4417, %4418) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4420 = tensor.empty() : tensor<1x64xf32> + %4421 = "ttir.relu"(%4419, %4420) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4422 = tensor.empty() : tensor<1x64xf32> + %4423 = "ttir.relu"(%4421, %4422) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4424 = tensor.empty() : tensor<1x64xf32> + %4425 = "ttir.relu"(%4423, %4424) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4426 = tensor.empty() : tensor<1x64xf32> + %4427 = "ttir.relu"(%4425, %4426) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4428 = tensor.empty() : tensor<1x64xf32> + %4429 = "ttir.relu"(%4427, %4428) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4430 = tensor.empty() : tensor<1x64xf32> + %4431 = "ttir.relu"(%4429, %4430) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4432 = tensor.empty() : tensor<1x64xf32> + %4433 = "ttir.relu"(%4431, %4432) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4434 = tensor.empty() : tensor<1x64xf32> + %4435 = "ttir.relu"(%4433, %4434) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4436 = tensor.empty() : tensor<1x64xf32> + %4437 = "ttir.relu"(%4435, %4436) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4438 = tensor.empty() : tensor<1x64xf32> + %4439 = "ttir.relu"(%4437, %4438) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4440 = tensor.empty() : tensor<1x64xf32> + %4441 = "ttir.relu"(%4439, %4440) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4442 = tensor.empty() : tensor<1x64xf32> + %4443 = "ttir.relu"(%4441, %4442) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4444 = tensor.empty() : tensor<1x64xf32> + %4445 = "ttir.relu"(%4443, %4444) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4446 = tensor.empty() : tensor<1x64xf32> + %4447 = "ttir.relu"(%4445, %4446) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4448 = tensor.empty() : tensor<1x64xf32> + %4449 = "ttir.relu"(%4447, %4448) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4450 = tensor.empty() : tensor<1x64xf32> + %4451 = "ttir.relu"(%4449, %4450) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4452 = tensor.empty() : tensor<1x64xf32> + %4453 = "ttir.relu"(%4451, %4452) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4454 = tensor.empty() : tensor<1x64xf32> + %4455 = "ttir.relu"(%4453, %4454) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4456 = tensor.empty() : tensor<1x64xf32> + %4457 = "ttir.relu"(%4455, %4456) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4458 = tensor.empty() : tensor<1x64xf32> + %4459 = "ttir.relu"(%4457, %4458) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4460 = tensor.empty() : tensor<1x64xf32> + %4461 = "ttir.relu"(%4459, %4460) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4462 = tensor.empty() : tensor<1x64xf32> + %4463 = "ttir.relu"(%4461, %4462) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4464 = tensor.empty() : tensor<1x64xf32> + %4465 = "ttir.relu"(%4463, %4464) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4466 = tensor.empty() : tensor<1x64xf32> + %4467 = "ttir.relu"(%4465, %4466) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4468 = tensor.empty() : tensor<1x64xf32> + %4469 = "ttir.relu"(%4467, %4468) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4470 = tensor.empty() : tensor<1x64xf32> + %4471 = "ttir.relu"(%4469, %4470) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4472 = tensor.empty() : tensor<1x64xf32> + %4473 = "ttir.relu"(%4471, %4472) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4474 = tensor.empty() : tensor<1x64xf32> + %4475 = "ttir.relu"(%4473, %4474) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4476 = tensor.empty() : tensor<1x64xf32> + %4477 = "ttir.relu"(%4475, %4476) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4478 = tensor.empty() : tensor<1x64xf32> + %4479 = "ttir.relu"(%4477, %4478) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4480 = tensor.empty() : tensor<1x64xf32> + %4481 = "ttir.relu"(%4479, %4480) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4482 = tensor.empty() : tensor<1x64xf32> + %4483 = "ttir.relu"(%4481, %4482) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4484 = tensor.empty() : tensor<1x64xf32> + %4485 = "ttir.relu"(%4483, %4484) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4486 = tensor.empty() : tensor<1x64xf32> + %4487 = "ttir.relu"(%4485, %4486) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4488 = tensor.empty() : tensor<1x64xf32> + %4489 = "ttir.relu"(%4487, %4488) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4490 = tensor.empty() : tensor<1x64xf32> + %4491 = "ttir.relu"(%4489, %4490) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4492 = tensor.empty() : tensor<1x64xf32> + %4493 = "ttir.relu"(%4491, %4492) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4494 = tensor.empty() : tensor<1x64xf32> + %4495 = "ttir.relu"(%4493, %4494) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4496 = tensor.empty() : tensor<1x64xf32> + %4497 = "ttir.relu"(%4495, %4496) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4498 = tensor.empty() : tensor<1x64xf32> + %4499 = "ttir.relu"(%4497, %4498) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4500 = tensor.empty() : tensor<1x64xf32> + %4501 = "ttir.relu"(%4499, %4500) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4502 = tensor.empty() : tensor<1x64xf32> + %4503 = "ttir.relu"(%4501, %4502) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4504 = tensor.empty() : tensor<1x64xf32> + %4505 = "ttir.relu"(%4503, %4504) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4506 = tensor.empty() : tensor<1x64xf32> + %4507 = "ttir.relu"(%4505, %4506) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4508 = tensor.empty() : tensor<1x64xf32> + %4509 = "ttir.relu"(%4507, %4508) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4510 = tensor.empty() : tensor<1x64xf32> + %4511 = "ttir.relu"(%4509, %4510) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4512 = tensor.empty() : tensor<1x64xf32> + %4513 = "ttir.relu"(%4511, %4512) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4514 = tensor.empty() : tensor<1x64xf32> + %4515 = "ttir.relu"(%4513, %4514) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4516 = tensor.empty() : tensor<1x64xf32> + %4517 = "ttir.relu"(%4515, %4516) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4518 = tensor.empty() : tensor<1x64xf32> + %4519 = "ttir.relu"(%4517, %4518) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4520 = tensor.empty() : tensor<1x64xf32> + %4521 = "ttir.relu"(%4519, %4520) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4522 = tensor.empty() : tensor<1x64xf32> + %4523 = "ttir.relu"(%4521, %4522) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4524 = tensor.empty() : tensor<1x64xf32> + %4525 = "ttir.relu"(%4523, %4524) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4526 = tensor.empty() : tensor<1x64xf32> + %4527 = "ttir.relu"(%4525, %4526) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4528 = tensor.empty() : tensor<1x64xf32> + %4529 = "ttir.relu"(%4527, %4528) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4530 = tensor.empty() : tensor<1x64xf32> + %4531 = "ttir.relu"(%4529, %4530) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4532 = tensor.empty() : tensor<1x64xf32> + %4533 = "ttir.relu"(%4531, %4532) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4534 = tensor.empty() : tensor<1x64xf32> + %4535 = "ttir.relu"(%4533, %4534) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4536 = tensor.empty() : tensor<1x64xf32> + %4537 = "ttir.relu"(%4535, %4536) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4538 = tensor.empty() : tensor<1x64xf32> + %4539 = "ttir.relu"(%4537, %4538) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4540 = tensor.empty() : tensor<1x64xf32> + %4541 = "ttir.relu"(%4539, %4540) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4542 = tensor.empty() : tensor<1x64xf32> + %4543 = "ttir.relu"(%4541, %4542) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4544 = tensor.empty() : tensor<1x64xf32> + %4545 = "ttir.relu"(%4543, %4544) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4546 = tensor.empty() : tensor<1x64xf32> + %4547 = "ttir.relu"(%4545, %4546) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4548 = tensor.empty() : tensor<1x64xf32> + %4549 = "ttir.relu"(%4547, %4548) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4550 = tensor.empty() : tensor<1x64xf32> + %4551 = "ttir.relu"(%4549, %4550) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4552 = tensor.empty() : tensor<1x64xf32> + %4553 = "ttir.relu"(%4551, %4552) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4554 = tensor.empty() : tensor<1x64xf32> + %4555 = "ttir.relu"(%4553, %4554) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4556 = tensor.empty() : tensor<1x64xf32> + %4557 = "ttir.relu"(%4555, %4556) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4558 = tensor.empty() : tensor<1x64xf32> + %4559 = "ttir.relu"(%4557, %4558) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4560 = tensor.empty() : tensor<1x64xf32> + %4561 = "ttir.relu"(%4559, %4560) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4562 = tensor.empty() : tensor<1x64xf32> + %4563 = "ttir.relu"(%4561, %4562) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4564 = tensor.empty() : tensor<1x64xf32> + %4565 = "ttir.relu"(%4563, %4564) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4566 = tensor.empty() : tensor<1x64xf32> + %4567 = "ttir.relu"(%4565, %4566) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4568 = tensor.empty() : tensor<1x64xf32> + %4569 = "ttir.relu"(%4567, %4568) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4570 = tensor.empty() : tensor<1x64xf32> + %4571 = "ttir.relu"(%4569, %4570) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4572 = tensor.empty() : tensor<1x64xf32> + %4573 = "ttir.relu"(%4571, %4572) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4574 = tensor.empty() : tensor<1x64xf32> + %4575 = "ttir.relu"(%4573, %4574) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4576 = tensor.empty() : tensor<1x64xf32> + %4577 = "ttir.relu"(%4575, %4576) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4578 = tensor.empty() : tensor<1x64xf32> + %4579 = "ttir.relu"(%4577, %4578) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4580 = tensor.empty() : tensor<1x64xf32> + %4581 = "ttir.relu"(%4579, %4580) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4582 = tensor.empty() : tensor<1x64xf32> + %4583 = "ttir.relu"(%4581, %4582) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4584 = tensor.empty() : tensor<1x64xf32> + %4585 = "ttir.relu"(%4583, %4584) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4586 = tensor.empty() : tensor<1x64xf32> + %4587 = "ttir.relu"(%4585, %4586) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4588 = tensor.empty() : tensor<1x64xf32> + %4589 = "ttir.relu"(%4587, %4588) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4590 = tensor.empty() : tensor<1x64xf32> + %4591 = "ttir.relu"(%4589, %4590) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4592 = tensor.empty() : tensor<1x64xf32> + %4593 = "ttir.relu"(%4591, %4592) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4594 = tensor.empty() : tensor<1x64xf32> + %4595 = "ttir.relu"(%4593, %4594) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4596 = tensor.empty() : tensor<1x64xf32> + %4597 = "ttir.relu"(%4595, %4596) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4598 = tensor.empty() : tensor<1x64xf32> + %4599 = "ttir.relu"(%4597, %4598) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4600 = tensor.empty() : tensor<1x64xf32> + %4601 = "ttir.relu"(%4599, %4600) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4602 = tensor.empty() : tensor<1x64xf32> + %4603 = "ttir.relu"(%4601, %4602) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4604 = tensor.empty() : tensor<1x64xf32> + %4605 = "ttir.relu"(%4603, %4604) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4606 = tensor.empty() : tensor<1x64xf32> + %4607 = "ttir.relu"(%4605, %4606) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4608 = tensor.empty() : tensor<1x64xf32> + %4609 = "ttir.relu"(%4607, %4608) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4610 = tensor.empty() : tensor<1x64xf32> + %4611 = "ttir.relu"(%4609, %4610) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4612 = tensor.empty() : tensor<1x64xf32> + %4613 = "ttir.relu"(%4611, %4612) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4614 = tensor.empty() : tensor<1x64xf32> + %4615 = "ttir.relu"(%4613, %4614) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4616 = tensor.empty() : tensor<1x64xf32> + %4617 = "ttir.relu"(%4615, %4616) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4618 = tensor.empty() : tensor<1x64xf32> + %4619 = "ttir.relu"(%4617, %4618) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4620 = tensor.empty() : tensor<1x64xf32> + %4621 = "ttir.relu"(%4619, %4620) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4622 = tensor.empty() : tensor<1x64xf32> + %4623 = "ttir.relu"(%4621, %4622) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4624 = tensor.empty() : tensor<1x64xf32> + %4625 = "ttir.relu"(%4623, %4624) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4626 = tensor.empty() : tensor<1x64xf32> + %4627 = "ttir.relu"(%4625, %4626) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4628 = tensor.empty() : tensor<1x64xf32> + %4629 = "ttir.relu"(%4627, %4628) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4630 = tensor.empty() : tensor<1x64xf32> + %4631 = "ttir.relu"(%4629, %4630) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4632 = tensor.empty() : tensor<1x64xf32> + %4633 = "ttir.relu"(%4631, %4632) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4634 = tensor.empty() : tensor<1x64xf32> + %4635 = "ttir.relu"(%4633, %4634) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4636 = tensor.empty() : tensor<1x64xf32> + %4637 = "ttir.relu"(%4635, %4636) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4638 = tensor.empty() : tensor<1x64xf32> + %4639 = "ttir.relu"(%4637, %4638) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4640 = tensor.empty() : tensor<1x64xf32> + %4641 = "ttir.relu"(%4639, %4640) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4642 = tensor.empty() : tensor<1x64xf32> + %4643 = "ttir.relu"(%4641, %4642) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4644 = tensor.empty() : tensor<1x64xf32> + %4645 = "ttir.relu"(%4643, %4644) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4646 = tensor.empty() : tensor<1x64xf32> + %4647 = "ttir.relu"(%4645, %4646) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4648 = tensor.empty() : tensor<1x64xf32> + %4649 = "ttir.relu"(%4647, %4648) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4650 = tensor.empty() : tensor<1x64xf32> + %4651 = "ttir.relu"(%4649, %4650) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4652 = tensor.empty() : tensor<1x64xf32> + %4653 = "ttir.relu"(%4651, %4652) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4654 = tensor.empty() : tensor<1x64xf32> + %4655 = "ttir.relu"(%4653, %4654) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4656 = tensor.empty() : tensor<1x64xf32> + %4657 = "ttir.relu"(%4655, %4656) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4658 = tensor.empty() : tensor<1x64xf32> + %4659 = "ttir.relu"(%4657, %4658) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4660 = tensor.empty() : tensor<1x64xf32> + %4661 = "ttir.relu"(%4659, %4660) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4662 = tensor.empty() : tensor<1x64xf32> + %4663 = "ttir.relu"(%4661, %4662) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4664 = tensor.empty() : tensor<1x64xf32> + %4665 = "ttir.relu"(%4663, %4664) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4666 = tensor.empty() : tensor<1x64xf32> + %4667 = "ttir.relu"(%4665, %4666) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4668 = tensor.empty() : tensor<1x64xf32> + %4669 = "ttir.relu"(%4667, %4668) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4670 = tensor.empty() : tensor<1x64xf32> + %4671 = "ttir.relu"(%4669, %4670) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4672 = tensor.empty() : tensor<1x64xf32> + %4673 = "ttir.relu"(%4671, %4672) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4674 = tensor.empty() : tensor<1x64xf32> + %4675 = "ttir.relu"(%4673, %4674) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4676 = tensor.empty() : tensor<1x64xf32> + %4677 = "ttir.relu"(%4675, %4676) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4678 = tensor.empty() : tensor<1x64xf32> + %4679 = "ttir.relu"(%4677, %4678) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4680 = tensor.empty() : tensor<1x64xf32> + %4681 = "ttir.relu"(%4679, %4680) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4682 = tensor.empty() : tensor<1x64xf32> + %4683 = "ttir.relu"(%4681, %4682) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4684 = tensor.empty() : tensor<1x64xf32> + %4685 = "ttir.relu"(%4683, %4684) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4686 = tensor.empty() : tensor<1x64xf32> + %4687 = "ttir.relu"(%4685, %4686) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4688 = tensor.empty() : tensor<1x64xf32> + %4689 = "ttir.relu"(%4687, %4688) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4690 = tensor.empty() : tensor<1x64xf32> + %4691 = "ttir.relu"(%4689, %4690) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4692 = tensor.empty() : tensor<1x64xf32> + %4693 = "ttir.relu"(%4691, %4692) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4694 = tensor.empty() : tensor<1x64xf32> + %4695 = "ttir.relu"(%4693, %4694) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4696 = tensor.empty() : tensor<1x64xf32> + %4697 = "ttir.relu"(%4695, %4696) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4698 = tensor.empty() : tensor<1x64xf32> + %4699 = "ttir.relu"(%4697, %4698) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4700 = tensor.empty() : tensor<1x64xf32> + %4701 = "ttir.relu"(%4699, %4700) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4702 = tensor.empty() : tensor<1x64xf32> + %4703 = "ttir.relu"(%4701, %4702) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4704 = tensor.empty() : tensor<1x64xf32> + %4705 = "ttir.relu"(%4703, %4704) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4706 = tensor.empty() : tensor<1x64xf32> + %4707 = "ttir.relu"(%4705, %4706) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4708 = tensor.empty() : tensor<1x64xf32> + %4709 = "ttir.relu"(%4707, %4708) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4710 = tensor.empty() : tensor<1x64xf32> + %4711 = "ttir.relu"(%4709, %4710) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4712 = tensor.empty() : tensor<1x64xf32> + %4713 = "ttir.relu"(%4711, %4712) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4714 = tensor.empty() : tensor<1x64xf32> + %4715 = "ttir.relu"(%4713, %4714) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4716 = tensor.empty() : tensor<1x64xf32> + %4717 = "ttir.relu"(%4715, %4716) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4718 = tensor.empty() : tensor<1x64xf32> + %4719 = "ttir.relu"(%4717, %4718) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4720 = tensor.empty() : tensor<1x64xf32> + %4721 = "ttir.relu"(%4719, %4720) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4722 = tensor.empty() : tensor<1x64xf32> + %4723 = "ttir.relu"(%4721, %4722) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4724 = tensor.empty() : tensor<1x64xf32> + %4725 = "ttir.relu"(%4723, %4724) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4726 = tensor.empty() : tensor<1x64xf32> + %4727 = "ttir.relu"(%4725, %4726) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4728 = tensor.empty() : tensor<1x64xf32> + %4729 = "ttir.relu"(%4727, %4728) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4730 = tensor.empty() : tensor<1x64xf32> + %4731 = "ttir.relu"(%4729, %4730) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4732 = tensor.empty() : tensor<1x64xf32> + %4733 = "ttir.relu"(%4731, %4732) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4734 = tensor.empty() : tensor<1x64xf32> + %4735 = "ttir.relu"(%4733, %4734) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4736 = tensor.empty() : tensor<1x64xf32> + %4737 = "ttir.relu"(%4735, %4736) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4738 = tensor.empty() : tensor<1x64xf32> + %4739 = "ttir.relu"(%4737, %4738) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4740 = tensor.empty() : tensor<1x64xf32> + %4741 = "ttir.relu"(%4739, %4740) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4742 = tensor.empty() : tensor<1x64xf32> + %4743 = "ttir.relu"(%4741, %4742) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4744 = tensor.empty() : tensor<1x64xf32> + %4745 = "ttir.relu"(%4743, %4744) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4746 = tensor.empty() : tensor<1x64xf32> + %4747 = "ttir.relu"(%4745, %4746) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4748 = tensor.empty() : tensor<1x64xf32> + %4749 = "ttir.relu"(%4747, %4748) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4750 = tensor.empty() : tensor<1x64xf32> + %4751 = "ttir.relu"(%4749, %4750) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4752 = tensor.empty() : tensor<1x64xf32> + %4753 = "ttir.relu"(%4751, %4752) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4754 = tensor.empty() : tensor<1x64xf32> + %4755 = "ttir.relu"(%4753, %4754) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4756 = tensor.empty() : tensor<1x64xf32> + %4757 = "ttir.relu"(%4755, %4756) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4758 = tensor.empty() : tensor<1x64xf32> + %4759 = "ttir.relu"(%4757, %4758) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4760 = tensor.empty() : tensor<1x64xf32> + %4761 = "ttir.relu"(%4759, %4760) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4762 = tensor.empty() : tensor<1x64xf32> + %4763 = "ttir.relu"(%4761, %4762) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4764 = tensor.empty() : tensor<1x64xf32> + %4765 = "ttir.relu"(%4763, %4764) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4766 = tensor.empty() : tensor<1x64xf32> + %4767 = "ttir.relu"(%4765, %4766) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4768 = tensor.empty() : tensor<1x64xf32> + %4769 = "ttir.relu"(%4767, %4768) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4770 = tensor.empty() : tensor<1x64xf32> + %4771 = "ttir.relu"(%4769, %4770) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4772 = tensor.empty() : tensor<1x64xf32> + %4773 = "ttir.relu"(%4771, %4772) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4774 = tensor.empty() : tensor<1x64xf32> + %4775 = "ttir.relu"(%4773, %4774) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4776 = tensor.empty() : tensor<1x64xf32> + %4777 = "ttir.relu"(%4775, %4776) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4778 = tensor.empty() : tensor<1x64xf32> + %4779 = "ttir.relu"(%4777, %4778) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4780 = tensor.empty() : tensor<1x64xf32> + %4781 = "ttir.relu"(%4779, %4780) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4782 = tensor.empty() : tensor<1x64xf32> + %4783 = "ttir.relu"(%4781, %4782) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4784 = tensor.empty() : tensor<1x64xf32> + %4785 = "ttir.relu"(%4783, %4784) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4786 = tensor.empty() : tensor<1x64xf32> + %4787 = "ttir.relu"(%4785, %4786) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4788 = tensor.empty() : tensor<1x64xf32> + %4789 = "ttir.relu"(%4787, %4788) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4790 = tensor.empty() : tensor<1x64xf32> + %4791 = "ttir.relu"(%4789, %4790) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4792 = tensor.empty() : tensor<1x64xf32> + %4793 = "ttir.relu"(%4791, %4792) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4794 = tensor.empty() : tensor<1x64xf32> + %4795 = "ttir.relu"(%4793, %4794) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4796 = tensor.empty() : tensor<1x64xf32> + %4797 = "ttir.relu"(%4795, %4796) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4798 = tensor.empty() : tensor<1x64xf32> + %4799 = "ttir.relu"(%4797, %4798) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4800 = tensor.empty() : tensor<1x64xf32> + %4801 = "ttir.relu"(%4799, %4800) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4802 = tensor.empty() : tensor<1x64xf32> + %4803 = "ttir.relu"(%4801, %4802) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4804 = tensor.empty() : tensor<1x64xf32> + %4805 = "ttir.relu"(%4803, %4804) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4806 = tensor.empty() : tensor<1x64xf32> + %4807 = "ttir.relu"(%4805, %4806) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4808 = tensor.empty() : tensor<1x64xf32> + %4809 = "ttir.relu"(%4807, %4808) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4810 = tensor.empty() : tensor<1x64xf32> + %4811 = "ttir.relu"(%4809, %4810) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4812 = tensor.empty() : tensor<1x64xf32> + %4813 = "ttir.relu"(%4811, %4812) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4814 = tensor.empty() : tensor<1x64xf32> + %4815 = "ttir.relu"(%4813, %4814) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4816 = tensor.empty() : tensor<1x64xf32> + %4817 = "ttir.relu"(%4815, %4816) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4818 = tensor.empty() : tensor<1x64xf32> + %4819 = "ttir.relu"(%4817, %4818) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4820 = tensor.empty() : tensor<1x64xf32> + %4821 = "ttir.relu"(%4819, %4820) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4822 = tensor.empty() : tensor<1x64xf32> + %4823 = "ttir.relu"(%4821, %4822) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4824 = tensor.empty() : tensor<1x64xf32> + %4825 = "ttir.relu"(%4823, %4824) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4826 = tensor.empty() : tensor<1x64xf32> + %4827 = "ttir.relu"(%4825, %4826) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4828 = tensor.empty() : tensor<1x64xf32> + %4829 = "ttir.relu"(%4827, %4828) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4830 = tensor.empty() : tensor<1x64xf32> + %4831 = "ttir.relu"(%4829, %4830) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4832 = tensor.empty() : tensor<1x64xf32> + %4833 = "ttir.relu"(%4831, %4832) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4834 = tensor.empty() : tensor<1x64xf32> + %4835 = "ttir.relu"(%4833, %4834) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4836 = tensor.empty() : tensor<1x64xf32> + %4837 = "ttir.relu"(%4835, %4836) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4838 = tensor.empty() : tensor<1x64xf32> + %4839 = "ttir.relu"(%4837, %4838) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4840 = tensor.empty() : tensor<1x64xf32> + %4841 = "ttir.relu"(%4839, %4840) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4842 = tensor.empty() : tensor<1x64xf32> + %4843 = "ttir.relu"(%4841, %4842) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4844 = tensor.empty() : tensor<1x64xf32> + %4845 = "ttir.relu"(%4843, %4844) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4846 = tensor.empty() : tensor<1x64xf32> + %4847 = "ttir.relu"(%4845, %4846) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4848 = tensor.empty() : tensor<1x64xf32> + %4849 = "ttir.relu"(%4847, %4848) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4850 = tensor.empty() : tensor<1x64xf32> + %4851 = "ttir.relu"(%4849, %4850) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4852 = tensor.empty() : tensor<1x64xf32> + %4853 = "ttir.relu"(%4851, %4852) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4854 = tensor.empty() : tensor<1x64xf32> + %4855 = "ttir.relu"(%4853, %4854) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4856 = tensor.empty() : tensor<1x64xf32> + %4857 = "ttir.relu"(%4855, %4856) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4858 = tensor.empty() : tensor<1x64xf32> + %4859 = "ttir.relu"(%4857, %4858) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4860 = tensor.empty() : tensor<1x64xf32> + %4861 = "ttir.relu"(%4859, %4860) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4862 = tensor.empty() : tensor<1x64xf32> + %4863 = "ttir.relu"(%4861, %4862) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4864 = tensor.empty() : tensor<1x64xf32> + %4865 = "ttir.relu"(%4863, %4864) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4866 = tensor.empty() : tensor<1x64xf32> + %4867 = "ttir.relu"(%4865, %4866) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4868 = tensor.empty() : tensor<1x64xf32> + %4869 = "ttir.relu"(%4867, %4868) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4870 = tensor.empty() : tensor<1x64xf32> + %4871 = "ttir.relu"(%4869, %4870) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4872 = tensor.empty() : tensor<1x64xf32> + %4873 = "ttir.relu"(%4871, %4872) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4874 = tensor.empty() : tensor<1x64xf32> + %4875 = "ttir.relu"(%4873, %4874) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4876 = tensor.empty() : tensor<1x64xf32> + %4877 = "ttir.relu"(%4875, %4876) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4878 = tensor.empty() : tensor<1x64xf32> + %4879 = "ttir.relu"(%4877, %4878) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4880 = tensor.empty() : tensor<1x64xf32> + %4881 = "ttir.relu"(%4879, %4880) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4882 = tensor.empty() : tensor<1x64xf32> + %4883 = "ttir.relu"(%4881, %4882) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4884 = tensor.empty() : tensor<1x64xf32> + %4885 = "ttir.relu"(%4883, %4884) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4886 = tensor.empty() : tensor<1x64xf32> + %4887 = "ttir.relu"(%4885, %4886) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4888 = tensor.empty() : tensor<1x64xf32> + %4889 = "ttir.relu"(%4887, %4888) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4890 = tensor.empty() : tensor<1x64xf32> + %4891 = "ttir.relu"(%4889, %4890) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4892 = tensor.empty() : tensor<1x64xf32> + %4893 = "ttir.relu"(%4891, %4892) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4894 = tensor.empty() : tensor<1x64xf32> + %4895 = "ttir.relu"(%4893, %4894) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4896 = tensor.empty() : tensor<1x64xf32> + %4897 = "ttir.relu"(%4895, %4896) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4898 = tensor.empty() : tensor<1x64xf32> + %4899 = "ttir.relu"(%4897, %4898) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4900 = tensor.empty() : tensor<1x64xf32> + %4901 = "ttir.relu"(%4899, %4900) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4902 = tensor.empty() : tensor<1x64xf32> + %4903 = "ttir.relu"(%4901, %4902) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4904 = tensor.empty() : tensor<1x64xf32> + %4905 = "ttir.relu"(%4903, %4904) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4906 = tensor.empty() : tensor<1x64xf32> + %4907 = "ttir.relu"(%4905, %4906) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4908 = tensor.empty() : tensor<1x64xf32> + %4909 = "ttir.relu"(%4907, %4908) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4910 = tensor.empty() : tensor<1x64xf32> + %4911 = "ttir.relu"(%4909, %4910) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4912 = tensor.empty() : tensor<1x64xf32> + %4913 = "ttir.relu"(%4911, %4912) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4914 = tensor.empty() : tensor<1x64xf32> + %4915 = "ttir.relu"(%4913, %4914) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4916 = tensor.empty() : tensor<1x64xf32> + %4917 = "ttir.relu"(%4915, %4916) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4918 = tensor.empty() : tensor<1x64xf32> + %4919 = "ttir.relu"(%4917, %4918) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4920 = tensor.empty() : tensor<1x64xf32> + %4921 = "ttir.relu"(%4919, %4920) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4922 = tensor.empty() : tensor<1x64xf32> + %4923 = "ttir.relu"(%4921, %4922) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4924 = tensor.empty() : tensor<1x64xf32> + %4925 = "ttir.relu"(%4923, %4924) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4926 = tensor.empty() : tensor<1x64xf32> + %4927 = "ttir.relu"(%4925, %4926) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4928 = tensor.empty() : tensor<1x64xf32> + %4929 = "ttir.relu"(%4927, %4928) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4930 = tensor.empty() : tensor<1x64xf32> + %4931 = "ttir.relu"(%4929, %4930) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4932 = tensor.empty() : tensor<1x64xf32> + %4933 = "ttir.relu"(%4931, %4932) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4934 = tensor.empty() : tensor<1x64xf32> + %4935 = "ttir.relu"(%4933, %4934) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4936 = tensor.empty() : tensor<1x64xf32> + %4937 = "ttir.relu"(%4935, %4936) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4938 = tensor.empty() : tensor<1x64xf32> + %4939 = "ttir.relu"(%4937, %4938) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4940 = tensor.empty() : tensor<1x64xf32> + %4941 = "ttir.relu"(%4939, %4940) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4942 = tensor.empty() : tensor<1x64xf32> + %4943 = "ttir.relu"(%4941, %4942) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4944 = tensor.empty() : tensor<1x64xf32> + %4945 = "ttir.relu"(%4943, %4944) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4946 = tensor.empty() : tensor<1x64xf32> + %4947 = "ttir.relu"(%4945, %4946) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4948 = tensor.empty() : tensor<1x64xf32> + %4949 = "ttir.relu"(%4947, %4948) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4950 = tensor.empty() : tensor<1x64xf32> + %4951 = "ttir.relu"(%4949, %4950) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4952 = tensor.empty() : tensor<1x64xf32> + %4953 = "ttir.relu"(%4951, %4952) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4954 = tensor.empty() : tensor<1x64xf32> + %4955 = "ttir.relu"(%4953, %4954) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4956 = tensor.empty() : tensor<1x64xf32> + %4957 = "ttir.relu"(%4955, %4956) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4958 = tensor.empty() : tensor<1x64xf32> + %4959 = "ttir.relu"(%4957, %4958) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4960 = tensor.empty() : tensor<1x64xf32> + %4961 = "ttir.relu"(%4959, %4960) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4962 = tensor.empty() : tensor<1x64xf32> + %4963 = "ttir.relu"(%4961, %4962) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4964 = tensor.empty() : tensor<1x64xf32> + %4965 = "ttir.relu"(%4963, %4964) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4966 = tensor.empty() : tensor<1x64xf32> + %4967 = "ttir.relu"(%4965, %4966) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4968 = tensor.empty() : tensor<1x64xf32> + %4969 = "ttir.relu"(%4967, %4968) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4970 = tensor.empty() : tensor<1x64xf32> + %4971 = "ttir.relu"(%4969, %4970) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4972 = tensor.empty() : tensor<1x64xf32> + %4973 = "ttir.relu"(%4971, %4972) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4974 = tensor.empty() : tensor<1x64xf32> + %4975 = "ttir.relu"(%4973, %4974) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4976 = tensor.empty() : tensor<1x64xf32> + %4977 = "ttir.relu"(%4975, %4976) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4978 = tensor.empty() : tensor<1x64xf32> + %4979 = "ttir.relu"(%4977, %4978) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4980 = tensor.empty() : tensor<1x64xf32> + %4981 = "ttir.relu"(%4979, %4980) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4982 = tensor.empty() : tensor<1x64xf32> + %4983 = "ttir.relu"(%4981, %4982) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4984 = tensor.empty() : tensor<1x64xf32> + %4985 = "ttir.relu"(%4983, %4984) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4986 = tensor.empty() : tensor<1x64xf32> + %4987 = "ttir.relu"(%4985, %4986) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4988 = tensor.empty() : tensor<1x64xf32> + %4989 = "ttir.relu"(%4987, %4988) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4990 = tensor.empty() : tensor<1x64xf32> + %4991 = "ttir.relu"(%4989, %4990) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4992 = tensor.empty() : tensor<1x64xf32> + %4993 = "ttir.relu"(%4991, %4992) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4994 = tensor.empty() : tensor<1x64xf32> + %4995 = "ttir.relu"(%4993, %4994) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4996 = tensor.empty() : tensor<1x64xf32> + %4997 = "ttir.relu"(%4995, %4996) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4998 = tensor.empty() : tensor<1x64xf32> + %4999 = "ttir.relu"(%4997, %4998) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5000 = tensor.empty() : tensor<1x64xf32> + %5001 = "ttir.relu"(%4999, %5000) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5002 = tensor.empty() : tensor<1x64xf32> + %5003 = "ttir.relu"(%5001, %5002) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5004 = tensor.empty() : tensor<1x64xf32> + %5005 = "ttir.relu"(%5003, %5004) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5006 = tensor.empty() : tensor<1x64xf32> + %5007 = "ttir.relu"(%5005, %5006) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5008 = tensor.empty() : tensor<1x64xf32> + %5009 = "ttir.relu"(%5007, %5008) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5010 = tensor.empty() : tensor<1x64xf32> + %5011 = "ttir.relu"(%5009, %5010) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5012 = tensor.empty() : tensor<1x64xf32> + %5013 = "ttir.relu"(%5011, %5012) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5014 = tensor.empty() : tensor<1x64xf32> + %5015 = "ttir.relu"(%5013, %5014) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5016 = tensor.empty() : tensor<1x64xf32> + %5017 = "ttir.relu"(%5015, %5016) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5018 = tensor.empty() : tensor<1x64xf32> + %5019 = "ttir.relu"(%5017, %5018) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5020 = tensor.empty() : tensor<1x64xf32> + %5021 = "ttir.relu"(%5019, %5020) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5022 = tensor.empty() : tensor<1x64xf32> + %5023 = "ttir.relu"(%5021, %5022) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5024 = tensor.empty() : tensor<1x64xf32> + %5025 = "ttir.relu"(%5023, %5024) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5026 = tensor.empty() : tensor<1x64xf32> + %5027 = "ttir.relu"(%5025, %5026) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5028 = tensor.empty() : tensor<1x64xf32> + %5029 = "ttir.relu"(%5027, %5028) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5030 = tensor.empty() : tensor<1x64xf32> + %5031 = "ttir.relu"(%5029, %5030) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5032 = tensor.empty() : tensor<1x64xf32> + %5033 = "ttir.relu"(%5031, %5032) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5034 = tensor.empty() : tensor<1x64xf32> + %5035 = "ttir.relu"(%5033, %5034) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5036 = tensor.empty() : tensor<1x64xf32> + %5037 = "ttir.relu"(%5035, %5036) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5038 = tensor.empty() : tensor<1x64xf32> + %5039 = "ttir.relu"(%5037, %5038) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5040 = tensor.empty() : tensor<1x64xf32> + %5041 = "ttir.relu"(%5039, %5040) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5042 = tensor.empty() : tensor<1x64xf32> + %5043 = "ttir.relu"(%5041, %5042) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5044 = tensor.empty() : tensor<1x64xf32> + %5045 = "ttir.relu"(%5043, %5044) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5046 = tensor.empty() : tensor<1x64xf32> + %5047 = "ttir.relu"(%5045, %5046) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5048 = tensor.empty() : tensor<1x64xf32> + %5049 = "ttir.relu"(%5047, %5048) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5050 = tensor.empty() : tensor<1x64xf32> + %5051 = "ttir.relu"(%5049, %5050) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5052 = tensor.empty() : tensor<1x64xf32> + %5053 = "ttir.relu"(%5051, %5052) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5054 = tensor.empty() : tensor<1x64xf32> + %5055 = "ttir.relu"(%5053, %5054) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5056 = tensor.empty() : tensor<1x64xf32> + %5057 = "ttir.relu"(%5055, %5056) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5058 = tensor.empty() : tensor<1x64xf32> + %5059 = "ttir.relu"(%5057, %5058) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5060 = tensor.empty() : tensor<1x64xf32> + %5061 = "ttir.relu"(%5059, %5060) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5062 = tensor.empty() : tensor<1x64xf32> + %5063 = "ttir.relu"(%5061, %5062) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5064 = tensor.empty() : tensor<1x64xf32> + %5065 = "ttir.relu"(%5063, %5064) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5066 = tensor.empty() : tensor<1x64xf32> + %5067 = "ttir.relu"(%5065, %5066) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5068 = tensor.empty() : tensor<1x64xf32> + %5069 = "ttir.relu"(%5067, %5068) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5070 = tensor.empty() : tensor<1x64xf32> + %5071 = "ttir.relu"(%5069, %5070) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5072 = tensor.empty() : tensor<1x64xf32> + %5073 = "ttir.relu"(%5071, %5072) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5074 = tensor.empty() : tensor<1x64xf32> + %5075 = "ttir.relu"(%5073, %5074) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5076 = tensor.empty() : tensor<1x64xf32> + %5077 = "ttir.relu"(%5075, %5076) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5078 = tensor.empty() : tensor<1x64xf32> + %5079 = "ttir.relu"(%5077, %5078) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5080 = tensor.empty() : tensor<1x64xf32> + %5081 = "ttir.relu"(%5079, %5080) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5082 = tensor.empty() : tensor<1x64xf32> + %5083 = "ttir.relu"(%5081, %5082) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5084 = tensor.empty() : tensor<1x64xf32> + %5085 = "ttir.relu"(%5083, %5084) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5086 = tensor.empty() : tensor<1x64xf32> + %5087 = "ttir.relu"(%5085, %5086) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5088 = tensor.empty() : tensor<1x64xf32> + %5089 = "ttir.relu"(%5087, %5088) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5090 = tensor.empty() : tensor<1x64xf32> + %5091 = "ttir.relu"(%5089, %5090) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5092 = tensor.empty() : tensor<1x64xf32> + %5093 = "ttir.relu"(%5091, %5092) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5094 = tensor.empty() : tensor<1x64xf32> + %5095 = "ttir.relu"(%5093, %5094) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5096 = tensor.empty() : tensor<1x64xf32> + %5097 = "ttir.relu"(%5095, %5096) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5098 = tensor.empty() : tensor<1x64xf32> + %5099 = "ttir.relu"(%5097, %5098) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5100 = tensor.empty() : tensor<1x64xf32> + %5101 = "ttir.relu"(%5099, %5100) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5102 = tensor.empty() : tensor<1x64xf32> + %5103 = "ttir.relu"(%5101, %5102) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5104 = tensor.empty() : tensor<1x64xf32> + %5105 = "ttir.relu"(%5103, %5104) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5106 = tensor.empty() : tensor<1x64xf32> + %5107 = "ttir.relu"(%5105, %5106) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5108 = tensor.empty() : tensor<1x64xf32> + %5109 = "ttir.relu"(%5107, %5108) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5110 = tensor.empty() : tensor<1x64xf32> + %5111 = "ttir.relu"(%5109, %5110) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5112 = tensor.empty() : tensor<1x64xf32> + %5113 = "ttir.relu"(%5111, %5112) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5114 = tensor.empty() : tensor<1x64xf32> + %5115 = "ttir.relu"(%5113, %5114) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5116 = tensor.empty() : tensor<1x64xf32> + %5117 = "ttir.relu"(%5115, %5116) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5118 = tensor.empty() : tensor<1x64xf32> + %5119 = "ttir.relu"(%5117, %5118) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5120 = tensor.empty() : tensor<1x64xf32> + %5121 = "ttir.relu"(%5119, %5120) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5122 = tensor.empty() : tensor<1x64xf32> + %5123 = "ttir.relu"(%5121, %5122) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5124 = tensor.empty() : tensor<1x64xf32> + %5125 = "ttir.relu"(%5123, %5124) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5126 = tensor.empty() : tensor<1x64xf32> + %5127 = "ttir.relu"(%5125, %5126) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5128 = tensor.empty() : tensor<1x64xf32> + %5129 = "ttir.relu"(%5127, %5128) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5130 = tensor.empty() : tensor<1x64xf32> + %5131 = "ttir.relu"(%5129, %5130) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5132 = tensor.empty() : tensor<1x64xf32> + %5133 = "ttir.relu"(%5131, %5132) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5134 = tensor.empty() : tensor<1x64xf32> + %5135 = "ttir.relu"(%5133, %5134) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5136 = tensor.empty() : tensor<1x64xf32> + %5137 = "ttir.relu"(%5135, %5136) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5138 = tensor.empty() : tensor<1x64xf32> + %5139 = "ttir.relu"(%5137, %5138) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5140 = tensor.empty() : tensor<1x64xf32> + %5141 = "ttir.relu"(%5139, %5140) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5142 = tensor.empty() : tensor<1x64xf32> + %5143 = "ttir.relu"(%5141, %5142) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5144 = tensor.empty() : tensor<1x64xf32> + %5145 = "ttir.relu"(%5143, %5144) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5146 = tensor.empty() : tensor<1x64xf32> + %5147 = "ttir.relu"(%5145, %5146) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5148 = tensor.empty() : tensor<1x64xf32> + %5149 = "ttir.relu"(%5147, %5148) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5150 = tensor.empty() : tensor<1x64xf32> + %5151 = "ttir.relu"(%5149, %5150) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5152 = tensor.empty() : tensor<1x64xf32> + %5153 = "ttir.relu"(%5151, %5152) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5154 = tensor.empty() : tensor<1x64xf32> + %5155 = "ttir.relu"(%5153, %5154) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5156 = tensor.empty() : tensor<1x64xf32> + %5157 = "ttir.relu"(%5155, %5156) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5158 = tensor.empty() : tensor<1x64xf32> + %5159 = "ttir.relu"(%5157, %5158) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5160 = tensor.empty() : tensor<1x64xf32> + %5161 = "ttir.relu"(%5159, %5160) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5162 = tensor.empty() : tensor<1x64xf32> + %5163 = "ttir.relu"(%5161, %5162) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5164 = tensor.empty() : tensor<1x64xf32> + %5165 = "ttir.relu"(%5163, %5164) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5166 = tensor.empty() : tensor<1x64xf32> + %5167 = "ttir.relu"(%5165, %5166) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5168 = tensor.empty() : tensor<1x64xf32> + %5169 = "ttir.relu"(%5167, %5168) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5170 = tensor.empty() : tensor<1x64xf32> + %5171 = "ttir.relu"(%5169, %5170) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5172 = tensor.empty() : tensor<1x64xf32> + %5173 = "ttir.relu"(%5171, %5172) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5174 = tensor.empty() : tensor<1x64xf32> + %5175 = "ttir.relu"(%5173, %5174) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5176 = tensor.empty() : tensor<1x64xf32> + %5177 = "ttir.relu"(%5175, %5176) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5178 = tensor.empty() : tensor<1x64xf32> + %5179 = "ttir.relu"(%5177, %5178) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5180 = tensor.empty() : tensor<1x64xf32> + %5181 = "ttir.relu"(%5179, %5180) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5182 = tensor.empty() : tensor<1x64xf32> + %5183 = "ttir.relu"(%5181, %5182) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5184 = tensor.empty() : tensor<1x64xf32> + %5185 = "ttir.relu"(%5183, %5184) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5186 = tensor.empty() : tensor<1x64xf32> + %5187 = "ttir.relu"(%5185, %5186) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5188 = tensor.empty() : tensor<1x64xf32> + %5189 = "ttir.relu"(%5187, %5188) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5190 = tensor.empty() : tensor<1x64xf32> + %5191 = "ttir.relu"(%5189, %5190) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5192 = tensor.empty() : tensor<1x64xf32> + %5193 = "ttir.relu"(%5191, %5192) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5194 = tensor.empty() : tensor<1x64xf32> + %5195 = "ttir.relu"(%5193, %5194) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5196 = tensor.empty() : tensor<1x64xf32> + %5197 = "ttir.relu"(%5195, %5196) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5198 = tensor.empty() : tensor<1x64xf32> + %5199 = "ttir.relu"(%5197, %5198) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5200 = tensor.empty() : tensor<1x64xf32> + %5201 = "ttir.relu"(%5199, %5200) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5202 = tensor.empty() : tensor<1x64xf32> + %5203 = "ttir.relu"(%5201, %5202) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5204 = tensor.empty() : tensor<1x64xf32> + %5205 = "ttir.relu"(%5203, %5204) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5206 = tensor.empty() : tensor<1x64xf32> + %5207 = "ttir.relu"(%5205, %5206) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5208 = tensor.empty() : tensor<1x64xf32> + %5209 = "ttir.relu"(%5207, %5208) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5210 = tensor.empty() : tensor<1x64xf32> + %5211 = "ttir.relu"(%5209, %5210) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5212 = tensor.empty() : tensor<1x64xf32> + %5213 = "ttir.relu"(%5211, %5212) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5214 = tensor.empty() : tensor<1x64xf32> + %5215 = "ttir.relu"(%5213, %5214) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5216 = tensor.empty() : tensor<1x64xf32> + %5217 = "ttir.relu"(%5215, %5216) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5218 = tensor.empty() : tensor<1x64xf32> + %5219 = "ttir.relu"(%5217, %5218) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5220 = tensor.empty() : tensor<1x64xf32> + %5221 = "ttir.relu"(%5219, %5220) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5222 = tensor.empty() : tensor<1x64xf32> + %5223 = "ttir.relu"(%5221, %5222) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5224 = tensor.empty() : tensor<1x64xf32> + %5225 = "ttir.relu"(%5223, %5224) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5226 = tensor.empty() : tensor<1x64xf32> + %5227 = "ttir.relu"(%5225, %5226) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5228 = tensor.empty() : tensor<1x64xf32> + %5229 = "ttir.relu"(%5227, %5228) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5230 = tensor.empty() : tensor<1x64xf32> + %5231 = "ttir.relu"(%5229, %5230) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5232 = tensor.empty() : tensor<1x64xf32> + %5233 = "ttir.relu"(%5231, %5232) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5234 = tensor.empty() : tensor<1x64xf32> + %5235 = "ttir.relu"(%5233, %5234) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5236 = tensor.empty() : tensor<1x64xf32> + %5237 = "ttir.relu"(%5235, %5236) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5238 = tensor.empty() : tensor<1x64xf32> + %5239 = "ttir.relu"(%5237, %5238) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5240 = tensor.empty() : tensor<1x64xf32> + %5241 = "ttir.relu"(%5239, %5240) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5242 = tensor.empty() : tensor<1x64xf32> + %5243 = "ttir.relu"(%5241, %5242) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5244 = tensor.empty() : tensor<1x64xf32> + %5245 = "ttir.relu"(%5243, %5244) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5246 = tensor.empty() : tensor<1x64xf32> + %5247 = "ttir.relu"(%5245, %5246) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5248 = tensor.empty() : tensor<1x64xf32> + %5249 = "ttir.relu"(%5247, %5248) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5250 = tensor.empty() : tensor<1x64xf32> + %5251 = "ttir.relu"(%5249, %5250) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5252 = tensor.empty() : tensor<1x64xf32> + %5253 = "ttir.relu"(%5251, %5252) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5254 = tensor.empty() : tensor<1x64xf32> + %5255 = "ttir.relu"(%5253, %5254) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5256 = tensor.empty() : tensor<1x64xf32> + %5257 = "ttir.relu"(%5255, %5256) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5258 = tensor.empty() : tensor<1x64xf32> + %5259 = "ttir.relu"(%5257, %5258) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5260 = tensor.empty() : tensor<1x64xf32> + %5261 = "ttir.relu"(%5259, %5260) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5262 = tensor.empty() : tensor<1x64xf32> + %5263 = "ttir.relu"(%5261, %5262) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5264 = tensor.empty() : tensor<1x64xf32> + %5265 = "ttir.relu"(%5263, %5264) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5266 = tensor.empty() : tensor<1x64xf32> + %5267 = "ttir.relu"(%5265, %5266) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5268 = tensor.empty() : tensor<1x64xf32> + %5269 = "ttir.relu"(%5267, %5268) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5270 = tensor.empty() : tensor<1x64xf32> + %5271 = "ttir.relu"(%5269, %5270) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5272 = tensor.empty() : tensor<1x64xf32> + %5273 = "ttir.relu"(%5271, %5272) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5274 = tensor.empty() : tensor<1x64xf32> + %5275 = "ttir.relu"(%5273, %5274) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5276 = tensor.empty() : tensor<1x64xf32> + %5277 = "ttir.relu"(%5275, %5276) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5278 = tensor.empty() : tensor<1x64xf32> + %5279 = "ttir.relu"(%5277, %5278) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5280 = tensor.empty() : tensor<1x64xf32> + %5281 = "ttir.relu"(%5279, %5280) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5282 = tensor.empty() : tensor<1x64xf32> + %5283 = "ttir.relu"(%5281, %5282) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5284 = tensor.empty() : tensor<1x64xf32> + %5285 = "ttir.relu"(%5283, %5284) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5286 = tensor.empty() : tensor<1x64xf32> + %5287 = "ttir.relu"(%5285, %5286) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5288 = tensor.empty() : tensor<1x64xf32> + %5289 = "ttir.relu"(%5287, %5288) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5290 = tensor.empty() : tensor<1x64xf32> + %5291 = "ttir.relu"(%5289, %5290) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5292 = tensor.empty() : tensor<1x64xf32> + %5293 = "ttir.relu"(%5291, %5292) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5294 = tensor.empty() : tensor<1x64xf32> + %5295 = "ttir.relu"(%5293, %5294) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5296 = tensor.empty() : tensor<1x64xf32> + %5297 = "ttir.relu"(%5295, %5296) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5298 = tensor.empty() : tensor<1x64xf32> + %5299 = "ttir.relu"(%5297, %5298) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5300 = tensor.empty() : tensor<1x64xf32> + %5301 = "ttir.relu"(%5299, %5300) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5302 = tensor.empty() : tensor<1x64xf32> + %5303 = "ttir.relu"(%5301, %5302) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5304 = tensor.empty() : tensor<1x64xf32> + %5305 = "ttir.relu"(%5303, %5304) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5306 = tensor.empty() : tensor<1x64xf32> + %5307 = "ttir.relu"(%5305, %5306) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5308 = tensor.empty() : tensor<1x64xf32> + %5309 = "ttir.relu"(%5307, %5308) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5310 = tensor.empty() : tensor<1x64xf32> + %5311 = "ttir.relu"(%5309, %5310) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5312 = tensor.empty() : tensor<1x64xf32> + %5313 = "ttir.relu"(%5311, %5312) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5314 = tensor.empty() : tensor<1x64xf32> + %5315 = "ttir.relu"(%5313, %5314) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5316 = tensor.empty() : tensor<1x64xf32> + %5317 = "ttir.relu"(%5315, %5316) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5318 = tensor.empty() : tensor<1x64xf32> + %5319 = "ttir.relu"(%5317, %5318) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5320 = tensor.empty() : tensor<1x64xf32> + %5321 = "ttir.relu"(%5319, %5320) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5322 = tensor.empty() : tensor<1x64xf32> + %5323 = "ttir.relu"(%5321, %5322) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5324 = tensor.empty() : tensor<1x64xf32> + %5325 = "ttir.relu"(%5323, %5324) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5326 = tensor.empty() : tensor<1x64xf32> + %5327 = "ttir.relu"(%5325, %5326) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5328 = tensor.empty() : tensor<1x64xf32> + %5329 = "ttir.relu"(%5327, %5328) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5330 = tensor.empty() : tensor<1x64xf32> + %5331 = "ttir.relu"(%5329, %5330) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5332 = tensor.empty() : tensor<1x64xf32> + %5333 = "ttir.relu"(%5331, %5332) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5334 = tensor.empty() : tensor<1x64xf32> + %5335 = "ttir.relu"(%5333, %5334) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5336 = tensor.empty() : tensor<1x64xf32> + %5337 = "ttir.relu"(%5335, %5336) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5338 = tensor.empty() : tensor<1x64xf32> + %5339 = "ttir.relu"(%5337, %5338) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5340 = tensor.empty() : tensor<1x64xf32> + %5341 = "ttir.relu"(%5339, %5340) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5342 = tensor.empty() : tensor<1x64xf32> + %5343 = "ttir.relu"(%5341, %5342) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5344 = tensor.empty() : tensor<1x64xf32> + %5345 = "ttir.relu"(%5343, %5344) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5346 = tensor.empty() : tensor<1x64xf32> + %5347 = "ttir.relu"(%5345, %5346) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5348 = tensor.empty() : tensor<1x64xf32> + %5349 = "ttir.relu"(%5347, %5348) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5350 = tensor.empty() : tensor<1x64xf32> + %5351 = "ttir.relu"(%5349, %5350) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5352 = tensor.empty() : tensor<1x64xf32> + %5353 = "ttir.relu"(%5351, %5352) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5354 = tensor.empty() : tensor<1x64xf32> + %5355 = "ttir.relu"(%5353, %5354) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5356 = tensor.empty() : tensor<1x64xf32> + %5357 = "ttir.relu"(%5355, %5356) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5358 = tensor.empty() : tensor<1x64xf32> + %5359 = "ttir.relu"(%5357, %5358) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5360 = tensor.empty() : tensor<1x64xf32> + %5361 = "ttir.relu"(%5359, %5360) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5362 = tensor.empty() : tensor<1x64xf32> + %5363 = "ttir.relu"(%5361, %5362) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5364 = tensor.empty() : tensor<1x64xf32> + %5365 = "ttir.relu"(%5363, %5364) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5366 = tensor.empty() : tensor<1x64xf32> + %5367 = "ttir.relu"(%5365, %5366) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5368 = tensor.empty() : tensor<1x64xf32> + %5369 = "ttir.relu"(%5367, %5368) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5370 = tensor.empty() : tensor<1x64xf32> + %5371 = "ttir.relu"(%5369, %5370) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5372 = tensor.empty() : tensor<1x64xf32> + %5373 = "ttir.relu"(%5371, %5372) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5374 = tensor.empty() : tensor<1x64xf32> + %5375 = "ttir.relu"(%5373, %5374) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5376 = tensor.empty() : tensor<1x64xf32> + %5377 = "ttir.relu"(%5375, %5376) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5378 = tensor.empty() : tensor<1x64xf32> + %5379 = "ttir.relu"(%5377, %5378) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5380 = tensor.empty() : tensor<1x64xf32> + %5381 = "ttir.relu"(%5379, %5380) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5382 = tensor.empty() : tensor<1x64xf32> + %5383 = "ttir.relu"(%5381, %5382) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5384 = tensor.empty() : tensor<1x64xf32> + %5385 = "ttir.relu"(%5383, %5384) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5386 = tensor.empty() : tensor<1x64xf32> + %5387 = "ttir.relu"(%5385, %5386) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5388 = tensor.empty() : tensor<1x64xf32> + %5389 = "ttir.relu"(%5387, %5388) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5390 = tensor.empty() : tensor<1x64xf32> + %5391 = "ttir.relu"(%5389, %5390) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5392 = tensor.empty() : tensor<1x64xf32> + %5393 = "ttir.relu"(%5391, %5392) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5394 = tensor.empty() : tensor<1x64xf32> + %5395 = "ttir.relu"(%5393, %5394) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5396 = tensor.empty() : tensor<1x64xf32> + %5397 = "ttir.relu"(%5395, %5396) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5398 = tensor.empty() : tensor<1x64xf32> + %5399 = "ttir.relu"(%5397, %5398) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5400 = tensor.empty() : tensor<1x64xf32> + %5401 = "ttir.relu"(%5399, %5400) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5402 = tensor.empty() : tensor<1x64xf32> + %5403 = "ttir.relu"(%5401, %5402) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5404 = tensor.empty() : tensor<1x64xf32> + %5405 = "ttir.relu"(%5403, %5404) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5406 = tensor.empty() : tensor<1x64xf32> + %5407 = "ttir.relu"(%5405, %5406) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5408 = tensor.empty() : tensor<1x64xf32> + %5409 = "ttir.relu"(%5407, %5408) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5410 = tensor.empty() : tensor<1x64xf32> + %5411 = "ttir.relu"(%5409, %5410) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5412 = tensor.empty() : tensor<1x64xf32> + %5413 = "ttir.relu"(%5411, %5412) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5414 = tensor.empty() : tensor<1x64xf32> + %5415 = "ttir.relu"(%5413, %5414) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5416 = tensor.empty() : tensor<1x64xf32> + %5417 = "ttir.relu"(%5415, %5416) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5418 = tensor.empty() : tensor<1x64xf32> + %5419 = "ttir.relu"(%5417, %5418) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5420 = tensor.empty() : tensor<1x64xf32> + %5421 = "ttir.relu"(%5419, %5420) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5422 = tensor.empty() : tensor<1x64xf32> + %5423 = "ttir.relu"(%5421, %5422) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5424 = tensor.empty() : tensor<1x64xf32> + %5425 = "ttir.relu"(%5423, %5424) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5426 = tensor.empty() : tensor<1x64xf32> + %5427 = "ttir.relu"(%5425, %5426) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5428 = tensor.empty() : tensor<1x64xf32> + %5429 = "ttir.relu"(%5427, %5428) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5430 = tensor.empty() : tensor<1x64xf32> + %5431 = "ttir.relu"(%5429, %5430) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5432 = tensor.empty() : tensor<1x64xf32> + %5433 = "ttir.relu"(%5431, %5432) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5434 = tensor.empty() : tensor<1x64xf32> + %5435 = "ttir.relu"(%5433, %5434) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5436 = tensor.empty() : tensor<1x64xf32> + %5437 = "ttir.relu"(%5435, %5436) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5438 = tensor.empty() : tensor<1x64xf32> + %5439 = "ttir.relu"(%5437, %5438) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5440 = tensor.empty() : tensor<1x64xf32> + %5441 = "ttir.relu"(%5439, %5440) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5442 = tensor.empty() : tensor<1x64xf32> + %5443 = "ttir.relu"(%5441, %5442) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5444 = tensor.empty() : tensor<1x64xf32> + %5445 = "ttir.relu"(%5443, %5444) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5446 = tensor.empty() : tensor<1x64xf32> + %5447 = "ttir.relu"(%5445, %5446) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5448 = tensor.empty() : tensor<1x64xf32> + %5449 = "ttir.relu"(%5447, %5448) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5450 = tensor.empty() : tensor<1x64xf32> + %5451 = "ttir.relu"(%5449, %5450) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5452 = tensor.empty() : tensor<1x64xf32> + %5453 = "ttir.relu"(%5451, %5452) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5454 = tensor.empty() : tensor<1x64xf32> + %5455 = "ttir.relu"(%5453, %5454) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5456 = tensor.empty() : tensor<1x64xf32> + %5457 = "ttir.relu"(%5455, %5456) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5458 = tensor.empty() : tensor<1x64xf32> + %5459 = "ttir.relu"(%5457, %5458) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5460 = tensor.empty() : tensor<1x64xf32> + %5461 = "ttir.relu"(%5459, %5460) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5462 = tensor.empty() : tensor<1x64xf32> + %5463 = "ttir.relu"(%5461, %5462) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5464 = tensor.empty() : tensor<1x64xf32> + %5465 = "ttir.relu"(%5463, %5464) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5466 = tensor.empty() : tensor<1x64xf32> + %5467 = "ttir.relu"(%5465, %5466) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5468 = tensor.empty() : tensor<1x64xf32> + %5469 = "ttir.relu"(%5467, %5468) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5470 = tensor.empty() : tensor<1x64xf32> + %5471 = "ttir.relu"(%5469, %5470) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5472 = tensor.empty() : tensor<1x64xf32> + %5473 = "ttir.relu"(%5471, %5472) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5474 = tensor.empty() : tensor<1x64xf32> + %5475 = "ttir.relu"(%5473, %5474) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5476 = tensor.empty() : tensor<1x64xf32> + %5477 = "ttir.relu"(%5475, %5476) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5478 = tensor.empty() : tensor<1x64xf32> + %5479 = "ttir.relu"(%5477, %5478) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5480 = tensor.empty() : tensor<1x64xf32> + %5481 = "ttir.relu"(%5479, %5480) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5482 = tensor.empty() : tensor<1x64xf32> + %5483 = "ttir.relu"(%5481, %5482) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5484 = tensor.empty() : tensor<1x64xf32> + %5485 = "ttir.relu"(%5483, %5484) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5486 = tensor.empty() : tensor<1x64xf32> + %5487 = "ttir.relu"(%5485, %5486) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5488 = tensor.empty() : tensor<1x64xf32> + %5489 = "ttir.relu"(%5487, %5488) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5490 = tensor.empty() : tensor<1x64xf32> + %5491 = "ttir.relu"(%5489, %5490) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5492 = tensor.empty() : tensor<1x64xf32> + %5493 = "ttir.relu"(%5491, %5492) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5494 = tensor.empty() : tensor<1x64xf32> + %5495 = "ttir.relu"(%5493, %5494) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5496 = tensor.empty() : tensor<1x64xf32> + %5497 = "ttir.relu"(%5495, %5496) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5498 = tensor.empty() : tensor<1x64xf32> + %5499 = "ttir.relu"(%5497, %5498) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5500 = tensor.empty() : tensor<1x64xf32> + %5501 = "ttir.relu"(%5499, %5500) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5502 = tensor.empty() : tensor<1x64xf32> + %5503 = "ttir.relu"(%5501, %5502) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5504 = tensor.empty() : tensor<1x64xf32> + %5505 = "ttir.relu"(%5503, %5504) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5506 = tensor.empty() : tensor<1x64xf32> + %5507 = "ttir.relu"(%5505, %5506) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5508 = tensor.empty() : tensor<1x64xf32> + %5509 = "ttir.relu"(%5507, %5508) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5510 = tensor.empty() : tensor<1x64xf32> + %5511 = "ttir.relu"(%5509, %5510) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5512 = tensor.empty() : tensor<1x64xf32> + %5513 = "ttir.relu"(%5511, %5512) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5514 = tensor.empty() : tensor<1x64xf32> + %5515 = "ttir.relu"(%5513, %5514) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5516 = tensor.empty() : tensor<1x64xf32> + %5517 = "ttir.relu"(%5515, %5516) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5518 = tensor.empty() : tensor<1x64xf32> + %5519 = "ttir.relu"(%5517, %5518) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5520 = tensor.empty() : tensor<1x64xf32> + %5521 = "ttir.relu"(%5519, %5520) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5522 = tensor.empty() : tensor<1x64xf32> + %5523 = "ttir.relu"(%5521, %5522) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5524 = tensor.empty() : tensor<1x64xf32> + %5525 = "ttir.relu"(%5523, %5524) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5526 = tensor.empty() : tensor<1x64xf32> + %5527 = "ttir.relu"(%5525, %5526) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5528 = tensor.empty() : tensor<1x64xf32> + %5529 = "ttir.relu"(%5527, %5528) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5530 = tensor.empty() : tensor<1x64xf32> + %5531 = "ttir.relu"(%5529, %5530) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5532 = tensor.empty() : tensor<1x64xf32> + %5533 = "ttir.relu"(%5531, %5532) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5534 = tensor.empty() : tensor<1x64xf32> + %5535 = "ttir.relu"(%5533, %5534) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5536 = tensor.empty() : tensor<1x64xf32> + %5537 = "ttir.relu"(%5535, %5536) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5538 = tensor.empty() : tensor<1x64xf32> + %5539 = "ttir.relu"(%5537, %5538) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5540 = tensor.empty() : tensor<1x64xf32> + %5541 = "ttir.relu"(%5539, %5540) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5542 = tensor.empty() : tensor<1x64xf32> + %5543 = "ttir.relu"(%5541, %5542) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5544 = tensor.empty() : tensor<1x64xf32> + %5545 = "ttir.relu"(%5543, %5544) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5546 = tensor.empty() : tensor<1x64xf32> + %5547 = "ttir.relu"(%5545, %5546) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5548 = tensor.empty() : tensor<1x64xf32> + %5549 = "ttir.relu"(%5547, %5548) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5550 = tensor.empty() : tensor<1x64xf32> + %5551 = "ttir.relu"(%5549, %5550) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5552 = tensor.empty() : tensor<1x64xf32> + %5553 = "ttir.relu"(%5551, %5552) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5554 = tensor.empty() : tensor<1x64xf32> + %5555 = "ttir.relu"(%5553, %5554) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5556 = tensor.empty() : tensor<1x64xf32> + %5557 = "ttir.relu"(%5555, %5556) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5558 = tensor.empty() : tensor<1x64xf32> + %5559 = "ttir.relu"(%5557, %5558) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5560 = tensor.empty() : tensor<1x64xf32> + %5561 = "ttir.relu"(%5559, %5560) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5562 = tensor.empty() : tensor<1x64xf32> + %5563 = "ttir.relu"(%5561, %5562) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5564 = tensor.empty() : tensor<1x64xf32> + %5565 = "ttir.relu"(%5563, %5564) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5566 = tensor.empty() : tensor<1x64xf32> + %5567 = "ttir.relu"(%5565, %5566) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5568 = tensor.empty() : tensor<1x64xf32> + %5569 = "ttir.relu"(%5567, %5568) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5570 = tensor.empty() : tensor<1x64xf32> + %5571 = "ttir.relu"(%5569, %5570) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5572 = tensor.empty() : tensor<1x64xf32> + %5573 = "ttir.relu"(%5571, %5572) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5574 = tensor.empty() : tensor<1x64xf32> + %5575 = "ttir.relu"(%5573, %5574) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5576 = tensor.empty() : tensor<1x64xf32> + %5577 = "ttir.relu"(%5575, %5576) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5578 = tensor.empty() : tensor<1x64xf32> + %5579 = "ttir.relu"(%5577, %5578) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5580 = tensor.empty() : tensor<1x64xf32> + %5581 = "ttir.relu"(%5579, %5580) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5582 = tensor.empty() : tensor<1x64xf32> + %5583 = "ttir.relu"(%5581, %5582) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5584 = tensor.empty() : tensor<1x64xf32> + %5585 = "ttir.relu"(%5583, %5584) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5586 = tensor.empty() : tensor<1x64xf32> + %5587 = "ttir.relu"(%5585, %5586) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5588 = tensor.empty() : tensor<1x64xf32> + %5589 = "ttir.relu"(%5587, %5588) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5590 = tensor.empty() : tensor<1x64xf32> + %5591 = "ttir.relu"(%5589, %5590) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5592 = tensor.empty() : tensor<1x64xf32> + %5593 = "ttir.relu"(%5591, %5592) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5594 = tensor.empty() : tensor<1x64xf32> + %5595 = "ttir.relu"(%5593, %5594) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5596 = tensor.empty() : tensor<1x64xf32> + %5597 = "ttir.relu"(%5595, %5596) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5598 = tensor.empty() : tensor<1x64xf32> + %5599 = "ttir.relu"(%5597, %5598) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5600 = tensor.empty() : tensor<1x64xf32> + %5601 = "ttir.relu"(%5599, %5600) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5602 = tensor.empty() : tensor<1x64xf32> + %5603 = "ttir.relu"(%5601, %5602) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5604 = tensor.empty() : tensor<1x64xf32> + %5605 = "ttir.relu"(%5603, %5604) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5606 = tensor.empty() : tensor<1x64xf32> + %5607 = "ttir.relu"(%5605, %5606) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5608 = tensor.empty() : tensor<1x64xf32> + %5609 = "ttir.relu"(%5607, %5608) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5610 = tensor.empty() : tensor<1x64xf32> + %5611 = "ttir.relu"(%5609, %5610) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5612 = tensor.empty() : tensor<1x64xf32> + %5613 = "ttir.relu"(%5611, %5612) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5614 = tensor.empty() : tensor<1x64xf32> + %5615 = "ttir.relu"(%5613, %5614) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5616 = tensor.empty() : tensor<1x64xf32> + %5617 = "ttir.relu"(%5615, %5616) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5618 = tensor.empty() : tensor<1x64xf32> + %5619 = "ttir.relu"(%5617, %5618) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5620 = tensor.empty() : tensor<1x64xf32> + %5621 = "ttir.relu"(%5619, %5620) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5622 = tensor.empty() : tensor<1x64xf32> + %5623 = "ttir.relu"(%5621, %5622) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5624 = tensor.empty() : tensor<1x64xf32> + %5625 = "ttir.relu"(%5623, %5624) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5626 = tensor.empty() : tensor<1x64xf32> + %5627 = "ttir.relu"(%5625, %5626) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5628 = tensor.empty() : tensor<1x64xf32> + %5629 = "ttir.relu"(%5627, %5628) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5630 = tensor.empty() : tensor<1x64xf32> + %5631 = "ttir.relu"(%5629, %5630) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5632 = tensor.empty() : tensor<1x64xf32> + %5633 = "ttir.relu"(%5631, %5632) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5634 = tensor.empty() : tensor<1x64xf32> + %5635 = "ttir.relu"(%5633, %5634) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5636 = tensor.empty() : tensor<1x64xf32> + %5637 = "ttir.relu"(%5635, %5636) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5638 = tensor.empty() : tensor<1x64xf32> + %5639 = "ttir.relu"(%5637, %5638) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5640 = tensor.empty() : tensor<1x64xf32> + %5641 = "ttir.relu"(%5639, %5640) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5642 = tensor.empty() : tensor<1x64xf32> + %5643 = "ttir.relu"(%5641, %5642) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5644 = tensor.empty() : tensor<1x64xf32> + %5645 = "ttir.relu"(%5643, %5644) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5646 = tensor.empty() : tensor<1x64xf32> + %5647 = "ttir.relu"(%5645, %5646) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5648 = tensor.empty() : tensor<1x64xf32> + %5649 = "ttir.relu"(%5647, %5648) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5650 = tensor.empty() : tensor<1x64xf32> + %5651 = "ttir.relu"(%5649, %5650) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5652 = tensor.empty() : tensor<1x64xf32> + %5653 = "ttir.relu"(%5651, %5652) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5654 = tensor.empty() : tensor<1x64xf32> + %5655 = "ttir.relu"(%5653, %5654) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5656 = tensor.empty() : tensor<1x64xf32> + %5657 = "ttir.relu"(%5655, %5656) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5658 = tensor.empty() : tensor<1x64xf32> + %5659 = "ttir.relu"(%5657, %5658) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5660 = tensor.empty() : tensor<1x64xf32> + %5661 = "ttir.relu"(%5659, %5660) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5662 = tensor.empty() : tensor<1x64xf32> + %5663 = "ttir.relu"(%5661, %5662) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5664 = tensor.empty() : tensor<1x64xf32> + %5665 = "ttir.relu"(%5663, %5664) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5666 = tensor.empty() : tensor<1x64xf32> + %5667 = "ttir.relu"(%5665, %5666) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5668 = tensor.empty() : tensor<1x64xf32> + %5669 = "ttir.relu"(%5667, %5668) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5670 = tensor.empty() : tensor<1x64xf32> + %5671 = "ttir.relu"(%5669, %5670) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5672 = tensor.empty() : tensor<1x64xf32> + %5673 = "ttir.relu"(%5671, %5672) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5674 = tensor.empty() : tensor<1x64xf32> + %5675 = "ttir.relu"(%5673, %5674) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5676 = tensor.empty() : tensor<1x64xf32> + %5677 = "ttir.relu"(%5675, %5676) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5678 = tensor.empty() : tensor<1x64xf32> + %5679 = "ttir.relu"(%5677, %5678) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5680 = tensor.empty() : tensor<1x64xf32> + %5681 = "ttir.relu"(%5679, %5680) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5682 = tensor.empty() : tensor<1x64xf32> + %5683 = "ttir.relu"(%5681, %5682) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5684 = tensor.empty() : tensor<1x64xf32> + %5685 = "ttir.relu"(%5683, %5684) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5686 = tensor.empty() : tensor<1x64xf32> + %5687 = "ttir.relu"(%5685, %5686) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5688 = tensor.empty() : tensor<1x64xf32> + %5689 = "ttir.relu"(%5687, %5688) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5690 = tensor.empty() : tensor<1x64xf32> + %5691 = "ttir.relu"(%5689, %5690) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5692 = tensor.empty() : tensor<1x64xf32> + %5693 = "ttir.relu"(%5691, %5692) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5694 = tensor.empty() : tensor<1x64xf32> + %5695 = "ttir.relu"(%5693, %5694) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5696 = tensor.empty() : tensor<1x64xf32> + %5697 = "ttir.relu"(%5695, %5696) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5698 = tensor.empty() : tensor<1x64xf32> + %5699 = "ttir.relu"(%5697, %5698) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5700 = tensor.empty() : tensor<1x64xf32> + %5701 = "ttir.relu"(%5699, %5700) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5702 = tensor.empty() : tensor<1x64xf32> + %5703 = "ttir.relu"(%5701, %5702) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5704 = tensor.empty() : tensor<1x64xf32> + %5705 = "ttir.relu"(%5703, %5704) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5706 = tensor.empty() : tensor<1x64xf32> + %5707 = "ttir.relu"(%5705, %5706) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5708 = tensor.empty() : tensor<1x64xf32> + %5709 = "ttir.relu"(%5707, %5708) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5710 = tensor.empty() : tensor<1x64xf32> + %5711 = "ttir.relu"(%5709, %5710) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5712 = tensor.empty() : tensor<1x64xf32> + %5713 = "ttir.relu"(%5711, %5712) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5714 = tensor.empty() : tensor<1x64xf32> + %5715 = "ttir.relu"(%5713, %5714) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5716 = tensor.empty() : tensor<1x64xf32> + %5717 = "ttir.relu"(%5715, %5716) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5718 = tensor.empty() : tensor<1x64xf32> + %5719 = "ttir.relu"(%5717, %5718) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5720 = tensor.empty() : tensor<1x64xf32> + %5721 = "ttir.relu"(%5719, %5720) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5722 = tensor.empty() : tensor<1x64xf32> + %5723 = "ttir.relu"(%5721, %5722) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5724 = tensor.empty() : tensor<1x64xf32> + %5725 = "ttir.relu"(%5723, %5724) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5726 = tensor.empty() : tensor<1x64xf32> + %5727 = "ttir.relu"(%5725, %5726) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5728 = tensor.empty() : tensor<1x64xf32> + %5729 = "ttir.relu"(%5727, %5728) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5730 = tensor.empty() : tensor<1x64xf32> + %5731 = "ttir.relu"(%5729, %5730) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5732 = tensor.empty() : tensor<1x64xf32> + %5733 = "ttir.relu"(%5731, %5732) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5734 = tensor.empty() : tensor<1x64xf32> + %5735 = "ttir.relu"(%5733, %5734) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5736 = tensor.empty() : tensor<1x64xf32> + %5737 = "ttir.relu"(%5735, %5736) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5738 = tensor.empty() : tensor<1x64xf32> + %5739 = "ttir.relu"(%5737, %5738) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5740 = tensor.empty() : tensor<1x64xf32> + %5741 = "ttir.relu"(%5739, %5740) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5742 = tensor.empty() : tensor<1x64xf32> + %5743 = "ttir.relu"(%5741, %5742) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5744 = tensor.empty() : tensor<1x64xf32> + %5745 = "ttir.relu"(%5743, %5744) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5746 = tensor.empty() : tensor<1x64xf32> + %5747 = "ttir.relu"(%5745, %5746) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5748 = tensor.empty() : tensor<1x64xf32> + %5749 = "ttir.relu"(%5747, %5748) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5750 = tensor.empty() : tensor<1x64xf32> + %5751 = "ttir.relu"(%5749, %5750) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5752 = tensor.empty() : tensor<1x64xf32> + %5753 = "ttir.relu"(%5751, %5752) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5754 = tensor.empty() : tensor<1x64xf32> + %5755 = "ttir.relu"(%5753, %5754) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5756 = tensor.empty() : tensor<1x64xf32> + %5757 = "ttir.relu"(%5755, %5756) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5758 = tensor.empty() : tensor<1x64xf32> + %5759 = "ttir.relu"(%5757, %5758) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5760 = tensor.empty() : tensor<1x64xf32> + %5761 = "ttir.relu"(%5759, %5760) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5762 = tensor.empty() : tensor<1x64xf32> + %5763 = "ttir.relu"(%5761, %5762) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5764 = tensor.empty() : tensor<1x64xf32> + %5765 = "ttir.relu"(%5763, %5764) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5766 = tensor.empty() : tensor<1x64xf32> + %5767 = "ttir.relu"(%5765, %5766) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5768 = tensor.empty() : tensor<1x64xf32> + %5769 = "ttir.relu"(%5767, %5768) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5770 = tensor.empty() : tensor<1x64xf32> + %5771 = "ttir.relu"(%5769, %5770) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5772 = tensor.empty() : tensor<1x64xf32> + %5773 = "ttir.relu"(%5771, %5772) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5774 = tensor.empty() : tensor<1x64xf32> + %5775 = "ttir.relu"(%5773, %5774) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5776 = tensor.empty() : tensor<1x64xf32> + %5777 = "ttir.relu"(%5775, %5776) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5778 = tensor.empty() : tensor<1x64xf32> + %5779 = "ttir.relu"(%5777, %5778) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5780 = tensor.empty() : tensor<1x64xf32> + %5781 = "ttir.relu"(%5779, %5780) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5782 = tensor.empty() : tensor<1x64xf32> + %5783 = "ttir.relu"(%5781, %5782) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5784 = tensor.empty() : tensor<1x64xf32> + %5785 = "ttir.relu"(%5783, %5784) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5786 = tensor.empty() : tensor<1x64xf32> + %5787 = "ttir.relu"(%5785, %5786) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5788 = tensor.empty() : tensor<1x64xf32> + %5789 = "ttir.relu"(%5787, %5788) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5790 = tensor.empty() : tensor<1x64xf32> + %5791 = "ttir.relu"(%5789, %5790) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5792 = tensor.empty() : tensor<1x64xf32> + %5793 = "ttir.relu"(%5791, %5792) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5794 = tensor.empty() : tensor<1x64xf32> + %5795 = "ttir.relu"(%5793, %5794) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5796 = tensor.empty() : tensor<1x64xf32> + %5797 = "ttir.relu"(%5795, %5796) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5798 = tensor.empty() : tensor<1x64xf32> + %5799 = "ttir.relu"(%5797, %5798) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5800 = tensor.empty() : tensor<1x64xf32> + %5801 = "ttir.relu"(%5799, %5800) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5802 = tensor.empty() : tensor<1x64xf32> + %5803 = "ttir.relu"(%5801, %5802) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5804 = tensor.empty() : tensor<1x64xf32> + %5805 = "ttir.relu"(%5803, %5804) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5806 = tensor.empty() : tensor<1x64xf32> + %5807 = "ttir.relu"(%5805, %5806) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5808 = tensor.empty() : tensor<1x64xf32> + %5809 = "ttir.relu"(%5807, %5808) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5810 = tensor.empty() : tensor<1x64xf32> + %5811 = "ttir.relu"(%5809, %5810) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5812 = tensor.empty() : tensor<1x64xf32> + %5813 = "ttir.relu"(%5811, %5812) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5814 = tensor.empty() : tensor<1x64xf32> + %5815 = "ttir.relu"(%5813, %5814) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5816 = tensor.empty() : tensor<1x64xf32> + %5817 = "ttir.relu"(%5815, %5816) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5818 = tensor.empty() : tensor<1x64xf32> + %5819 = "ttir.relu"(%5817, %5818) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5820 = tensor.empty() : tensor<1x64xf32> + %5821 = "ttir.relu"(%5819, %5820) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5822 = tensor.empty() : tensor<1x64xf32> + %5823 = "ttir.relu"(%5821, %5822) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5824 = tensor.empty() : tensor<1x64xf32> + %5825 = "ttir.relu"(%5823, %5824) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5826 = tensor.empty() : tensor<1x64xf32> + %5827 = "ttir.relu"(%5825, %5826) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5828 = tensor.empty() : tensor<1x64xf32> + %5829 = "ttir.relu"(%5827, %5828) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5830 = tensor.empty() : tensor<1x64xf32> + %5831 = "ttir.relu"(%5829, %5830) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5832 = tensor.empty() : tensor<1x64xf32> + %5833 = "ttir.relu"(%5831, %5832) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5834 = tensor.empty() : tensor<1x64xf32> + %5835 = "ttir.relu"(%5833, %5834) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5836 = tensor.empty() : tensor<1x64xf32> + %5837 = "ttir.relu"(%5835, %5836) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5838 = tensor.empty() : tensor<1x64xf32> + %5839 = "ttir.relu"(%5837, %5838) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5840 = tensor.empty() : tensor<1x64xf32> + %5841 = "ttir.relu"(%5839, %5840) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5842 = tensor.empty() : tensor<1x64xf32> + %5843 = "ttir.relu"(%5841, %5842) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5844 = tensor.empty() : tensor<1x64xf32> + %5845 = "ttir.relu"(%5843, %5844) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5846 = tensor.empty() : tensor<1x64xf32> + %5847 = "ttir.relu"(%5845, %5846) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5848 = tensor.empty() : tensor<1x64xf32> + %5849 = "ttir.relu"(%5847, %5848) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5850 = tensor.empty() : tensor<1x64xf32> + %5851 = "ttir.relu"(%5849, %5850) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5852 = tensor.empty() : tensor<1x64xf32> + %5853 = "ttir.relu"(%5851, %5852) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5854 = tensor.empty() : tensor<1x64xf32> + %5855 = "ttir.relu"(%5853, %5854) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5856 = tensor.empty() : tensor<1x64xf32> + %5857 = "ttir.relu"(%5855, %5856) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5858 = tensor.empty() : tensor<1x64xf32> + %5859 = "ttir.relu"(%5857, %5858) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5860 = tensor.empty() : tensor<1x64xf32> + %5861 = "ttir.relu"(%5859, %5860) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5862 = tensor.empty() : tensor<1x64xf32> + %5863 = "ttir.relu"(%5861, %5862) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5864 = tensor.empty() : tensor<1x64xf32> + %5865 = "ttir.relu"(%5863, %5864) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5866 = tensor.empty() : tensor<1x64xf32> + %5867 = "ttir.relu"(%5865, %5866) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5868 = tensor.empty() : tensor<1x64xf32> + %5869 = "ttir.relu"(%5867, %5868) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5870 = tensor.empty() : tensor<1x64xf32> + %5871 = "ttir.relu"(%5869, %5870) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5872 = tensor.empty() : tensor<1x64xf32> + %5873 = "ttir.relu"(%5871, %5872) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5874 = tensor.empty() : tensor<1x64xf32> + %5875 = "ttir.relu"(%5873, %5874) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5876 = tensor.empty() : tensor<1x64xf32> + %5877 = "ttir.relu"(%5875, %5876) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5878 = tensor.empty() : tensor<1x64xf32> + %5879 = "ttir.relu"(%5877, %5878) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5880 = tensor.empty() : tensor<1x64xf32> + %5881 = "ttir.relu"(%5879, %5880) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5882 = tensor.empty() : tensor<1x64xf32> + %5883 = "ttir.relu"(%5881, %5882) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5884 = tensor.empty() : tensor<1x64xf32> + %5885 = "ttir.relu"(%5883, %5884) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5886 = tensor.empty() : tensor<1x64xf32> + %5887 = "ttir.relu"(%5885, %5886) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5888 = tensor.empty() : tensor<1x64xf32> + %5889 = "ttir.relu"(%5887, %5888) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5890 = tensor.empty() : tensor<1x64xf32> + %5891 = "ttir.relu"(%5889, %5890) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5892 = tensor.empty() : tensor<1x64xf32> + %5893 = "ttir.relu"(%5891, %5892) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5894 = tensor.empty() : tensor<1x64xf32> + %5895 = "ttir.relu"(%5893, %5894) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5896 = tensor.empty() : tensor<1x64xf32> + %5897 = "ttir.relu"(%5895, %5896) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5898 = tensor.empty() : tensor<1x64xf32> + %5899 = "ttir.relu"(%5897, %5898) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5900 = tensor.empty() : tensor<1x64xf32> + %5901 = "ttir.relu"(%5899, %5900) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5902 = tensor.empty() : tensor<1x64xf32> + %5903 = "ttir.relu"(%5901, %5902) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5904 = tensor.empty() : tensor<1x64xf32> + %5905 = "ttir.relu"(%5903, %5904) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5906 = tensor.empty() : tensor<1x64xf32> + %5907 = "ttir.relu"(%5905, %5906) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5908 = tensor.empty() : tensor<1x64xf32> + %5909 = "ttir.relu"(%5907, %5908) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5910 = tensor.empty() : tensor<1x64xf32> + %5911 = "ttir.relu"(%5909, %5910) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5912 = tensor.empty() : tensor<1x64xf32> + %5913 = "ttir.relu"(%5911, %5912) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5914 = tensor.empty() : tensor<1x64xf32> + %5915 = "ttir.relu"(%5913, %5914) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5916 = tensor.empty() : tensor<1x64xf32> + %5917 = "ttir.relu"(%5915, %5916) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5918 = tensor.empty() : tensor<1x64xf32> + %5919 = "ttir.relu"(%5917, %5918) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5920 = tensor.empty() : tensor<1x64xf32> + %5921 = "ttir.relu"(%5919, %5920) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5922 = tensor.empty() : tensor<1x64xf32> + %5923 = "ttir.relu"(%5921, %5922) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5924 = tensor.empty() : tensor<1x64xf32> + %5925 = "ttir.relu"(%5923, %5924) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5926 = tensor.empty() : tensor<1x64xf32> + %5927 = "ttir.relu"(%5925, %5926) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5928 = tensor.empty() : tensor<1x64xf32> + %5929 = "ttir.relu"(%5927, %5928) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5930 = tensor.empty() : tensor<1x64xf32> + %5931 = "ttir.relu"(%5929, %5930) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5932 = tensor.empty() : tensor<1x64xf32> + %5933 = "ttir.relu"(%5931, %5932) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5934 = tensor.empty() : tensor<1x64xf32> + %5935 = "ttir.relu"(%5933, %5934) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5936 = tensor.empty() : tensor<1x64xf32> + %5937 = "ttir.relu"(%5935, %5936) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5938 = tensor.empty() : tensor<1x64xf32> + %5939 = "ttir.relu"(%5937, %5938) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5940 = tensor.empty() : tensor<1x64xf32> + %5941 = "ttir.relu"(%5939, %5940) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5942 = tensor.empty() : tensor<1x64xf32> + %5943 = "ttir.relu"(%5941, %5942) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5944 = tensor.empty() : tensor<1x64xf32> + %5945 = "ttir.relu"(%5943, %5944) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5946 = tensor.empty() : tensor<1x64xf32> + %5947 = "ttir.relu"(%5945, %5946) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5948 = tensor.empty() : tensor<1x64xf32> + %5949 = "ttir.relu"(%5947, %5948) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5950 = tensor.empty() : tensor<1x64xf32> + %5951 = "ttir.relu"(%5949, %5950) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5952 = tensor.empty() : tensor<1x64xf32> + %5953 = "ttir.relu"(%5951, %5952) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5954 = tensor.empty() : tensor<1x64xf32> + %5955 = "ttir.relu"(%5953, %5954) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5956 = tensor.empty() : tensor<1x64xf32> + %5957 = "ttir.relu"(%5955, %5956) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5958 = tensor.empty() : tensor<1x64xf32> + %5959 = "ttir.relu"(%5957, %5958) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5960 = tensor.empty() : tensor<1x64xf32> + %5961 = "ttir.relu"(%5959, %5960) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5962 = tensor.empty() : tensor<1x64xf32> + %5963 = "ttir.relu"(%5961, %5962) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5964 = tensor.empty() : tensor<1x64xf32> + %5965 = "ttir.relu"(%5963, %5964) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5966 = tensor.empty() : tensor<1x64xf32> + %5967 = "ttir.relu"(%5965, %5966) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5968 = tensor.empty() : tensor<1x64xf32> + %5969 = "ttir.relu"(%5967, %5968) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5970 = tensor.empty() : tensor<1x64xf32> + %5971 = "ttir.relu"(%5969, %5970) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5972 = tensor.empty() : tensor<1x64xf32> + %5973 = "ttir.relu"(%5971, %5972) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5974 = tensor.empty() : tensor<1x64xf32> + %5975 = "ttir.relu"(%5973, %5974) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5976 = tensor.empty() : tensor<1x64xf32> + %5977 = "ttir.relu"(%5975, %5976) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5978 = tensor.empty() : tensor<1x64xf32> + %5979 = "ttir.relu"(%5977, %5978) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5980 = tensor.empty() : tensor<1x64xf32> + %5981 = "ttir.relu"(%5979, %5980) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5982 = tensor.empty() : tensor<1x64xf32> + %5983 = "ttir.relu"(%5981, %5982) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5984 = tensor.empty() : tensor<1x64xf32> + %5985 = "ttir.relu"(%5983, %5984) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5986 = tensor.empty() : tensor<1x64xf32> + %5987 = "ttir.relu"(%5985, %5986) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5988 = tensor.empty() : tensor<1x64xf32> + %5989 = "ttir.relu"(%5987, %5988) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5990 = tensor.empty() : tensor<1x64xf32> + %5991 = "ttir.relu"(%5989, %5990) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5992 = tensor.empty() : tensor<1x64xf32> + %5993 = "ttir.relu"(%5991, %5992) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5994 = tensor.empty() : tensor<1x64xf32> + %5995 = "ttir.relu"(%5993, %5994) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5996 = tensor.empty() : tensor<1x64xf32> + %5997 = "ttir.relu"(%5995, %5996) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %5998 = tensor.empty() : tensor<1x64xf32> + %5999 = "ttir.relu"(%5997, %5998) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6000 = tensor.empty() : tensor<1x64xf32> + %6001 = "ttir.relu"(%5999, %6000) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6002 = tensor.empty() : tensor<1x64xf32> + %6003 = "ttir.relu"(%6001, %6002) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6004 = tensor.empty() : tensor<1x64xf32> + %6005 = "ttir.relu"(%6003, %6004) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6006 = tensor.empty() : tensor<1x64xf32> + %6007 = "ttir.relu"(%6005, %6006) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6008 = tensor.empty() : tensor<1x64xf32> + %6009 = "ttir.relu"(%6007, %6008) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6010 = tensor.empty() : tensor<1x64xf32> + %6011 = "ttir.relu"(%6009, %6010) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6012 = tensor.empty() : tensor<1x64xf32> + %6013 = "ttir.relu"(%6011, %6012) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6014 = tensor.empty() : tensor<1x64xf32> + %6015 = "ttir.relu"(%6013, %6014) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6016 = tensor.empty() : tensor<1x64xf32> + %6017 = "ttir.relu"(%6015, %6016) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6018 = tensor.empty() : tensor<1x64xf32> + %6019 = "ttir.relu"(%6017, %6018) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6020 = tensor.empty() : tensor<1x64xf32> + %6021 = "ttir.relu"(%6019, %6020) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6022 = tensor.empty() : tensor<1x64xf32> + %6023 = "ttir.relu"(%6021, %6022) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6024 = tensor.empty() : tensor<1x64xf32> + %6025 = "ttir.relu"(%6023, %6024) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6026 = tensor.empty() : tensor<1x64xf32> + %6027 = "ttir.relu"(%6025, %6026) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6028 = tensor.empty() : tensor<1x64xf32> + %6029 = "ttir.relu"(%6027, %6028) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6030 = tensor.empty() : tensor<1x64xf32> + %6031 = "ttir.relu"(%6029, %6030) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6032 = tensor.empty() : tensor<1x64xf32> + %6033 = "ttir.relu"(%6031, %6032) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6034 = tensor.empty() : tensor<1x64xf32> + %6035 = "ttir.relu"(%6033, %6034) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6036 = tensor.empty() : tensor<1x64xf32> + %6037 = "ttir.relu"(%6035, %6036) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6038 = tensor.empty() : tensor<1x64xf32> + %6039 = "ttir.relu"(%6037, %6038) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6040 = tensor.empty() : tensor<1x64xf32> + %6041 = "ttir.relu"(%6039, %6040) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6042 = tensor.empty() : tensor<1x64xf32> + %6043 = "ttir.relu"(%6041, %6042) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6044 = tensor.empty() : tensor<1x64xf32> + %6045 = "ttir.relu"(%6043, %6044) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6046 = tensor.empty() : tensor<1x64xf32> + %6047 = "ttir.relu"(%6045, %6046) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6048 = tensor.empty() : tensor<1x64xf32> + %6049 = "ttir.relu"(%6047, %6048) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6050 = tensor.empty() : tensor<1x64xf32> + %6051 = "ttir.relu"(%6049, %6050) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6052 = tensor.empty() : tensor<1x64xf32> + %6053 = "ttir.relu"(%6051, %6052) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6054 = tensor.empty() : tensor<1x64xf32> + %6055 = "ttir.relu"(%6053, %6054) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6056 = tensor.empty() : tensor<1x64xf32> + %6057 = "ttir.relu"(%6055, %6056) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6058 = tensor.empty() : tensor<1x64xf32> + %6059 = "ttir.relu"(%6057, %6058) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6060 = tensor.empty() : tensor<1x64xf32> + %6061 = "ttir.relu"(%6059, %6060) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6062 = tensor.empty() : tensor<1x64xf32> + %6063 = "ttir.relu"(%6061, %6062) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6064 = tensor.empty() : tensor<1x64xf32> + %6065 = "ttir.relu"(%6063, %6064) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6066 = tensor.empty() : tensor<1x64xf32> + %6067 = "ttir.relu"(%6065, %6066) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6068 = tensor.empty() : tensor<1x64xf32> + %6069 = "ttir.relu"(%6067, %6068) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6070 = tensor.empty() : tensor<1x64xf32> + %6071 = "ttir.relu"(%6069, %6070) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6072 = tensor.empty() : tensor<1x64xf32> + %6073 = "ttir.relu"(%6071, %6072) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6074 = tensor.empty() : tensor<1x64xf32> + %6075 = "ttir.relu"(%6073, %6074) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6076 = tensor.empty() : tensor<1x64xf32> + %6077 = "ttir.relu"(%6075, %6076) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6078 = tensor.empty() : tensor<1x64xf32> + %6079 = "ttir.relu"(%6077, %6078) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6080 = tensor.empty() : tensor<1x64xf32> + %6081 = "ttir.relu"(%6079, %6080) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6082 = tensor.empty() : tensor<1x64xf32> + %6083 = "ttir.relu"(%6081, %6082) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6084 = tensor.empty() : tensor<1x64xf32> + %6085 = "ttir.relu"(%6083, %6084) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6086 = tensor.empty() : tensor<1x64xf32> + %6087 = "ttir.relu"(%6085, %6086) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6088 = tensor.empty() : tensor<1x64xf32> + %6089 = "ttir.relu"(%6087, %6088) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6090 = tensor.empty() : tensor<1x64xf32> + %6091 = "ttir.relu"(%6089, %6090) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6092 = tensor.empty() : tensor<1x64xf32> + %6093 = "ttir.relu"(%6091, %6092) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6094 = tensor.empty() : tensor<1x64xf32> + %6095 = "ttir.relu"(%6093, %6094) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6096 = tensor.empty() : tensor<1x64xf32> + %6097 = "ttir.relu"(%6095, %6096) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6098 = tensor.empty() : tensor<1x64xf32> + %6099 = "ttir.relu"(%6097, %6098) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6100 = tensor.empty() : tensor<1x64xf32> + %6101 = "ttir.relu"(%6099, %6100) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6102 = tensor.empty() : tensor<1x64xf32> + %6103 = "ttir.relu"(%6101, %6102) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6104 = tensor.empty() : tensor<1x64xf32> + %6105 = "ttir.relu"(%6103, %6104) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6106 = tensor.empty() : tensor<1x64xf32> + %6107 = "ttir.relu"(%6105, %6106) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6108 = tensor.empty() : tensor<1x64xf32> + %6109 = "ttir.relu"(%6107, %6108) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6110 = tensor.empty() : tensor<1x64xf32> + %6111 = "ttir.relu"(%6109, %6110) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6112 = tensor.empty() : tensor<1x64xf32> + %6113 = "ttir.relu"(%6111, %6112) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6114 = tensor.empty() : tensor<1x64xf32> + %6115 = "ttir.relu"(%6113, %6114) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6116 = tensor.empty() : tensor<1x64xf32> + %6117 = "ttir.relu"(%6115, %6116) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6118 = tensor.empty() : tensor<1x64xf32> + %6119 = "ttir.relu"(%6117, %6118) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6120 = tensor.empty() : tensor<1x64xf32> + %6121 = "ttir.relu"(%6119, %6120) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6122 = tensor.empty() : tensor<1x64xf32> + %6123 = "ttir.relu"(%6121, %6122) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6124 = tensor.empty() : tensor<1x64xf32> + %6125 = "ttir.relu"(%6123, %6124) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6126 = tensor.empty() : tensor<1x64xf32> + %6127 = "ttir.relu"(%6125, %6126) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6128 = tensor.empty() : tensor<1x64xf32> + %6129 = "ttir.relu"(%6127, %6128) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6130 = tensor.empty() : tensor<1x64xf32> + %6131 = "ttir.relu"(%6129, %6130) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6132 = tensor.empty() : tensor<1x64xf32> + %6133 = "ttir.relu"(%6131, %6132) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6134 = tensor.empty() : tensor<1x64xf32> + %6135 = "ttir.relu"(%6133, %6134) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6136 = tensor.empty() : tensor<1x64xf32> + %6137 = "ttir.relu"(%6135, %6136) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6138 = tensor.empty() : tensor<1x64xf32> + %6139 = "ttir.relu"(%6137, %6138) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6140 = tensor.empty() : tensor<1x64xf32> + %6141 = "ttir.relu"(%6139, %6140) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6142 = tensor.empty() : tensor<1x64xf32> + %6143 = "ttir.relu"(%6141, %6142) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6144 = tensor.empty() : tensor<1x64xf32> + %6145 = "ttir.relu"(%6143, %6144) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6146 = tensor.empty() : tensor<1x64xf32> + %6147 = "ttir.relu"(%6145, %6146) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6148 = tensor.empty() : tensor<1x64xf32> + %6149 = "ttir.relu"(%6147, %6148) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6150 = tensor.empty() : tensor<1x64xf32> + %6151 = "ttir.relu"(%6149, %6150) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6152 = tensor.empty() : tensor<1x64xf32> + %6153 = "ttir.relu"(%6151, %6152) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6154 = tensor.empty() : tensor<1x64xf32> + %6155 = "ttir.relu"(%6153, %6154) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6156 = tensor.empty() : tensor<1x64xf32> + %6157 = "ttir.relu"(%6155, %6156) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6158 = tensor.empty() : tensor<1x64xf32> + %6159 = "ttir.relu"(%6157, %6158) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6160 = tensor.empty() : tensor<1x64xf32> + %6161 = "ttir.relu"(%6159, %6160) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6162 = tensor.empty() : tensor<1x64xf32> + %6163 = "ttir.relu"(%6161, %6162) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6164 = tensor.empty() : tensor<1x64xf32> + %6165 = "ttir.relu"(%6163, %6164) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6166 = tensor.empty() : tensor<1x64xf32> + %6167 = "ttir.relu"(%6165, %6166) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6168 = tensor.empty() : tensor<1x64xf32> + %6169 = "ttir.relu"(%6167, %6168) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6170 = tensor.empty() : tensor<1x64xf32> + %6171 = "ttir.relu"(%6169, %6170) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6172 = tensor.empty() : tensor<1x64xf32> + %6173 = "ttir.relu"(%6171, %6172) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6174 = tensor.empty() : tensor<1x64xf32> + %6175 = "ttir.relu"(%6173, %6174) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6176 = tensor.empty() : tensor<1x64xf32> + %6177 = "ttir.relu"(%6175, %6176) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6178 = tensor.empty() : tensor<1x64xf32> + %6179 = "ttir.relu"(%6177, %6178) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6180 = tensor.empty() : tensor<1x64xf32> + %6181 = "ttir.relu"(%6179, %6180) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6182 = tensor.empty() : tensor<1x64xf32> + %6183 = "ttir.relu"(%6181, %6182) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6184 = tensor.empty() : tensor<1x64xf32> + %6185 = "ttir.relu"(%6183, %6184) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6186 = tensor.empty() : tensor<1x64xf32> + %6187 = "ttir.relu"(%6185, %6186) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6188 = tensor.empty() : tensor<1x64xf32> + %6189 = "ttir.relu"(%6187, %6188) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6190 = tensor.empty() : tensor<1x64xf32> + %6191 = "ttir.relu"(%6189, %6190) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6192 = tensor.empty() : tensor<1x64xf32> + %6193 = "ttir.relu"(%6191, %6192) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6194 = tensor.empty() : tensor<1x64xf32> + %6195 = "ttir.relu"(%6193, %6194) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6196 = tensor.empty() : tensor<1x64xf32> + %6197 = "ttir.relu"(%6195, %6196) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6198 = tensor.empty() : tensor<1x64xf32> + %6199 = "ttir.relu"(%6197, %6198) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6200 = tensor.empty() : tensor<1x64xf32> + %6201 = "ttir.relu"(%6199, %6200) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6202 = tensor.empty() : tensor<1x64xf32> + %6203 = "ttir.relu"(%6201, %6202) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6204 = tensor.empty() : tensor<1x64xf32> + %6205 = "ttir.relu"(%6203, %6204) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6206 = tensor.empty() : tensor<1x64xf32> + %6207 = "ttir.relu"(%6205, %6206) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6208 = tensor.empty() : tensor<1x64xf32> + %6209 = "ttir.relu"(%6207, %6208) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6210 = tensor.empty() : tensor<1x64xf32> + %6211 = "ttir.relu"(%6209, %6210) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6212 = tensor.empty() : tensor<1x64xf32> + %6213 = "ttir.relu"(%6211, %6212) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6214 = tensor.empty() : tensor<1x64xf32> + %6215 = "ttir.relu"(%6213, %6214) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6216 = tensor.empty() : tensor<1x64xf32> + %6217 = "ttir.relu"(%6215, %6216) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6218 = tensor.empty() : tensor<1x64xf32> + %6219 = "ttir.relu"(%6217, %6218) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6220 = tensor.empty() : tensor<1x64xf32> + %6221 = "ttir.relu"(%6219, %6220) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6222 = tensor.empty() : tensor<1x64xf32> + %6223 = "ttir.relu"(%6221, %6222) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6224 = tensor.empty() : tensor<1x64xf32> + %6225 = "ttir.relu"(%6223, %6224) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6226 = tensor.empty() : tensor<1x64xf32> + %6227 = "ttir.relu"(%6225, %6226) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6228 = tensor.empty() : tensor<1x64xf32> + %6229 = "ttir.relu"(%6227, %6228) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6230 = tensor.empty() : tensor<1x64xf32> + %6231 = "ttir.relu"(%6229, %6230) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6232 = tensor.empty() : tensor<1x64xf32> + %6233 = "ttir.relu"(%6231, %6232) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6234 = tensor.empty() : tensor<1x64xf32> + %6235 = "ttir.relu"(%6233, %6234) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6236 = tensor.empty() : tensor<1x64xf32> + %6237 = "ttir.relu"(%6235, %6236) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6238 = tensor.empty() : tensor<1x64xf32> + %6239 = "ttir.relu"(%6237, %6238) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6240 = tensor.empty() : tensor<1x64xf32> + %6241 = "ttir.relu"(%6239, %6240) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6242 = tensor.empty() : tensor<1x64xf32> + %6243 = "ttir.relu"(%6241, %6242) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6244 = tensor.empty() : tensor<1x64xf32> + %6245 = "ttir.relu"(%6243, %6244) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6246 = tensor.empty() : tensor<1x64xf32> + %6247 = "ttir.relu"(%6245, %6246) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6248 = tensor.empty() : tensor<1x64xf32> + %6249 = "ttir.relu"(%6247, %6248) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6250 = tensor.empty() : tensor<1x64xf32> + %6251 = "ttir.relu"(%6249, %6250) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6252 = tensor.empty() : tensor<1x64xf32> + %6253 = "ttir.relu"(%6251, %6252) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6254 = tensor.empty() : tensor<1x64xf32> + %6255 = "ttir.relu"(%6253, %6254) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6256 = tensor.empty() : tensor<1x64xf32> + %6257 = "ttir.relu"(%6255, %6256) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6258 = tensor.empty() : tensor<1x64xf32> + %6259 = "ttir.relu"(%6257, %6258) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6260 = tensor.empty() : tensor<1x64xf32> + %6261 = "ttir.relu"(%6259, %6260) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6262 = tensor.empty() : tensor<1x64xf32> + %6263 = "ttir.relu"(%6261, %6262) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6264 = tensor.empty() : tensor<1x64xf32> + %6265 = "ttir.relu"(%6263, %6264) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6266 = tensor.empty() : tensor<1x64xf32> + %6267 = "ttir.relu"(%6265, %6266) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6268 = tensor.empty() : tensor<1x64xf32> + %6269 = "ttir.relu"(%6267, %6268) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6270 = tensor.empty() : tensor<1x64xf32> + %6271 = "ttir.relu"(%6269, %6270) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6272 = tensor.empty() : tensor<1x64xf32> + %6273 = "ttir.relu"(%6271, %6272) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6274 = tensor.empty() : tensor<1x64xf32> + %6275 = "ttir.relu"(%6273, %6274) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6276 = tensor.empty() : tensor<1x64xf32> + %6277 = "ttir.relu"(%6275, %6276) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6278 = tensor.empty() : tensor<1x64xf32> + %6279 = "ttir.relu"(%6277, %6278) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6280 = tensor.empty() : tensor<1x64xf32> + %6281 = "ttir.relu"(%6279, %6280) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6282 = tensor.empty() : tensor<1x64xf32> + %6283 = "ttir.relu"(%6281, %6282) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6284 = tensor.empty() : tensor<1x64xf32> + %6285 = "ttir.relu"(%6283, %6284) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6286 = tensor.empty() : tensor<1x64xf32> + %6287 = "ttir.relu"(%6285, %6286) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6288 = tensor.empty() : tensor<1x64xf32> + %6289 = "ttir.relu"(%6287, %6288) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6290 = tensor.empty() : tensor<1x64xf32> + %6291 = "ttir.relu"(%6289, %6290) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6292 = tensor.empty() : tensor<1x64xf32> + %6293 = "ttir.relu"(%6291, %6292) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6294 = tensor.empty() : tensor<1x64xf32> + %6295 = "ttir.relu"(%6293, %6294) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6296 = tensor.empty() : tensor<1x64xf32> + %6297 = "ttir.relu"(%6295, %6296) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6298 = tensor.empty() : tensor<1x64xf32> + %6299 = "ttir.relu"(%6297, %6298) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6300 = tensor.empty() : tensor<1x64xf32> + %6301 = "ttir.relu"(%6299, %6300) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6302 = tensor.empty() : tensor<1x64xf32> + %6303 = "ttir.relu"(%6301, %6302) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6304 = tensor.empty() : tensor<1x64xf32> + %6305 = "ttir.relu"(%6303, %6304) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6306 = tensor.empty() : tensor<1x64xf32> + %6307 = "ttir.relu"(%6305, %6306) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6308 = tensor.empty() : tensor<1x64xf32> + %6309 = "ttir.relu"(%6307, %6308) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6310 = tensor.empty() : tensor<1x64xf32> + %6311 = "ttir.relu"(%6309, %6310) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6312 = tensor.empty() : tensor<1x64xf32> + %6313 = "ttir.relu"(%6311, %6312) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6314 = tensor.empty() : tensor<1x64xf32> + %6315 = "ttir.relu"(%6313, %6314) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6316 = tensor.empty() : tensor<1x64xf32> + %6317 = "ttir.relu"(%6315, %6316) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6318 = tensor.empty() : tensor<1x64xf32> + %6319 = "ttir.relu"(%6317, %6318) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6320 = tensor.empty() : tensor<1x64xf32> + %6321 = "ttir.relu"(%6319, %6320) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6322 = tensor.empty() : tensor<1x64xf32> + %6323 = "ttir.relu"(%6321, %6322) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6324 = tensor.empty() : tensor<1x64xf32> + %6325 = "ttir.relu"(%6323, %6324) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6326 = tensor.empty() : tensor<1x64xf32> + %6327 = "ttir.relu"(%6325, %6326) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6328 = tensor.empty() : tensor<1x64xf32> + %6329 = "ttir.relu"(%6327, %6328) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6330 = tensor.empty() : tensor<1x64xf32> + %6331 = "ttir.relu"(%6329, %6330) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6332 = tensor.empty() : tensor<1x64xf32> + %6333 = "ttir.relu"(%6331, %6332) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6334 = tensor.empty() : tensor<1x64xf32> + %6335 = "ttir.relu"(%6333, %6334) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6336 = tensor.empty() : tensor<1x64xf32> + %6337 = "ttir.relu"(%6335, %6336) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6338 = tensor.empty() : tensor<1x64xf32> + %6339 = "ttir.relu"(%6337, %6338) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6340 = tensor.empty() : tensor<1x64xf32> + %6341 = "ttir.relu"(%6339, %6340) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6342 = tensor.empty() : tensor<1x64xf32> + %6343 = "ttir.relu"(%6341, %6342) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6344 = tensor.empty() : tensor<1x64xf32> + %6345 = "ttir.relu"(%6343, %6344) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6346 = tensor.empty() : tensor<1x64xf32> + %6347 = "ttir.relu"(%6345, %6346) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6348 = tensor.empty() : tensor<1x64xf32> + %6349 = "ttir.relu"(%6347, %6348) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6350 = tensor.empty() : tensor<1x64xf32> + %6351 = "ttir.relu"(%6349, %6350) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6352 = tensor.empty() : tensor<1x64xf32> + %6353 = "ttir.relu"(%6351, %6352) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6354 = tensor.empty() : tensor<1x64xf32> + %6355 = "ttir.relu"(%6353, %6354) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6356 = tensor.empty() : tensor<1x64xf32> + %6357 = "ttir.relu"(%6355, %6356) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6358 = tensor.empty() : tensor<1x64xf32> + %6359 = "ttir.relu"(%6357, %6358) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6360 = tensor.empty() : tensor<1x64xf32> + %6361 = "ttir.relu"(%6359, %6360) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6362 = tensor.empty() : tensor<1x64xf32> + %6363 = "ttir.relu"(%6361, %6362) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6364 = tensor.empty() : tensor<1x64xf32> + %6365 = "ttir.relu"(%6363, %6364) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6366 = tensor.empty() : tensor<1x64xf32> + %6367 = "ttir.relu"(%6365, %6366) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6368 = tensor.empty() : tensor<1x64xf32> + %6369 = "ttir.relu"(%6367, %6368) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6370 = tensor.empty() : tensor<1x64xf32> + %6371 = "ttir.relu"(%6369, %6370) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6372 = tensor.empty() : tensor<1x64xf32> + %6373 = "ttir.relu"(%6371, %6372) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6374 = tensor.empty() : tensor<1x64xf32> + %6375 = "ttir.relu"(%6373, %6374) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6376 = tensor.empty() : tensor<1x64xf32> + %6377 = "ttir.relu"(%6375, %6376) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6378 = tensor.empty() : tensor<1x64xf32> + %6379 = "ttir.relu"(%6377, %6378) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6380 = tensor.empty() : tensor<1x64xf32> + %6381 = "ttir.relu"(%6379, %6380) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6382 = tensor.empty() : tensor<1x64xf32> + %6383 = "ttir.relu"(%6381, %6382) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6384 = tensor.empty() : tensor<1x64xf32> + %6385 = "ttir.relu"(%6383, %6384) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6386 = tensor.empty() : tensor<1x64xf32> + %6387 = "ttir.relu"(%6385, %6386) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6388 = tensor.empty() : tensor<1x64xf32> + %6389 = "ttir.relu"(%6387, %6388) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6390 = tensor.empty() : tensor<1x64xf32> + %6391 = "ttir.relu"(%6389, %6390) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6392 = tensor.empty() : tensor<1x64xf32> + %6393 = "ttir.relu"(%6391, %6392) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6394 = tensor.empty() : tensor<1x64xf32> + %6395 = "ttir.relu"(%6393, %6394) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6396 = tensor.empty() : tensor<1x64xf32> + %6397 = "ttir.relu"(%6395, %6396) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6398 = tensor.empty() : tensor<1x64xf32> + %6399 = "ttir.relu"(%6397, %6398) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6400 = tensor.empty() : tensor<1x64xf32> + %6401 = "ttir.relu"(%6399, %6400) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6402 = tensor.empty() : tensor<1x64xf32> + %6403 = "ttir.relu"(%6401, %6402) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6404 = tensor.empty() : tensor<1x64xf32> + %6405 = "ttir.relu"(%6403, %6404) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6406 = tensor.empty() : tensor<1x64xf32> + %6407 = "ttir.relu"(%6405, %6406) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6408 = tensor.empty() : tensor<1x64xf32> + %6409 = "ttir.relu"(%6407, %6408) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6410 = tensor.empty() : tensor<1x64xf32> + %6411 = "ttir.relu"(%6409, %6410) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6412 = tensor.empty() : tensor<1x64xf32> + %6413 = "ttir.relu"(%6411, %6412) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6414 = tensor.empty() : tensor<1x64xf32> + %6415 = "ttir.relu"(%6413, %6414) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6416 = tensor.empty() : tensor<1x64xf32> + %6417 = "ttir.relu"(%6415, %6416) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6418 = tensor.empty() : tensor<1x64xf32> + %6419 = "ttir.relu"(%6417, %6418) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6420 = tensor.empty() : tensor<1x64xf32> + %6421 = "ttir.relu"(%6419, %6420) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6422 = tensor.empty() : tensor<1x64xf32> + %6423 = "ttir.relu"(%6421, %6422) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6424 = tensor.empty() : tensor<1x64xf32> + %6425 = "ttir.relu"(%6423, %6424) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6426 = tensor.empty() : tensor<1x64xf32> + %6427 = "ttir.relu"(%6425, %6426) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6428 = tensor.empty() : tensor<1x64xf32> + %6429 = "ttir.relu"(%6427, %6428) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6430 = tensor.empty() : tensor<1x64xf32> + %6431 = "ttir.relu"(%6429, %6430) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6432 = tensor.empty() : tensor<1x64xf32> + %6433 = "ttir.relu"(%6431, %6432) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6434 = tensor.empty() : tensor<1x64xf32> + %6435 = "ttir.relu"(%6433, %6434) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6436 = tensor.empty() : tensor<1x64xf32> + %6437 = "ttir.relu"(%6435, %6436) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6438 = tensor.empty() : tensor<1x64xf32> + %6439 = "ttir.relu"(%6437, %6438) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6440 = tensor.empty() : tensor<1x64xf32> + %6441 = "ttir.relu"(%6439, %6440) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6442 = tensor.empty() : tensor<1x64xf32> + %6443 = "ttir.relu"(%6441, %6442) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6444 = tensor.empty() : tensor<1x64xf32> + %6445 = "ttir.relu"(%6443, %6444) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6446 = tensor.empty() : tensor<1x64xf32> + %6447 = "ttir.relu"(%6445, %6446) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6448 = tensor.empty() : tensor<1x64xf32> + %6449 = "ttir.relu"(%6447, %6448) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6450 = tensor.empty() : tensor<1x64xf32> + %6451 = "ttir.relu"(%6449, %6450) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6452 = tensor.empty() : tensor<1x64xf32> + %6453 = "ttir.relu"(%6451, %6452) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6454 = tensor.empty() : tensor<1x64xf32> + %6455 = "ttir.relu"(%6453, %6454) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6456 = tensor.empty() : tensor<1x64xf32> + %6457 = "ttir.relu"(%6455, %6456) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6458 = tensor.empty() : tensor<1x64xf32> + %6459 = "ttir.relu"(%6457, %6458) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6460 = tensor.empty() : tensor<1x64xf32> + %6461 = "ttir.relu"(%6459, %6460) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6462 = tensor.empty() : tensor<1x64xf32> + %6463 = "ttir.relu"(%6461, %6462) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6464 = tensor.empty() : tensor<1x64xf32> + %6465 = "ttir.relu"(%6463, %6464) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6466 = tensor.empty() : tensor<1x64xf32> + %6467 = "ttir.relu"(%6465, %6466) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6468 = tensor.empty() : tensor<1x64xf32> + %6469 = "ttir.relu"(%6467, %6468) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6470 = tensor.empty() : tensor<1x64xf32> + %6471 = "ttir.relu"(%6469, %6470) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6472 = tensor.empty() : tensor<1x64xf32> + %6473 = "ttir.relu"(%6471, %6472) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6474 = tensor.empty() : tensor<1x64xf32> + %6475 = "ttir.relu"(%6473, %6474) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6476 = tensor.empty() : tensor<1x64xf32> + %6477 = "ttir.relu"(%6475, %6476) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6478 = tensor.empty() : tensor<1x64xf32> + %6479 = "ttir.relu"(%6477, %6478) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6480 = tensor.empty() : tensor<1x64xf32> + %6481 = "ttir.relu"(%6479, %6480) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6482 = tensor.empty() : tensor<1x64xf32> + %6483 = "ttir.relu"(%6481, %6482) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6484 = tensor.empty() : tensor<1x64xf32> + %6485 = "ttir.relu"(%6483, %6484) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6486 = tensor.empty() : tensor<1x64xf32> + %6487 = "ttir.relu"(%6485, %6486) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6488 = tensor.empty() : tensor<1x64xf32> + %6489 = "ttir.relu"(%6487, %6488) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6490 = tensor.empty() : tensor<1x64xf32> + %6491 = "ttir.relu"(%6489, %6490) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6492 = tensor.empty() : tensor<1x64xf32> + %6493 = "ttir.relu"(%6491, %6492) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6494 = tensor.empty() : tensor<1x64xf32> + %6495 = "ttir.relu"(%6493, %6494) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6496 = tensor.empty() : tensor<1x64xf32> + %6497 = "ttir.relu"(%6495, %6496) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6498 = tensor.empty() : tensor<1x64xf32> + %6499 = "ttir.relu"(%6497, %6498) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6500 = tensor.empty() : tensor<1x64xf32> + %6501 = "ttir.relu"(%6499, %6500) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6502 = tensor.empty() : tensor<1x64xf32> + %6503 = "ttir.relu"(%6501, %6502) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6504 = tensor.empty() : tensor<1x64xf32> + %6505 = "ttir.relu"(%6503, %6504) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6506 = tensor.empty() : tensor<1x64xf32> + %6507 = "ttir.relu"(%6505, %6506) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6508 = tensor.empty() : tensor<1x64xf32> + %6509 = "ttir.relu"(%6507, %6508) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6510 = tensor.empty() : tensor<1x64xf32> + %6511 = "ttir.relu"(%6509, %6510) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6512 = tensor.empty() : tensor<1x64xf32> + %6513 = "ttir.relu"(%6511, %6512) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6514 = tensor.empty() : tensor<1x64xf32> + %6515 = "ttir.relu"(%6513, %6514) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6516 = tensor.empty() : tensor<1x64xf32> + %6517 = "ttir.relu"(%6515, %6516) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6518 = tensor.empty() : tensor<1x64xf32> + %6519 = "ttir.relu"(%6517, %6518) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6520 = tensor.empty() : tensor<1x64xf32> + %6521 = "ttir.relu"(%6519, %6520) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6522 = tensor.empty() : tensor<1x64xf32> + %6523 = "ttir.relu"(%6521, %6522) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6524 = tensor.empty() : tensor<1x64xf32> + %6525 = "ttir.relu"(%6523, %6524) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6526 = tensor.empty() : tensor<1x64xf32> + %6527 = "ttir.relu"(%6525, %6526) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6528 = tensor.empty() : tensor<1x64xf32> + %6529 = "ttir.relu"(%6527, %6528) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6530 = tensor.empty() : tensor<1x64xf32> + %6531 = "ttir.relu"(%6529, %6530) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6532 = tensor.empty() : tensor<1x64xf32> + %6533 = "ttir.relu"(%6531, %6532) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6534 = tensor.empty() : tensor<1x64xf32> + %6535 = "ttir.relu"(%6533, %6534) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6536 = tensor.empty() : tensor<1x64xf32> + %6537 = "ttir.relu"(%6535, %6536) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6538 = tensor.empty() : tensor<1x64xf32> + %6539 = "ttir.relu"(%6537, %6538) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6540 = tensor.empty() : tensor<1x64xf32> + %6541 = "ttir.relu"(%6539, %6540) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6542 = tensor.empty() : tensor<1x64xf32> + %6543 = "ttir.relu"(%6541, %6542) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6544 = tensor.empty() : tensor<1x64xf32> + %6545 = "ttir.relu"(%6543, %6544) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6546 = tensor.empty() : tensor<1x64xf32> + %6547 = "ttir.relu"(%6545, %6546) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6548 = tensor.empty() : tensor<1x64xf32> + %6549 = "ttir.relu"(%6547, %6548) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6550 = tensor.empty() : tensor<1x64xf32> + %6551 = "ttir.relu"(%6549, %6550) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6552 = tensor.empty() : tensor<1x64xf32> + %6553 = "ttir.relu"(%6551, %6552) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6554 = tensor.empty() : tensor<1x64xf32> + %6555 = "ttir.relu"(%6553, %6554) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6556 = tensor.empty() : tensor<1x64xf32> + %6557 = "ttir.relu"(%6555, %6556) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6558 = tensor.empty() : tensor<1x64xf32> + %6559 = "ttir.relu"(%6557, %6558) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6560 = tensor.empty() : tensor<1x64xf32> + %6561 = "ttir.relu"(%6559, %6560) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6562 = tensor.empty() : tensor<1x64xf32> + %6563 = "ttir.relu"(%6561, %6562) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6564 = tensor.empty() : tensor<1x64xf32> + %6565 = "ttir.relu"(%6563, %6564) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6566 = tensor.empty() : tensor<1x64xf32> + %6567 = "ttir.relu"(%6565, %6566) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6568 = tensor.empty() : tensor<1x64xf32> + %6569 = "ttir.relu"(%6567, %6568) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6570 = tensor.empty() : tensor<1x64xf32> + %6571 = "ttir.relu"(%6569, %6570) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6572 = tensor.empty() : tensor<1x64xf32> + %6573 = "ttir.relu"(%6571, %6572) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6574 = tensor.empty() : tensor<1x64xf32> + %6575 = "ttir.relu"(%6573, %6574) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6576 = tensor.empty() : tensor<1x64xf32> + %6577 = "ttir.relu"(%6575, %6576) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6578 = tensor.empty() : tensor<1x64xf32> + %6579 = "ttir.relu"(%6577, %6578) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6580 = tensor.empty() : tensor<1x64xf32> + %6581 = "ttir.relu"(%6579, %6580) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6582 = tensor.empty() : tensor<1x64xf32> + %6583 = "ttir.relu"(%6581, %6582) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6584 = tensor.empty() : tensor<1x64xf32> + %6585 = "ttir.relu"(%6583, %6584) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6586 = tensor.empty() : tensor<1x64xf32> + %6587 = "ttir.relu"(%6585, %6586) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6588 = tensor.empty() : tensor<1x64xf32> + %6589 = "ttir.relu"(%6587, %6588) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6590 = tensor.empty() : tensor<1x64xf32> + %6591 = "ttir.relu"(%6589, %6590) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6592 = tensor.empty() : tensor<1x64xf32> + %6593 = "ttir.relu"(%6591, %6592) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6594 = tensor.empty() : tensor<1x64xf32> + %6595 = "ttir.relu"(%6593, %6594) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6596 = tensor.empty() : tensor<1x64xf32> + %6597 = "ttir.relu"(%6595, %6596) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6598 = tensor.empty() : tensor<1x64xf32> + %6599 = "ttir.relu"(%6597, %6598) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6600 = tensor.empty() : tensor<1x64xf32> + %6601 = "ttir.relu"(%6599, %6600) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6602 = tensor.empty() : tensor<1x64xf32> + %6603 = "ttir.relu"(%6601, %6602) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6604 = tensor.empty() : tensor<1x64xf32> + %6605 = "ttir.relu"(%6603, %6604) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6606 = tensor.empty() : tensor<1x64xf32> + %6607 = "ttir.relu"(%6605, %6606) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6608 = tensor.empty() : tensor<1x64xf32> + %6609 = "ttir.relu"(%6607, %6608) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6610 = tensor.empty() : tensor<1x64xf32> + %6611 = "ttir.relu"(%6609, %6610) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6612 = tensor.empty() : tensor<1x64xf32> + %6613 = "ttir.relu"(%6611, %6612) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6614 = tensor.empty() : tensor<1x64xf32> + %6615 = "ttir.relu"(%6613, %6614) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6616 = tensor.empty() : tensor<1x64xf32> + %6617 = "ttir.relu"(%6615, %6616) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6618 = tensor.empty() : tensor<1x64xf32> + %6619 = "ttir.relu"(%6617, %6618) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6620 = tensor.empty() : tensor<1x64xf32> + %6621 = "ttir.relu"(%6619, %6620) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6622 = tensor.empty() : tensor<1x64xf32> + %6623 = "ttir.relu"(%6621, %6622) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6624 = tensor.empty() : tensor<1x64xf32> + %6625 = "ttir.relu"(%6623, %6624) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6626 = tensor.empty() : tensor<1x64xf32> + %6627 = "ttir.relu"(%6625, %6626) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6628 = tensor.empty() : tensor<1x64xf32> + %6629 = "ttir.relu"(%6627, %6628) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6630 = tensor.empty() : tensor<1x64xf32> + %6631 = "ttir.relu"(%6629, %6630) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6632 = tensor.empty() : tensor<1x64xf32> + %6633 = "ttir.relu"(%6631, %6632) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6634 = tensor.empty() : tensor<1x64xf32> + %6635 = "ttir.relu"(%6633, %6634) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6636 = tensor.empty() : tensor<1x64xf32> + %6637 = "ttir.relu"(%6635, %6636) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6638 = tensor.empty() : tensor<1x64xf32> + %6639 = "ttir.relu"(%6637, %6638) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6640 = tensor.empty() : tensor<1x64xf32> + %6641 = "ttir.relu"(%6639, %6640) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6642 = tensor.empty() : tensor<1x64xf32> + %6643 = "ttir.relu"(%6641, %6642) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6644 = tensor.empty() : tensor<1x64xf32> + %6645 = "ttir.relu"(%6643, %6644) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6646 = tensor.empty() : tensor<1x64xf32> + %6647 = "ttir.relu"(%6645, %6646) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6648 = tensor.empty() : tensor<1x64xf32> + %6649 = "ttir.relu"(%6647, %6648) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6650 = tensor.empty() : tensor<1x64xf32> + %6651 = "ttir.relu"(%6649, %6650) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6652 = tensor.empty() : tensor<1x64xf32> + %6653 = "ttir.relu"(%6651, %6652) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6654 = tensor.empty() : tensor<1x64xf32> + %6655 = "ttir.relu"(%6653, %6654) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6656 = tensor.empty() : tensor<1x64xf32> + %6657 = "ttir.relu"(%6655, %6656) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6658 = tensor.empty() : tensor<1x64xf32> + %6659 = "ttir.relu"(%6657, %6658) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6660 = tensor.empty() : tensor<1x64xf32> + %6661 = "ttir.relu"(%6659, %6660) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6662 = tensor.empty() : tensor<1x64xf32> + %6663 = "ttir.relu"(%6661, %6662) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6664 = tensor.empty() : tensor<1x64xf32> + %6665 = "ttir.relu"(%6663, %6664) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6666 = tensor.empty() : tensor<1x64xf32> + %6667 = "ttir.relu"(%6665, %6666) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6668 = tensor.empty() : tensor<1x64xf32> + %6669 = "ttir.relu"(%6667, %6668) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6670 = tensor.empty() : tensor<1x64xf32> + %6671 = "ttir.relu"(%6669, %6670) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6672 = tensor.empty() : tensor<1x64xf32> + %6673 = "ttir.relu"(%6671, %6672) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6674 = tensor.empty() : tensor<1x64xf32> + %6675 = "ttir.relu"(%6673, %6674) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6676 = tensor.empty() : tensor<1x64xf32> + %6677 = "ttir.relu"(%6675, %6676) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6678 = tensor.empty() : tensor<1x64xf32> + %6679 = "ttir.relu"(%6677, %6678) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6680 = tensor.empty() : tensor<1x64xf32> + %6681 = "ttir.relu"(%6679, %6680) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6682 = tensor.empty() : tensor<1x64xf32> + %6683 = "ttir.relu"(%6681, %6682) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6684 = tensor.empty() : tensor<1x64xf32> + %6685 = "ttir.relu"(%6683, %6684) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6686 = tensor.empty() : tensor<1x64xf32> + %6687 = "ttir.relu"(%6685, %6686) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6688 = tensor.empty() : tensor<1x64xf32> + %6689 = "ttir.relu"(%6687, %6688) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6690 = tensor.empty() : tensor<1x64xf32> + %6691 = "ttir.relu"(%6689, %6690) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6692 = tensor.empty() : tensor<1x64xf32> + %6693 = "ttir.relu"(%6691, %6692) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6694 = tensor.empty() : tensor<1x64xf32> + %6695 = "ttir.relu"(%6693, %6694) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6696 = tensor.empty() : tensor<1x64xf32> + %6697 = "ttir.relu"(%6695, %6696) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6698 = tensor.empty() : tensor<1x64xf32> + %6699 = "ttir.relu"(%6697, %6698) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6700 = tensor.empty() : tensor<1x64xf32> + %6701 = "ttir.relu"(%6699, %6700) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6702 = tensor.empty() : tensor<1x64xf32> + %6703 = "ttir.relu"(%6701, %6702) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6704 = tensor.empty() : tensor<1x64xf32> + %6705 = "ttir.relu"(%6703, %6704) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6706 = tensor.empty() : tensor<1x64xf32> + %6707 = "ttir.relu"(%6705, %6706) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6708 = tensor.empty() : tensor<1x64xf32> + %6709 = "ttir.relu"(%6707, %6708) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6710 = tensor.empty() : tensor<1x64xf32> + %6711 = "ttir.relu"(%6709, %6710) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6712 = tensor.empty() : tensor<1x64xf32> + %6713 = "ttir.relu"(%6711, %6712) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6714 = tensor.empty() : tensor<1x64xf32> + %6715 = "ttir.relu"(%6713, %6714) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6716 = tensor.empty() : tensor<1x64xf32> + %6717 = "ttir.relu"(%6715, %6716) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6718 = tensor.empty() : tensor<1x64xf32> + %6719 = "ttir.relu"(%6717, %6718) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6720 = tensor.empty() : tensor<1x64xf32> + %6721 = "ttir.relu"(%6719, %6720) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6722 = tensor.empty() : tensor<1x64xf32> + %6723 = "ttir.relu"(%6721, %6722) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6724 = tensor.empty() : tensor<1x64xf32> + %6725 = "ttir.relu"(%6723, %6724) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6726 = tensor.empty() : tensor<1x64xf32> + %6727 = "ttir.relu"(%6725, %6726) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6728 = tensor.empty() : tensor<1x64xf32> + %6729 = "ttir.relu"(%6727, %6728) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6730 = tensor.empty() : tensor<1x64xf32> + %6731 = "ttir.relu"(%6729, %6730) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6732 = tensor.empty() : tensor<1x64xf32> + %6733 = "ttir.relu"(%6731, %6732) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6734 = tensor.empty() : tensor<1x64xf32> + %6735 = "ttir.relu"(%6733, %6734) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6736 = tensor.empty() : tensor<1x64xf32> + %6737 = "ttir.relu"(%6735, %6736) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6738 = tensor.empty() : tensor<1x64xf32> + %6739 = "ttir.relu"(%6737, %6738) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6740 = tensor.empty() : tensor<1x64xf32> + %6741 = "ttir.relu"(%6739, %6740) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6742 = tensor.empty() : tensor<1x64xf32> + %6743 = "ttir.relu"(%6741, %6742) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6744 = tensor.empty() : tensor<1x64xf32> + %6745 = "ttir.relu"(%6743, %6744) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6746 = tensor.empty() : tensor<1x64xf32> + %6747 = "ttir.relu"(%6745, %6746) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6748 = tensor.empty() : tensor<1x64xf32> + %6749 = "ttir.relu"(%6747, %6748) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6750 = tensor.empty() : tensor<1x64xf32> + %6751 = "ttir.relu"(%6749, %6750) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6752 = tensor.empty() : tensor<1x64xf32> + %6753 = "ttir.relu"(%6751, %6752) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6754 = tensor.empty() : tensor<1x64xf32> + %6755 = "ttir.relu"(%6753, %6754) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6756 = tensor.empty() : tensor<1x64xf32> + %6757 = "ttir.relu"(%6755, %6756) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6758 = tensor.empty() : tensor<1x64xf32> + %6759 = "ttir.relu"(%6757, %6758) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6760 = tensor.empty() : tensor<1x64xf32> + %6761 = "ttir.relu"(%6759, %6760) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6762 = tensor.empty() : tensor<1x64xf32> + %6763 = "ttir.relu"(%6761, %6762) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6764 = tensor.empty() : tensor<1x64xf32> + %6765 = "ttir.relu"(%6763, %6764) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6766 = tensor.empty() : tensor<1x64xf32> + %6767 = "ttir.relu"(%6765, %6766) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6768 = tensor.empty() : tensor<1x64xf32> + %6769 = "ttir.relu"(%6767, %6768) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6770 = tensor.empty() : tensor<1x64xf32> + %6771 = "ttir.relu"(%6769, %6770) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6772 = tensor.empty() : tensor<1x64xf32> + %6773 = "ttir.relu"(%6771, %6772) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6774 = tensor.empty() : tensor<1x64xf32> + %6775 = "ttir.relu"(%6773, %6774) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6776 = tensor.empty() : tensor<1x64xf32> + %6777 = "ttir.relu"(%6775, %6776) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6778 = tensor.empty() : tensor<1x64xf32> + %6779 = "ttir.relu"(%6777, %6778) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6780 = tensor.empty() : tensor<1x64xf32> + %6781 = "ttir.relu"(%6779, %6780) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6782 = tensor.empty() : tensor<1x64xf32> + %6783 = "ttir.relu"(%6781, %6782) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6784 = tensor.empty() : tensor<1x64xf32> + %6785 = "ttir.relu"(%6783, %6784) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6786 = tensor.empty() : tensor<1x64xf32> + %6787 = "ttir.relu"(%6785, %6786) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6788 = tensor.empty() : tensor<1x64xf32> + %6789 = "ttir.relu"(%6787, %6788) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6790 = tensor.empty() : tensor<1x64xf32> + %6791 = "ttir.relu"(%6789, %6790) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6792 = tensor.empty() : tensor<1x64xf32> + %6793 = "ttir.relu"(%6791, %6792) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6794 = tensor.empty() : tensor<1x64xf32> + %6795 = "ttir.relu"(%6793, %6794) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6796 = tensor.empty() : tensor<1x64xf32> + %6797 = "ttir.relu"(%6795, %6796) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6798 = tensor.empty() : tensor<1x64xf32> + %6799 = "ttir.relu"(%6797, %6798) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6800 = tensor.empty() : tensor<1x64xf32> + %6801 = "ttir.relu"(%6799, %6800) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6802 = tensor.empty() : tensor<1x64xf32> + %6803 = "ttir.relu"(%6801, %6802) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6804 = tensor.empty() : tensor<1x64xf32> + %6805 = "ttir.relu"(%6803, %6804) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6806 = tensor.empty() : tensor<1x64xf32> + %6807 = "ttir.relu"(%6805, %6806) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6808 = tensor.empty() : tensor<1x64xf32> + %6809 = "ttir.relu"(%6807, %6808) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6810 = tensor.empty() : tensor<1x64xf32> + %6811 = "ttir.relu"(%6809, %6810) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6812 = tensor.empty() : tensor<1x64xf32> + %6813 = "ttir.relu"(%6811, %6812) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6814 = tensor.empty() : tensor<1x64xf32> + %6815 = "ttir.relu"(%6813, %6814) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6816 = tensor.empty() : tensor<1x64xf32> + %6817 = "ttir.relu"(%6815, %6816) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6818 = tensor.empty() : tensor<1x64xf32> + %6819 = "ttir.relu"(%6817, %6818) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6820 = tensor.empty() : tensor<1x64xf32> + %6821 = "ttir.relu"(%6819, %6820) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6822 = tensor.empty() : tensor<1x64xf32> + %6823 = "ttir.relu"(%6821, %6822) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6824 = tensor.empty() : tensor<1x64xf32> + %6825 = "ttir.relu"(%6823, %6824) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6826 = tensor.empty() : tensor<1x64xf32> + %6827 = "ttir.relu"(%6825, %6826) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6828 = tensor.empty() : tensor<1x64xf32> + %6829 = "ttir.relu"(%6827, %6828) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6830 = tensor.empty() : tensor<1x64xf32> + %6831 = "ttir.relu"(%6829, %6830) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6832 = tensor.empty() : tensor<1x64xf32> + %6833 = "ttir.relu"(%6831, %6832) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6834 = tensor.empty() : tensor<1x64xf32> + %6835 = "ttir.relu"(%6833, %6834) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6836 = tensor.empty() : tensor<1x64xf32> + %6837 = "ttir.relu"(%6835, %6836) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6838 = tensor.empty() : tensor<1x64xf32> + %6839 = "ttir.relu"(%6837, %6838) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6840 = tensor.empty() : tensor<1x64xf32> + %6841 = "ttir.relu"(%6839, %6840) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6842 = tensor.empty() : tensor<1x64xf32> + %6843 = "ttir.relu"(%6841, %6842) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6844 = tensor.empty() : tensor<1x64xf32> + %6845 = "ttir.relu"(%6843, %6844) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6846 = tensor.empty() : tensor<1x64xf32> + %6847 = "ttir.relu"(%6845, %6846) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6848 = tensor.empty() : tensor<1x64xf32> + %6849 = "ttir.relu"(%6847, %6848) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6850 = tensor.empty() : tensor<1x64xf32> + %6851 = "ttir.relu"(%6849, %6850) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6852 = tensor.empty() : tensor<1x64xf32> + %6853 = "ttir.relu"(%6851, %6852) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6854 = tensor.empty() : tensor<1x64xf32> + %6855 = "ttir.relu"(%6853, %6854) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6856 = tensor.empty() : tensor<1x64xf32> + %6857 = "ttir.relu"(%6855, %6856) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6858 = tensor.empty() : tensor<1x64xf32> + %6859 = "ttir.relu"(%6857, %6858) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6860 = tensor.empty() : tensor<1x64xf32> + %6861 = "ttir.relu"(%6859, %6860) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6862 = tensor.empty() : tensor<1x64xf32> + %6863 = "ttir.relu"(%6861, %6862) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6864 = tensor.empty() : tensor<1x64xf32> + %6865 = "ttir.relu"(%6863, %6864) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6866 = tensor.empty() : tensor<1x64xf32> + %6867 = "ttir.relu"(%6865, %6866) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6868 = tensor.empty() : tensor<1x64xf32> + %6869 = "ttir.relu"(%6867, %6868) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6870 = tensor.empty() : tensor<1x64xf32> + %6871 = "ttir.relu"(%6869, %6870) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6872 = tensor.empty() : tensor<1x64xf32> + %6873 = "ttir.relu"(%6871, %6872) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6874 = tensor.empty() : tensor<1x64xf32> + %6875 = "ttir.relu"(%6873, %6874) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6876 = tensor.empty() : tensor<1x64xf32> + %6877 = "ttir.relu"(%6875, %6876) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6878 = tensor.empty() : tensor<1x64xf32> + %6879 = "ttir.relu"(%6877, %6878) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6880 = tensor.empty() : tensor<1x64xf32> + %6881 = "ttir.relu"(%6879, %6880) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6882 = tensor.empty() : tensor<1x64xf32> + %6883 = "ttir.relu"(%6881, %6882) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6884 = tensor.empty() : tensor<1x64xf32> + %6885 = "ttir.relu"(%6883, %6884) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6886 = tensor.empty() : tensor<1x64xf32> + %6887 = "ttir.relu"(%6885, %6886) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6888 = tensor.empty() : tensor<1x64xf32> + %6889 = "ttir.relu"(%6887, %6888) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6890 = tensor.empty() : tensor<1x64xf32> + %6891 = "ttir.relu"(%6889, %6890) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6892 = tensor.empty() : tensor<1x64xf32> + %6893 = "ttir.relu"(%6891, %6892) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6894 = tensor.empty() : tensor<1x64xf32> + %6895 = "ttir.relu"(%6893, %6894) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6896 = tensor.empty() : tensor<1x64xf32> + %6897 = "ttir.relu"(%6895, %6896) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6898 = tensor.empty() : tensor<1x64xf32> + %6899 = "ttir.relu"(%6897, %6898) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6900 = tensor.empty() : tensor<1x64xf32> + %6901 = "ttir.relu"(%6899, %6900) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6902 = tensor.empty() : tensor<1x64xf32> + %6903 = "ttir.relu"(%6901, %6902) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6904 = tensor.empty() : tensor<1x64xf32> + %6905 = "ttir.relu"(%6903, %6904) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6906 = tensor.empty() : tensor<1x64xf32> + %6907 = "ttir.relu"(%6905, %6906) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6908 = tensor.empty() : tensor<1x64xf32> + %6909 = "ttir.relu"(%6907, %6908) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6910 = tensor.empty() : tensor<1x64xf32> + %6911 = "ttir.relu"(%6909, %6910) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6912 = tensor.empty() : tensor<1x64xf32> + %6913 = "ttir.relu"(%6911, %6912) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6914 = tensor.empty() : tensor<1x64xf32> + %6915 = "ttir.relu"(%6913, %6914) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6916 = tensor.empty() : tensor<1x64xf32> + %6917 = "ttir.relu"(%6915, %6916) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6918 = tensor.empty() : tensor<1x64xf32> + %6919 = "ttir.relu"(%6917, %6918) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6920 = tensor.empty() : tensor<1x64xf32> + %6921 = "ttir.relu"(%6919, %6920) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6922 = tensor.empty() : tensor<1x64xf32> + %6923 = "ttir.relu"(%6921, %6922) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6924 = tensor.empty() : tensor<1x64xf32> + %6925 = "ttir.relu"(%6923, %6924) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6926 = tensor.empty() : tensor<1x64xf32> + %6927 = "ttir.relu"(%6925, %6926) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6928 = tensor.empty() : tensor<1x64xf32> + %6929 = "ttir.relu"(%6927, %6928) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6930 = tensor.empty() : tensor<1x64xf32> + %6931 = "ttir.relu"(%6929, %6930) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6932 = tensor.empty() : tensor<1x64xf32> + %6933 = "ttir.relu"(%6931, %6932) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6934 = tensor.empty() : tensor<1x64xf32> + %6935 = "ttir.relu"(%6933, %6934) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6936 = tensor.empty() : tensor<1x64xf32> + %6937 = "ttir.relu"(%6935, %6936) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6938 = tensor.empty() : tensor<1x64xf32> + %6939 = "ttir.relu"(%6937, %6938) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6940 = tensor.empty() : tensor<1x64xf32> + %6941 = "ttir.relu"(%6939, %6940) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6942 = tensor.empty() : tensor<1x64xf32> + %6943 = "ttir.relu"(%6941, %6942) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6944 = tensor.empty() : tensor<1x64xf32> + %6945 = "ttir.relu"(%6943, %6944) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6946 = tensor.empty() : tensor<1x64xf32> + %6947 = "ttir.relu"(%6945, %6946) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6948 = tensor.empty() : tensor<1x64xf32> + %6949 = "ttir.relu"(%6947, %6948) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6950 = tensor.empty() : tensor<1x64xf32> + %6951 = "ttir.relu"(%6949, %6950) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6952 = tensor.empty() : tensor<1x64xf32> + %6953 = "ttir.relu"(%6951, %6952) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6954 = tensor.empty() : tensor<1x64xf32> + %6955 = "ttir.relu"(%6953, %6954) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6956 = tensor.empty() : tensor<1x64xf32> + %6957 = "ttir.relu"(%6955, %6956) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6958 = tensor.empty() : tensor<1x64xf32> + %6959 = "ttir.relu"(%6957, %6958) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6960 = tensor.empty() : tensor<1x64xf32> + %6961 = "ttir.relu"(%6959, %6960) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6962 = tensor.empty() : tensor<1x64xf32> + %6963 = "ttir.relu"(%6961, %6962) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6964 = tensor.empty() : tensor<1x64xf32> + %6965 = "ttir.relu"(%6963, %6964) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6966 = tensor.empty() : tensor<1x64xf32> + %6967 = "ttir.relu"(%6965, %6966) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6968 = tensor.empty() : tensor<1x64xf32> + %6969 = "ttir.relu"(%6967, %6968) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6970 = tensor.empty() : tensor<1x64xf32> + %6971 = "ttir.relu"(%6969, %6970) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6972 = tensor.empty() : tensor<1x64xf32> + %6973 = "ttir.relu"(%6971, %6972) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6974 = tensor.empty() : tensor<1x64xf32> + %6975 = "ttir.relu"(%6973, %6974) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6976 = tensor.empty() : tensor<1x64xf32> + %6977 = "ttir.relu"(%6975, %6976) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6978 = tensor.empty() : tensor<1x64xf32> + %6979 = "ttir.relu"(%6977, %6978) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6980 = tensor.empty() : tensor<1x64xf32> + %6981 = "ttir.relu"(%6979, %6980) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6982 = tensor.empty() : tensor<1x64xf32> + %6983 = "ttir.relu"(%6981, %6982) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6984 = tensor.empty() : tensor<1x64xf32> + %6985 = "ttir.relu"(%6983, %6984) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6986 = tensor.empty() : tensor<1x64xf32> + %6987 = "ttir.relu"(%6985, %6986) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6988 = tensor.empty() : tensor<1x64xf32> + %6989 = "ttir.relu"(%6987, %6988) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6990 = tensor.empty() : tensor<1x64xf32> + %6991 = "ttir.relu"(%6989, %6990) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6992 = tensor.empty() : tensor<1x64xf32> + %6993 = "ttir.relu"(%6991, %6992) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6994 = tensor.empty() : tensor<1x64xf32> + %6995 = "ttir.relu"(%6993, %6994) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6996 = tensor.empty() : tensor<1x64xf32> + %6997 = "ttir.relu"(%6995, %6996) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6998 = tensor.empty() : tensor<1x64xf32> + %6999 = "ttir.relu"(%6997, %6998) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7000 = tensor.empty() : tensor<1x64xf32> + %7001 = "ttir.relu"(%6999, %7000) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7002 = tensor.empty() : tensor<1x64xf32> + %7003 = "ttir.relu"(%7001, %7002) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7004 = tensor.empty() : tensor<1x64xf32> + %7005 = "ttir.relu"(%7003, %7004) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7006 = tensor.empty() : tensor<1x64xf32> + %7007 = "ttir.relu"(%7005, %7006) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7008 = tensor.empty() : tensor<1x64xf32> + %7009 = "ttir.relu"(%7007, %7008) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7010 = tensor.empty() : tensor<1x64xf32> + %7011 = "ttir.relu"(%7009, %7010) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7012 = tensor.empty() : tensor<1x64xf32> + %7013 = "ttir.relu"(%7011, %7012) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7014 = tensor.empty() : tensor<1x64xf32> + %7015 = "ttir.relu"(%7013, %7014) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7016 = tensor.empty() : tensor<1x64xf32> + %7017 = "ttir.relu"(%7015, %7016) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7018 = tensor.empty() : tensor<1x64xf32> + %7019 = "ttir.relu"(%7017, %7018) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7020 = tensor.empty() : tensor<1x64xf32> + %7021 = "ttir.relu"(%7019, %7020) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7022 = tensor.empty() : tensor<1x64xf32> + %7023 = "ttir.relu"(%7021, %7022) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7024 = tensor.empty() : tensor<1x64xf32> + %7025 = "ttir.relu"(%7023, %7024) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7026 = tensor.empty() : tensor<1x64xf32> + %7027 = "ttir.relu"(%7025, %7026) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7028 = tensor.empty() : tensor<1x64xf32> + %7029 = "ttir.relu"(%7027, %7028) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7030 = tensor.empty() : tensor<1x64xf32> + %7031 = "ttir.relu"(%7029, %7030) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7032 = tensor.empty() : tensor<1x64xf32> + %7033 = "ttir.relu"(%7031, %7032) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7034 = tensor.empty() : tensor<1x64xf32> + %7035 = "ttir.relu"(%7033, %7034) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7036 = tensor.empty() : tensor<1x64xf32> + %7037 = "ttir.relu"(%7035, %7036) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7038 = tensor.empty() : tensor<1x64xf32> + %7039 = "ttir.relu"(%7037, %7038) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7040 = tensor.empty() : tensor<1x64xf32> + %7041 = "ttir.relu"(%7039, %7040) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7042 = tensor.empty() : tensor<1x64xf32> + %7043 = "ttir.relu"(%7041, %7042) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7044 = tensor.empty() : tensor<1x64xf32> + %7045 = "ttir.relu"(%7043, %7044) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7046 = tensor.empty() : tensor<1x64xf32> + %7047 = "ttir.relu"(%7045, %7046) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7048 = tensor.empty() : tensor<1x64xf32> + %7049 = "ttir.relu"(%7047, %7048) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7050 = tensor.empty() : tensor<1x64xf32> + %7051 = "ttir.relu"(%7049, %7050) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7052 = tensor.empty() : tensor<1x64xf32> + %7053 = "ttir.relu"(%7051, %7052) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7054 = tensor.empty() : tensor<1x64xf32> + %7055 = "ttir.relu"(%7053, %7054) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7056 = tensor.empty() : tensor<1x64xf32> + %7057 = "ttir.relu"(%7055, %7056) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7058 = tensor.empty() : tensor<1x64xf32> + %7059 = "ttir.relu"(%7057, %7058) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7060 = tensor.empty() : tensor<1x64xf32> + %7061 = "ttir.relu"(%7059, %7060) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7062 = tensor.empty() : tensor<1x64xf32> + %7063 = "ttir.relu"(%7061, %7062) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7064 = tensor.empty() : tensor<1x64xf32> + %7065 = "ttir.relu"(%7063, %7064) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7066 = tensor.empty() : tensor<1x64xf32> + %7067 = "ttir.relu"(%7065, %7066) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7068 = tensor.empty() : tensor<1x64xf32> + %7069 = "ttir.relu"(%7067, %7068) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7070 = tensor.empty() : tensor<1x64xf32> + %7071 = "ttir.relu"(%7069, %7070) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7072 = tensor.empty() : tensor<1x64xf32> + %7073 = "ttir.relu"(%7071, %7072) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7074 = tensor.empty() : tensor<1x64xf32> + %7075 = "ttir.relu"(%7073, %7074) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7076 = tensor.empty() : tensor<1x64xf32> + %7077 = "ttir.relu"(%7075, %7076) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7078 = tensor.empty() : tensor<1x64xf32> + %7079 = "ttir.relu"(%7077, %7078) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7080 = tensor.empty() : tensor<1x64xf32> + %7081 = "ttir.relu"(%7079, %7080) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7082 = tensor.empty() : tensor<1x64xf32> + %7083 = "ttir.relu"(%7081, %7082) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7084 = tensor.empty() : tensor<1x64xf32> + %7085 = "ttir.relu"(%7083, %7084) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7086 = tensor.empty() : tensor<1x64xf32> + %7087 = "ttir.relu"(%7085, %7086) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7088 = tensor.empty() : tensor<1x64xf32> + %7089 = "ttir.relu"(%7087, %7088) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7090 = tensor.empty() : tensor<1x64xf32> + %7091 = "ttir.relu"(%7089, %7090) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7092 = tensor.empty() : tensor<1x64xf32> + %7093 = "ttir.relu"(%7091, %7092) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7094 = tensor.empty() : tensor<1x64xf32> + %7095 = "ttir.relu"(%7093, %7094) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7096 = tensor.empty() : tensor<1x64xf32> + %7097 = "ttir.relu"(%7095, %7096) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7098 = tensor.empty() : tensor<1x64xf32> + %7099 = "ttir.relu"(%7097, %7098) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7100 = tensor.empty() : tensor<1x64xf32> + %7101 = "ttir.relu"(%7099, %7100) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7102 = tensor.empty() : tensor<1x64xf32> + %7103 = "ttir.relu"(%7101, %7102) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7104 = tensor.empty() : tensor<1x64xf32> + %7105 = "ttir.relu"(%7103, %7104) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7106 = tensor.empty() : tensor<1x64xf32> + %7107 = "ttir.relu"(%7105, %7106) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7108 = tensor.empty() : tensor<1x64xf32> + %7109 = "ttir.relu"(%7107, %7108) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7110 = tensor.empty() : tensor<1x64xf32> + %7111 = "ttir.relu"(%7109, %7110) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7112 = tensor.empty() : tensor<1x64xf32> + %7113 = "ttir.relu"(%7111, %7112) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7114 = tensor.empty() : tensor<1x64xf32> + %7115 = "ttir.relu"(%7113, %7114) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7116 = tensor.empty() : tensor<1x64xf32> + %7117 = "ttir.relu"(%7115, %7116) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7118 = tensor.empty() : tensor<1x64xf32> + %7119 = "ttir.relu"(%7117, %7118) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7120 = tensor.empty() : tensor<1x64xf32> + %7121 = "ttir.relu"(%7119, %7120) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7122 = tensor.empty() : tensor<1x64xf32> + %7123 = "ttir.relu"(%7121, %7122) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7124 = tensor.empty() : tensor<1x64xf32> + %7125 = "ttir.relu"(%7123, %7124) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7126 = tensor.empty() : tensor<1x64xf32> + %7127 = "ttir.relu"(%7125, %7126) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7128 = tensor.empty() : tensor<1x64xf32> + %7129 = "ttir.relu"(%7127, %7128) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7130 = tensor.empty() : tensor<1x64xf32> + %7131 = "ttir.relu"(%7129, %7130) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7132 = tensor.empty() : tensor<1x64xf32> + %7133 = "ttir.relu"(%7131, %7132) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7134 = tensor.empty() : tensor<1x64xf32> + %7135 = "ttir.relu"(%7133, %7134) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7136 = tensor.empty() : tensor<1x64xf32> + %7137 = "ttir.relu"(%7135, %7136) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7138 = tensor.empty() : tensor<1x64xf32> + %7139 = "ttir.relu"(%7137, %7138) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7140 = tensor.empty() : tensor<1x64xf32> + %7141 = "ttir.relu"(%7139, %7140) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7142 = tensor.empty() : tensor<1x64xf32> + %7143 = "ttir.relu"(%7141, %7142) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7144 = tensor.empty() : tensor<1x64xf32> + %7145 = "ttir.relu"(%7143, %7144) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7146 = tensor.empty() : tensor<1x64xf32> + %7147 = "ttir.relu"(%7145, %7146) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7148 = tensor.empty() : tensor<1x64xf32> + %7149 = "ttir.relu"(%7147, %7148) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7150 = tensor.empty() : tensor<1x64xf32> + %7151 = "ttir.relu"(%7149, %7150) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7152 = tensor.empty() : tensor<1x64xf32> + %7153 = "ttir.relu"(%7151, %7152) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7154 = tensor.empty() : tensor<1x64xf32> + %7155 = "ttir.relu"(%7153, %7154) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7156 = tensor.empty() : tensor<1x64xf32> + %7157 = "ttir.relu"(%7155, %7156) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7158 = tensor.empty() : tensor<1x64xf32> + %7159 = "ttir.relu"(%7157, %7158) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7160 = tensor.empty() : tensor<1x64xf32> + %7161 = "ttir.relu"(%7159, %7160) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7162 = tensor.empty() : tensor<1x64xf32> + %7163 = "ttir.relu"(%7161, %7162) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7164 = tensor.empty() : tensor<1x64xf32> + %7165 = "ttir.relu"(%7163, %7164) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7166 = tensor.empty() : tensor<1x64xf32> + %7167 = "ttir.relu"(%7165, %7166) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7168 = tensor.empty() : tensor<1x64xf32> + %7169 = "ttir.relu"(%7167, %7168) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7170 = tensor.empty() : tensor<1x64xf32> + %7171 = "ttir.relu"(%7169, %7170) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7172 = tensor.empty() : tensor<1x64xf32> + %7173 = "ttir.relu"(%7171, %7172) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7174 = tensor.empty() : tensor<1x64xf32> + %7175 = "ttir.relu"(%7173, %7174) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7176 = tensor.empty() : tensor<1x64xf32> + %7177 = "ttir.relu"(%7175, %7176) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7178 = tensor.empty() : tensor<1x64xf32> + %7179 = "ttir.relu"(%7177, %7178) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7180 = tensor.empty() : tensor<1x64xf32> + %7181 = "ttir.relu"(%7179, %7180) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7182 = tensor.empty() : tensor<1x64xf32> + %7183 = "ttir.relu"(%7181, %7182) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7184 = tensor.empty() : tensor<1x64xf32> + %7185 = "ttir.relu"(%7183, %7184) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7186 = tensor.empty() : tensor<1x64xf32> + %7187 = "ttir.relu"(%7185, %7186) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7188 = tensor.empty() : tensor<1x64xf32> + %7189 = "ttir.relu"(%7187, %7188) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7190 = tensor.empty() : tensor<1x64xf32> + %7191 = "ttir.relu"(%7189, %7190) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7192 = tensor.empty() : tensor<1x64xf32> + %7193 = "ttir.relu"(%7191, %7192) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7194 = tensor.empty() : tensor<1x64xf32> + %7195 = "ttir.relu"(%7193, %7194) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7196 = tensor.empty() : tensor<1x64xf32> + %7197 = "ttir.relu"(%7195, %7196) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7198 = tensor.empty() : tensor<1x64xf32> + %7199 = "ttir.relu"(%7197, %7198) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7200 = tensor.empty() : tensor<1x64xf32> + %7201 = "ttir.relu"(%7199, %7200) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7202 = tensor.empty() : tensor<1x64xf32> + %7203 = "ttir.relu"(%7201, %7202) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7204 = tensor.empty() : tensor<1x64xf32> + %7205 = "ttir.relu"(%7203, %7204) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7206 = tensor.empty() : tensor<1x64xf32> + %7207 = "ttir.relu"(%7205, %7206) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7208 = tensor.empty() : tensor<1x64xf32> + %7209 = "ttir.relu"(%7207, %7208) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7210 = tensor.empty() : tensor<1x64xf32> + %7211 = "ttir.relu"(%7209, %7210) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7212 = tensor.empty() : tensor<1x64xf32> + %7213 = "ttir.relu"(%7211, %7212) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7214 = tensor.empty() : tensor<1x64xf32> + %7215 = "ttir.relu"(%7213, %7214) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7216 = tensor.empty() : tensor<1x64xf32> + %7217 = "ttir.relu"(%7215, %7216) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7218 = tensor.empty() : tensor<1x64xf32> + %7219 = "ttir.relu"(%7217, %7218) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7220 = tensor.empty() : tensor<1x64xf32> + %7221 = "ttir.relu"(%7219, %7220) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7222 = tensor.empty() : tensor<1x64xf32> + %7223 = "ttir.relu"(%7221, %7222) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7224 = tensor.empty() : tensor<1x64xf32> + %7225 = "ttir.relu"(%7223, %7224) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7226 = tensor.empty() : tensor<1x64xf32> + %7227 = "ttir.relu"(%7225, %7226) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7228 = tensor.empty() : tensor<1x64xf32> + %7229 = "ttir.relu"(%7227, %7228) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7230 = tensor.empty() : tensor<1x64xf32> + %7231 = "ttir.relu"(%7229, %7230) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7232 = tensor.empty() : tensor<1x64xf32> + %7233 = "ttir.relu"(%7231, %7232) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7234 = tensor.empty() : tensor<1x64xf32> + %7235 = "ttir.relu"(%7233, %7234) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7236 = tensor.empty() : tensor<1x64xf32> + %7237 = "ttir.relu"(%7235, %7236) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7238 = tensor.empty() : tensor<1x64xf32> + %7239 = "ttir.relu"(%7237, %7238) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7240 = tensor.empty() : tensor<1x64xf32> + %7241 = "ttir.relu"(%7239, %7240) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7242 = tensor.empty() : tensor<1x64xf32> + %7243 = "ttir.relu"(%7241, %7242) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7244 = tensor.empty() : tensor<1x64xf32> + %7245 = "ttir.relu"(%7243, %7244) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7246 = tensor.empty() : tensor<1x64xf32> + %7247 = "ttir.relu"(%7245, %7246) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7248 = tensor.empty() : tensor<1x64xf32> + %7249 = "ttir.relu"(%7247, %7248) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7250 = tensor.empty() : tensor<1x64xf32> + %7251 = "ttir.relu"(%7249, %7250) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7252 = tensor.empty() : tensor<1x64xf32> + %7253 = "ttir.relu"(%7251, %7252) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7254 = tensor.empty() : tensor<1x64xf32> + %7255 = "ttir.relu"(%7253, %7254) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7256 = tensor.empty() : tensor<1x64xf32> + %7257 = "ttir.relu"(%7255, %7256) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7258 = tensor.empty() : tensor<1x64xf32> + %7259 = "ttir.relu"(%7257, %7258) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7260 = tensor.empty() : tensor<1x64xf32> + %7261 = "ttir.relu"(%7259, %7260) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7262 = tensor.empty() : tensor<1x64xf32> + %7263 = "ttir.relu"(%7261, %7262) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7264 = tensor.empty() : tensor<1x64xf32> + %7265 = "ttir.relu"(%7263, %7264) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7266 = tensor.empty() : tensor<1x64xf32> + %7267 = "ttir.relu"(%7265, %7266) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7268 = tensor.empty() : tensor<1x64xf32> + %7269 = "ttir.relu"(%7267, %7268) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7270 = tensor.empty() : tensor<1x64xf32> + %7271 = "ttir.relu"(%7269, %7270) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7272 = tensor.empty() : tensor<1x64xf32> + %7273 = "ttir.relu"(%7271, %7272) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7274 = tensor.empty() : tensor<1x64xf32> + %7275 = "ttir.relu"(%7273, %7274) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7276 = tensor.empty() : tensor<1x64xf32> + %7277 = "ttir.relu"(%7275, %7276) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7278 = tensor.empty() : tensor<1x64xf32> + %7279 = "ttir.relu"(%7277, %7278) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7280 = tensor.empty() : tensor<1x64xf32> + %7281 = "ttir.relu"(%7279, %7280) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7282 = tensor.empty() : tensor<1x64xf32> + %7283 = "ttir.relu"(%7281, %7282) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7284 = tensor.empty() : tensor<1x64xf32> + %7285 = "ttir.relu"(%7283, %7284) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7286 = tensor.empty() : tensor<1x64xf32> + %7287 = "ttir.relu"(%7285, %7286) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7288 = tensor.empty() : tensor<1x64xf32> + %7289 = "ttir.relu"(%7287, %7288) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7290 = tensor.empty() : tensor<1x64xf32> + %7291 = "ttir.relu"(%7289, %7290) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7292 = tensor.empty() : tensor<1x64xf32> + %7293 = "ttir.relu"(%7291, %7292) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7294 = tensor.empty() : tensor<1x64xf32> + %7295 = "ttir.relu"(%7293, %7294) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7296 = tensor.empty() : tensor<1x64xf32> + %7297 = "ttir.relu"(%7295, %7296) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7298 = tensor.empty() : tensor<1x64xf32> + %7299 = "ttir.relu"(%7297, %7298) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7300 = tensor.empty() : tensor<1x64xf32> + %7301 = "ttir.relu"(%7299, %7300) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7302 = tensor.empty() : tensor<1x64xf32> + %7303 = "ttir.relu"(%7301, %7302) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7304 = tensor.empty() : tensor<1x64xf32> + %7305 = "ttir.relu"(%7303, %7304) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7306 = tensor.empty() : tensor<1x64xf32> + %7307 = "ttir.relu"(%7305, %7306) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7308 = tensor.empty() : tensor<1x64xf32> + %7309 = "ttir.relu"(%7307, %7308) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7310 = tensor.empty() : tensor<1x64xf32> + %7311 = "ttir.relu"(%7309, %7310) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7312 = tensor.empty() : tensor<1x64xf32> + %7313 = "ttir.relu"(%7311, %7312) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7314 = tensor.empty() : tensor<1x64xf32> + %7315 = "ttir.relu"(%7313, %7314) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7316 = tensor.empty() : tensor<1x64xf32> + %7317 = "ttir.relu"(%7315, %7316) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7318 = tensor.empty() : tensor<1x64xf32> + %7319 = "ttir.relu"(%7317, %7318) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7320 = tensor.empty() : tensor<1x64xf32> + %7321 = "ttir.relu"(%7319, %7320) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7322 = tensor.empty() : tensor<1x64xf32> + %7323 = "ttir.relu"(%7321, %7322) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7324 = tensor.empty() : tensor<1x64xf32> + %7325 = "ttir.relu"(%7323, %7324) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7326 = tensor.empty() : tensor<1x64xf32> + %7327 = "ttir.relu"(%7325, %7326) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7328 = tensor.empty() : tensor<1x64xf32> + %7329 = "ttir.relu"(%7327, %7328) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7330 = tensor.empty() : tensor<1x64xf32> + %7331 = "ttir.relu"(%7329, %7330) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7332 = tensor.empty() : tensor<1x64xf32> + %7333 = "ttir.relu"(%7331, %7332) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7334 = tensor.empty() : tensor<1x64xf32> + %7335 = "ttir.relu"(%7333, %7334) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7336 = tensor.empty() : tensor<1x64xf32> + %7337 = "ttir.relu"(%7335, %7336) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7338 = tensor.empty() : tensor<1x64xf32> + %7339 = "ttir.relu"(%7337, %7338) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7340 = tensor.empty() : tensor<1x64xf32> + %7341 = "ttir.relu"(%7339, %7340) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7342 = tensor.empty() : tensor<1x64xf32> + %7343 = "ttir.relu"(%7341, %7342) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7344 = tensor.empty() : tensor<1x64xf32> + %7345 = "ttir.relu"(%7343, %7344) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7346 = tensor.empty() : tensor<1x64xf32> + %7347 = "ttir.relu"(%7345, %7346) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7348 = tensor.empty() : tensor<1x64xf32> + %7349 = "ttir.relu"(%7347, %7348) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7350 = tensor.empty() : tensor<1x64xf32> + %7351 = "ttir.relu"(%7349, %7350) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7352 = tensor.empty() : tensor<1x64xf32> + %7353 = "ttir.relu"(%7351, %7352) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7354 = tensor.empty() : tensor<1x64xf32> + %7355 = "ttir.relu"(%7353, %7354) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7356 = tensor.empty() : tensor<1x64xf32> + %7357 = "ttir.relu"(%7355, %7356) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7358 = tensor.empty() : tensor<1x64xf32> + %7359 = "ttir.relu"(%7357, %7358) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7360 = tensor.empty() : tensor<1x64xf32> + %7361 = "ttir.relu"(%7359, %7360) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7362 = tensor.empty() : tensor<1x64xf32> + %7363 = "ttir.relu"(%7361, %7362) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7364 = tensor.empty() : tensor<1x64xf32> + %7365 = "ttir.relu"(%7363, %7364) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7366 = tensor.empty() : tensor<1x64xf32> + %7367 = "ttir.relu"(%7365, %7366) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7368 = tensor.empty() : tensor<1x64xf32> + %7369 = "ttir.relu"(%7367, %7368) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7370 = tensor.empty() : tensor<1x64xf32> + %7371 = "ttir.relu"(%7369, %7370) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7372 = tensor.empty() : tensor<1x64xf32> + %7373 = "ttir.relu"(%7371, %7372) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7374 = tensor.empty() : tensor<1x64xf32> + %7375 = "ttir.relu"(%7373, %7374) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7376 = tensor.empty() : tensor<1x64xf32> + %7377 = "ttir.relu"(%7375, %7376) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7378 = tensor.empty() : tensor<1x64xf32> + %7379 = "ttir.relu"(%7377, %7378) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7380 = tensor.empty() : tensor<1x64xf32> + %7381 = "ttir.relu"(%7379, %7380) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7382 = tensor.empty() : tensor<1x64xf32> + %7383 = "ttir.relu"(%7381, %7382) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7384 = tensor.empty() : tensor<1x64xf32> + %7385 = "ttir.relu"(%7383, %7384) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7386 = tensor.empty() : tensor<1x64xf32> + %7387 = "ttir.relu"(%7385, %7386) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7388 = tensor.empty() : tensor<1x64xf32> + %7389 = "ttir.relu"(%7387, %7388) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7390 = tensor.empty() : tensor<1x64xf32> + %7391 = "ttir.relu"(%7389, %7390) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7392 = tensor.empty() : tensor<1x64xf32> + %7393 = "ttir.relu"(%7391, %7392) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7394 = tensor.empty() : tensor<1x64xf32> + %7395 = "ttir.relu"(%7393, %7394) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7396 = tensor.empty() : tensor<1x64xf32> + %7397 = "ttir.relu"(%7395, %7396) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7398 = tensor.empty() : tensor<1x64xf32> + %7399 = "ttir.relu"(%7397, %7398) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7400 = tensor.empty() : tensor<1x64xf32> + %7401 = "ttir.relu"(%7399, %7400) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7402 = tensor.empty() : tensor<1x64xf32> + %7403 = "ttir.relu"(%7401, %7402) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7404 = tensor.empty() : tensor<1x64xf32> + %7405 = "ttir.relu"(%7403, %7404) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7406 = tensor.empty() : tensor<1x64xf32> + %7407 = "ttir.relu"(%7405, %7406) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7408 = tensor.empty() : tensor<1x64xf32> + %7409 = "ttir.relu"(%7407, %7408) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7410 = tensor.empty() : tensor<1x64xf32> + %7411 = "ttir.relu"(%7409, %7410) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7412 = tensor.empty() : tensor<1x64xf32> + %7413 = "ttir.relu"(%7411, %7412) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7414 = tensor.empty() : tensor<1x64xf32> + %7415 = "ttir.relu"(%7413, %7414) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7416 = tensor.empty() : tensor<1x64xf32> + %7417 = "ttir.relu"(%7415, %7416) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7418 = tensor.empty() : tensor<1x64xf32> + %7419 = "ttir.relu"(%7417, %7418) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7420 = tensor.empty() : tensor<1x64xf32> + %7421 = "ttir.relu"(%7419, %7420) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7422 = tensor.empty() : tensor<1x64xf32> + %7423 = "ttir.relu"(%7421, %7422) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7424 = tensor.empty() : tensor<1x64xf32> + %7425 = "ttir.relu"(%7423, %7424) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7426 = tensor.empty() : tensor<1x64xf32> + %7427 = "ttir.relu"(%7425, %7426) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7428 = tensor.empty() : tensor<1x64xf32> + %7429 = "ttir.relu"(%7427, %7428) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7430 = tensor.empty() : tensor<1x64xf32> + %7431 = "ttir.relu"(%7429, %7430) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7432 = tensor.empty() : tensor<1x64xf32> + %7433 = "ttir.relu"(%7431, %7432) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7434 = tensor.empty() : tensor<1x64xf32> + %7435 = "ttir.relu"(%7433, %7434) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7436 = tensor.empty() : tensor<1x64xf32> + %7437 = "ttir.relu"(%7435, %7436) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7438 = tensor.empty() : tensor<1x64xf32> + %7439 = "ttir.relu"(%7437, %7438) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7440 = tensor.empty() : tensor<1x64xf32> + %7441 = "ttir.relu"(%7439, %7440) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7442 = tensor.empty() : tensor<1x64xf32> + %7443 = "ttir.relu"(%7441, %7442) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7444 = tensor.empty() : tensor<1x64xf32> + %7445 = "ttir.relu"(%7443, %7444) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7446 = tensor.empty() : tensor<1x64xf32> + %7447 = "ttir.relu"(%7445, %7446) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7448 = tensor.empty() : tensor<1x64xf32> + %7449 = "ttir.relu"(%7447, %7448) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7450 = tensor.empty() : tensor<1x64xf32> + %7451 = "ttir.relu"(%7449, %7450) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7452 = tensor.empty() : tensor<1x64xf32> + %7453 = "ttir.relu"(%7451, %7452) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7454 = tensor.empty() : tensor<1x64xf32> + %7455 = "ttir.relu"(%7453, %7454) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7456 = tensor.empty() : tensor<1x64xf32> + %7457 = "ttir.relu"(%7455, %7456) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7458 = tensor.empty() : tensor<1x64xf32> + %7459 = "ttir.relu"(%7457, %7458) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7460 = tensor.empty() : tensor<1x64xf32> + %7461 = "ttir.relu"(%7459, %7460) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7462 = tensor.empty() : tensor<1x64xf32> + %7463 = "ttir.relu"(%7461, %7462) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7464 = tensor.empty() : tensor<1x64xf32> + %7465 = "ttir.relu"(%7463, %7464) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7466 = tensor.empty() : tensor<1x64xf32> + %7467 = "ttir.relu"(%7465, %7466) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7468 = tensor.empty() : tensor<1x64xf32> + %7469 = "ttir.relu"(%7467, %7468) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7470 = tensor.empty() : tensor<1x64xf32> + %7471 = "ttir.relu"(%7469, %7470) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7472 = tensor.empty() : tensor<1x64xf32> + %7473 = "ttir.relu"(%7471, %7472) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7474 = tensor.empty() : tensor<1x64xf32> + %7475 = "ttir.relu"(%7473, %7474) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7476 = tensor.empty() : tensor<1x64xf32> + %7477 = "ttir.relu"(%7475, %7476) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7478 = tensor.empty() : tensor<1x64xf32> + %7479 = "ttir.relu"(%7477, %7478) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7480 = tensor.empty() : tensor<1x64xf32> + %7481 = "ttir.relu"(%7479, %7480) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7482 = tensor.empty() : tensor<1x64xf32> + %7483 = "ttir.relu"(%7481, %7482) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7484 = tensor.empty() : tensor<1x64xf32> + %7485 = "ttir.relu"(%7483, %7484) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7486 = tensor.empty() : tensor<1x64xf32> + %7487 = "ttir.relu"(%7485, %7486) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7488 = tensor.empty() : tensor<1x64xf32> + %7489 = "ttir.relu"(%7487, %7488) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7490 = tensor.empty() : tensor<1x64xf32> + %7491 = "ttir.relu"(%7489, %7490) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7492 = tensor.empty() : tensor<1x64xf32> + %7493 = "ttir.relu"(%7491, %7492) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7494 = tensor.empty() : tensor<1x64xf32> + %7495 = "ttir.relu"(%7493, %7494) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7496 = tensor.empty() : tensor<1x64xf32> + %7497 = "ttir.relu"(%7495, %7496) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7498 = tensor.empty() : tensor<1x64xf32> + %7499 = "ttir.relu"(%7497, %7498) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7500 = tensor.empty() : tensor<1x64xf32> + %7501 = "ttir.relu"(%7499, %7500) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7502 = tensor.empty() : tensor<1x64xf32> + %7503 = "ttir.relu"(%7501, %7502) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7504 = tensor.empty() : tensor<1x64xf32> + %7505 = "ttir.relu"(%7503, %7504) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7506 = tensor.empty() : tensor<1x64xf32> + %7507 = "ttir.relu"(%7505, %7506) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7508 = tensor.empty() : tensor<1x64xf32> + %7509 = "ttir.relu"(%7507, %7508) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7510 = tensor.empty() : tensor<1x64xf32> + %7511 = "ttir.relu"(%7509, %7510) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7512 = tensor.empty() : tensor<1x64xf32> + %7513 = "ttir.relu"(%7511, %7512) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7514 = tensor.empty() : tensor<1x64xf32> + %7515 = "ttir.relu"(%7513, %7514) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7516 = tensor.empty() : tensor<1x64xf32> + %7517 = "ttir.relu"(%7515, %7516) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7518 = tensor.empty() : tensor<1x64xf32> + %7519 = "ttir.relu"(%7517, %7518) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7520 = tensor.empty() : tensor<1x64xf32> + %7521 = "ttir.relu"(%7519, %7520) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7522 = tensor.empty() : tensor<1x64xf32> + %7523 = "ttir.relu"(%7521, %7522) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7524 = tensor.empty() : tensor<1x64xf32> + %7525 = "ttir.relu"(%7523, %7524) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7526 = tensor.empty() : tensor<1x64xf32> + %7527 = "ttir.relu"(%7525, %7526) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7528 = tensor.empty() : tensor<1x64xf32> + %7529 = "ttir.relu"(%7527, %7528) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7530 = tensor.empty() : tensor<1x64xf32> + %7531 = "ttir.relu"(%7529, %7530) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7532 = tensor.empty() : tensor<1x64xf32> + %7533 = "ttir.relu"(%7531, %7532) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7534 = tensor.empty() : tensor<1x64xf32> + %7535 = "ttir.relu"(%7533, %7534) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7536 = tensor.empty() : tensor<1x64xf32> + %7537 = "ttir.relu"(%7535, %7536) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7538 = tensor.empty() : tensor<1x64xf32> + %7539 = "ttir.relu"(%7537, %7538) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7540 = tensor.empty() : tensor<1x64xf32> + %7541 = "ttir.relu"(%7539, %7540) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7542 = tensor.empty() : tensor<1x64xf32> + %7543 = "ttir.relu"(%7541, %7542) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7544 = tensor.empty() : tensor<1x64xf32> + %7545 = "ttir.relu"(%7543, %7544) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7546 = tensor.empty() : tensor<1x64xf32> + %7547 = "ttir.relu"(%7545, %7546) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7548 = tensor.empty() : tensor<1x64xf32> + %7549 = "ttir.relu"(%7547, %7548) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7550 = tensor.empty() : tensor<1x64xf32> + %7551 = "ttir.relu"(%7549, %7550) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7552 = tensor.empty() : tensor<1x64xf32> + %7553 = "ttir.relu"(%7551, %7552) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7554 = tensor.empty() : tensor<1x64xf32> + %7555 = "ttir.relu"(%7553, %7554) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7556 = tensor.empty() : tensor<1x64xf32> + %7557 = "ttir.relu"(%7555, %7556) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7558 = tensor.empty() : tensor<1x64xf32> + %7559 = "ttir.relu"(%7557, %7558) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7560 = tensor.empty() : tensor<1x64xf32> + %7561 = "ttir.relu"(%7559, %7560) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7562 = tensor.empty() : tensor<1x64xf32> + %7563 = "ttir.relu"(%7561, %7562) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7564 = tensor.empty() : tensor<1x64xf32> + %7565 = "ttir.relu"(%7563, %7564) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7566 = tensor.empty() : tensor<1x64xf32> + %7567 = "ttir.relu"(%7565, %7566) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7568 = tensor.empty() : tensor<1x64xf32> + %7569 = "ttir.relu"(%7567, %7568) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7570 = tensor.empty() : tensor<1x64xf32> + %7571 = "ttir.relu"(%7569, %7570) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7572 = tensor.empty() : tensor<1x64xf32> + %7573 = "ttir.relu"(%7571, %7572) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7574 = tensor.empty() : tensor<1x64xf32> + %7575 = "ttir.relu"(%7573, %7574) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7576 = tensor.empty() : tensor<1x64xf32> + %7577 = "ttir.relu"(%7575, %7576) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7578 = tensor.empty() : tensor<1x64xf32> + %7579 = "ttir.relu"(%7577, %7578) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7580 = tensor.empty() : tensor<1x64xf32> + %7581 = "ttir.relu"(%7579, %7580) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7582 = tensor.empty() : tensor<1x64xf32> + %7583 = "ttir.relu"(%7581, %7582) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7584 = tensor.empty() : tensor<1x64xf32> + %7585 = "ttir.relu"(%7583, %7584) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7586 = tensor.empty() : tensor<1x64xf32> + %7587 = "ttir.relu"(%7585, %7586) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7588 = tensor.empty() : tensor<1x64xf32> + %7589 = "ttir.relu"(%7587, %7588) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7590 = tensor.empty() : tensor<1x64xf32> + %7591 = "ttir.relu"(%7589, %7590) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7592 = tensor.empty() : tensor<1x64xf32> + %7593 = "ttir.relu"(%7591, %7592) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7594 = tensor.empty() : tensor<1x64xf32> + %7595 = "ttir.relu"(%7593, %7594) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7596 = tensor.empty() : tensor<1x64xf32> + %7597 = "ttir.relu"(%7595, %7596) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7598 = tensor.empty() : tensor<1x64xf32> + %7599 = "ttir.relu"(%7597, %7598) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7600 = tensor.empty() : tensor<1x64xf32> + %7601 = "ttir.relu"(%7599, %7600) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7602 = tensor.empty() : tensor<1x64xf32> + %7603 = "ttir.relu"(%7601, %7602) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7604 = tensor.empty() : tensor<1x64xf32> + %7605 = "ttir.relu"(%7603, %7604) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7606 = tensor.empty() : tensor<1x64xf32> + %7607 = "ttir.relu"(%7605, %7606) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7608 = tensor.empty() : tensor<1x64xf32> + %7609 = "ttir.relu"(%7607, %7608) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7610 = tensor.empty() : tensor<1x64xf32> + %7611 = "ttir.relu"(%7609, %7610) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7612 = tensor.empty() : tensor<1x64xf32> + %7613 = "ttir.relu"(%7611, %7612) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7614 = tensor.empty() : tensor<1x64xf32> + %7615 = "ttir.relu"(%7613, %7614) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7616 = tensor.empty() : tensor<1x64xf32> + %7617 = "ttir.relu"(%7615, %7616) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7618 = tensor.empty() : tensor<1x64xf32> + %7619 = "ttir.relu"(%7617, %7618) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7620 = tensor.empty() : tensor<1x64xf32> + %7621 = "ttir.relu"(%7619, %7620) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7622 = tensor.empty() : tensor<1x64xf32> + %7623 = "ttir.relu"(%7621, %7622) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7624 = tensor.empty() : tensor<1x64xf32> + %7625 = "ttir.relu"(%7623, %7624) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7626 = tensor.empty() : tensor<1x64xf32> + %7627 = "ttir.relu"(%7625, %7626) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7628 = tensor.empty() : tensor<1x64xf32> + %7629 = "ttir.relu"(%7627, %7628) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7630 = tensor.empty() : tensor<1x64xf32> + %7631 = "ttir.relu"(%7629, %7630) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7632 = tensor.empty() : tensor<1x64xf32> + %7633 = "ttir.relu"(%7631, %7632) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7634 = tensor.empty() : tensor<1x64xf32> + %7635 = "ttir.relu"(%7633, %7634) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7636 = tensor.empty() : tensor<1x64xf32> + %7637 = "ttir.relu"(%7635, %7636) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7638 = tensor.empty() : tensor<1x64xf32> + %7639 = "ttir.relu"(%7637, %7638) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7640 = tensor.empty() : tensor<1x64xf32> + %7641 = "ttir.relu"(%7639, %7640) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7642 = tensor.empty() : tensor<1x64xf32> + %7643 = "ttir.relu"(%7641, %7642) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7644 = tensor.empty() : tensor<1x64xf32> + %7645 = "ttir.relu"(%7643, %7644) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7646 = tensor.empty() : tensor<1x64xf32> + %7647 = "ttir.relu"(%7645, %7646) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7648 = tensor.empty() : tensor<1x64xf32> + %7649 = "ttir.relu"(%7647, %7648) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7650 = tensor.empty() : tensor<1x64xf32> + %7651 = "ttir.relu"(%7649, %7650) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7652 = tensor.empty() : tensor<1x64xf32> + %7653 = "ttir.relu"(%7651, %7652) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7654 = tensor.empty() : tensor<1x64xf32> + %7655 = "ttir.relu"(%7653, %7654) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7656 = tensor.empty() : tensor<1x64xf32> + %7657 = "ttir.relu"(%7655, %7656) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7658 = tensor.empty() : tensor<1x64xf32> + %7659 = "ttir.relu"(%7657, %7658) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7660 = tensor.empty() : tensor<1x64xf32> + %7661 = "ttir.relu"(%7659, %7660) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7662 = tensor.empty() : tensor<1x64xf32> + %7663 = "ttir.relu"(%7661, %7662) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7664 = tensor.empty() : tensor<1x64xf32> + %7665 = "ttir.relu"(%7663, %7664) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7666 = tensor.empty() : tensor<1x64xf32> + %7667 = "ttir.relu"(%7665, %7666) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7668 = tensor.empty() : tensor<1x64xf32> + %7669 = "ttir.relu"(%7667, %7668) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7670 = tensor.empty() : tensor<1x64xf32> + %7671 = "ttir.relu"(%7669, %7670) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7672 = tensor.empty() : tensor<1x64xf32> + %7673 = "ttir.relu"(%7671, %7672) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7674 = tensor.empty() : tensor<1x64xf32> + %7675 = "ttir.relu"(%7673, %7674) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7676 = tensor.empty() : tensor<1x64xf32> + %7677 = "ttir.relu"(%7675, %7676) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7678 = tensor.empty() : tensor<1x64xf32> + %7679 = "ttir.relu"(%7677, %7678) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7680 = tensor.empty() : tensor<1x64xf32> + %7681 = "ttir.relu"(%7679, %7680) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7682 = tensor.empty() : tensor<1x64xf32> + %7683 = "ttir.relu"(%7681, %7682) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7684 = tensor.empty() : tensor<1x64xf32> + %7685 = "ttir.relu"(%7683, %7684) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7686 = tensor.empty() : tensor<1x64xf32> + %7687 = "ttir.relu"(%7685, %7686) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7688 = tensor.empty() : tensor<1x64xf32> + %7689 = "ttir.relu"(%7687, %7688) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7690 = tensor.empty() : tensor<1x64xf32> + %7691 = "ttir.relu"(%7689, %7690) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7692 = tensor.empty() : tensor<1x64xf32> + %7693 = "ttir.relu"(%7691, %7692) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7694 = tensor.empty() : tensor<1x64xf32> + %7695 = "ttir.relu"(%7693, %7694) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7696 = tensor.empty() : tensor<1x64xf32> + %7697 = "ttir.relu"(%7695, %7696) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7698 = tensor.empty() : tensor<1x64xf32> + %7699 = "ttir.relu"(%7697, %7698) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7700 = tensor.empty() : tensor<1x64xf32> + %7701 = "ttir.relu"(%7699, %7700) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7702 = tensor.empty() : tensor<1x64xf32> + %7703 = "ttir.relu"(%7701, %7702) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7704 = tensor.empty() : tensor<1x64xf32> + %7705 = "ttir.relu"(%7703, %7704) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7706 = tensor.empty() : tensor<1x64xf32> + %7707 = "ttir.relu"(%7705, %7706) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7708 = tensor.empty() : tensor<1x64xf32> + %7709 = "ttir.relu"(%7707, %7708) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7710 = tensor.empty() : tensor<1x64xf32> + %7711 = "ttir.relu"(%7709, %7710) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7712 = tensor.empty() : tensor<1x64xf32> + %7713 = "ttir.relu"(%7711, %7712) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7714 = tensor.empty() : tensor<1x64xf32> + %7715 = "ttir.relu"(%7713, %7714) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7716 = tensor.empty() : tensor<1x64xf32> + %7717 = "ttir.relu"(%7715, %7716) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7718 = tensor.empty() : tensor<1x64xf32> + %7719 = "ttir.relu"(%7717, %7718) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7720 = tensor.empty() : tensor<1x64xf32> + %7721 = "ttir.relu"(%7719, %7720) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7722 = tensor.empty() : tensor<1x64xf32> + %7723 = "ttir.relu"(%7721, %7722) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7724 = tensor.empty() : tensor<1x64xf32> + %7725 = "ttir.relu"(%7723, %7724) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7726 = tensor.empty() : tensor<1x64xf32> + %7727 = "ttir.relu"(%7725, %7726) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7728 = tensor.empty() : tensor<1x64xf32> + %7729 = "ttir.relu"(%7727, %7728) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7730 = tensor.empty() : tensor<1x64xf32> + %7731 = "ttir.relu"(%7729, %7730) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7732 = tensor.empty() : tensor<1x64xf32> + %7733 = "ttir.relu"(%7731, %7732) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7734 = tensor.empty() : tensor<1x64xf32> + %7735 = "ttir.relu"(%7733, %7734) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7736 = tensor.empty() : tensor<1x64xf32> + %7737 = "ttir.relu"(%7735, %7736) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7738 = tensor.empty() : tensor<1x64xf32> + %7739 = "ttir.relu"(%7737, %7738) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7740 = tensor.empty() : tensor<1x64xf32> + %7741 = "ttir.relu"(%7739, %7740) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7742 = tensor.empty() : tensor<1x64xf32> + %7743 = "ttir.relu"(%7741, %7742) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7744 = tensor.empty() : tensor<1x64xf32> + %7745 = "ttir.relu"(%7743, %7744) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7746 = tensor.empty() : tensor<1x64xf32> + %7747 = "ttir.relu"(%7745, %7746) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7748 = tensor.empty() : tensor<1x64xf32> + %7749 = "ttir.relu"(%7747, %7748) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7750 = tensor.empty() : tensor<1x64xf32> + %7751 = "ttir.relu"(%7749, %7750) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7752 = tensor.empty() : tensor<1x64xf32> + %7753 = "ttir.relu"(%7751, %7752) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7754 = tensor.empty() : tensor<1x64xf32> + %7755 = "ttir.relu"(%7753, %7754) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7756 = tensor.empty() : tensor<1x64xf32> + %7757 = "ttir.relu"(%7755, %7756) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7758 = tensor.empty() : tensor<1x64xf32> + %7759 = "ttir.relu"(%7757, %7758) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7760 = tensor.empty() : tensor<1x64xf32> + %7761 = "ttir.relu"(%7759, %7760) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7762 = tensor.empty() : tensor<1x64xf32> + %7763 = "ttir.relu"(%7761, %7762) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7764 = tensor.empty() : tensor<1x64xf32> + %7765 = "ttir.relu"(%7763, %7764) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7766 = tensor.empty() : tensor<1x64xf32> + %7767 = "ttir.relu"(%7765, %7766) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7768 = tensor.empty() : tensor<1x64xf32> + %7769 = "ttir.relu"(%7767, %7768) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7770 = tensor.empty() : tensor<1x64xf32> + %7771 = "ttir.relu"(%7769, %7770) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7772 = tensor.empty() : tensor<1x64xf32> + %7773 = "ttir.relu"(%7771, %7772) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7774 = tensor.empty() : tensor<1x64xf32> + %7775 = "ttir.relu"(%7773, %7774) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7776 = tensor.empty() : tensor<1x64xf32> + %7777 = "ttir.relu"(%7775, %7776) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7778 = tensor.empty() : tensor<1x64xf32> + %7779 = "ttir.relu"(%7777, %7778) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7780 = tensor.empty() : tensor<1x64xf32> + %7781 = "ttir.relu"(%7779, %7780) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7782 = tensor.empty() : tensor<1x64xf32> + %7783 = "ttir.relu"(%7781, %7782) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7784 = tensor.empty() : tensor<1x64xf32> + %7785 = "ttir.relu"(%7783, %7784) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7786 = tensor.empty() : tensor<1x64xf32> + %7787 = "ttir.relu"(%7785, %7786) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7788 = tensor.empty() : tensor<1x64xf32> + %7789 = "ttir.relu"(%7787, %7788) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7790 = tensor.empty() : tensor<1x64xf32> + %7791 = "ttir.relu"(%7789, %7790) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7792 = tensor.empty() : tensor<1x64xf32> + %7793 = "ttir.relu"(%7791, %7792) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7794 = tensor.empty() : tensor<1x64xf32> + %7795 = "ttir.relu"(%7793, %7794) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7796 = tensor.empty() : tensor<1x64xf32> + %7797 = "ttir.relu"(%7795, %7796) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7798 = tensor.empty() : tensor<1x64xf32> + %7799 = "ttir.relu"(%7797, %7798) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7800 = tensor.empty() : tensor<1x64xf32> + %7801 = "ttir.relu"(%7799, %7800) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7802 = tensor.empty() : tensor<1x64xf32> + %7803 = "ttir.relu"(%7801, %7802) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7804 = tensor.empty() : tensor<1x64xf32> + %7805 = "ttir.relu"(%7803, %7804) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7806 = tensor.empty() : tensor<1x64xf32> + %7807 = "ttir.relu"(%7805, %7806) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7808 = tensor.empty() : tensor<1x64xf32> + %7809 = "ttir.relu"(%7807, %7808) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7810 = tensor.empty() : tensor<1x64xf32> + %7811 = "ttir.relu"(%7809, %7810) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7812 = tensor.empty() : tensor<1x64xf32> + %7813 = "ttir.relu"(%7811, %7812) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7814 = tensor.empty() : tensor<1x64xf32> + %7815 = "ttir.relu"(%7813, %7814) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7816 = tensor.empty() : tensor<1x64xf32> + %7817 = "ttir.relu"(%7815, %7816) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7818 = tensor.empty() : tensor<1x64xf32> + %7819 = "ttir.relu"(%7817, %7818) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7820 = tensor.empty() : tensor<1x64xf32> + %7821 = "ttir.relu"(%7819, %7820) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7822 = tensor.empty() : tensor<1x64xf32> + %7823 = "ttir.relu"(%7821, %7822) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7824 = tensor.empty() : tensor<1x64xf32> + %7825 = "ttir.relu"(%7823, %7824) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7826 = tensor.empty() : tensor<1x64xf32> + %7827 = "ttir.relu"(%7825, %7826) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7828 = tensor.empty() : tensor<1x64xf32> + %7829 = "ttir.relu"(%7827, %7828) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7830 = tensor.empty() : tensor<1x64xf32> + %7831 = "ttir.relu"(%7829, %7830) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7832 = tensor.empty() : tensor<1x64xf32> + %7833 = "ttir.relu"(%7831, %7832) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7834 = tensor.empty() : tensor<1x64xf32> + %7835 = "ttir.relu"(%7833, %7834) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7836 = tensor.empty() : tensor<1x64xf32> + %7837 = "ttir.relu"(%7835, %7836) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7838 = tensor.empty() : tensor<1x64xf32> + %7839 = "ttir.relu"(%7837, %7838) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7840 = tensor.empty() : tensor<1x64xf32> + %7841 = "ttir.relu"(%7839, %7840) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7842 = tensor.empty() : tensor<1x64xf32> + %7843 = "ttir.relu"(%7841, %7842) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7844 = tensor.empty() : tensor<1x64xf32> + %7845 = "ttir.relu"(%7843, %7844) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7846 = tensor.empty() : tensor<1x64xf32> + %7847 = "ttir.relu"(%7845, %7846) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7848 = tensor.empty() : tensor<1x64xf32> + %7849 = "ttir.relu"(%7847, %7848) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7850 = tensor.empty() : tensor<1x64xf32> + %7851 = "ttir.relu"(%7849, %7850) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7852 = tensor.empty() : tensor<1x64xf32> + %7853 = "ttir.relu"(%7851, %7852) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7854 = tensor.empty() : tensor<1x64xf32> + %7855 = "ttir.relu"(%7853, %7854) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7856 = tensor.empty() : tensor<1x64xf32> + %7857 = "ttir.relu"(%7855, %7856) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7858 = tensor.empty() : tensor<1x64xf32> + %7859 = "ttir.relu"(%7857, %7858) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7860 = tensor.empty() : tensor<1x64xf32> + %7861 = "ttir.relu"(%7859, %7860) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7862 = tensor.empty() : tensor<1x64xf32> + %7863 = "ttir.relu"(%7861, %7862) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7864 = tensor.empty() : tensor<1x64xf32> + %7865 = "ttir.relu"(%7863, %7864) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7866 = tensor.empty() : tensor<1x64xf32> + %7867 = "ttir.relu"(%7865, %7866) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7868 = tensor.empty() : tensor<1x64xf32> + %7869 = "ttir.relu"(%7867, %7868) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7870 = tensor.empty() : tensor<1x64xf32> + %7871 = "ttir.relu"(%7869, %7870) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7872 = tensor.empty() : tensor<1x64xf32> + %7873 = "ttir.relu"(%7871, %7872) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7874 = tensor.empty() : tensor<1x64xf32> + %7875 = "ttir.relu"(%7873, %7874) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7876 = tensor.empty() : tensor<1x64xf32> + %7877 = "ttir.relu"(%7875, %7876) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7878 = tensor.empty() : tensor<1x64xf32> + %7879 = "ttir.relu"(%7877, %7878) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7880 = tensor.empty() : tensor<1x64xf32> + %7881 = "ttir.relu"(%7879, %7880) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7882 = tensor.empty() : tensor<1x64xf32> + %7883 = "ttir.relu"(%7881, %7882) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7884 = tensor.empty() : tensor<1x64xf32> + %7885 = "ttir.relu"(%7883, %7884) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7886 = tensor.empty() : tensor<1x64xf32> + %7887 = "ttir.relu"(%7885, %7886) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7888 = tensor.empty() : tensor<1x64xf32> + %7889 = "ttir.relu"(%7887, %7888) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7890 = tensor.empty() : tensor<1x64xf32> + %7891 = "ttir.relu"(%7889, %7890) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7892 = tensor.empty() : tensor<1x64xf32> + %7893 = "ttir.relu"(%7891, %7892) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7894 = tensor.empty() : tensor<1x64xf32> + %7895 = "ttir.relu"(%7893, %7894) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7896 = tensor.empty() : tensor<1x64xf32> + %7897 = "ttir.relu"(%7895, %7896) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7898 = tensor.empty() : tensor<1x64xf32> + %7899 = "ttir.relu"(%7897, %7898) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7900 = tensor.empty() : tensor<1x64xf32> + %7901 = "ttir.relu"(%7899, %7900) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7902 = tensor.empty() : tensor<1x64xf32> + %7903 = "ttir.relu"(%7901, %7902) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7904 = tensor.empty() : tensor<1x64xf32> + %7905 = "ttir.relu"(%7903, %7904) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7906 = tensor.empty() : tensor<1x64xf32> + %7907 = "ttir.relu"(%7905, %7906) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7908 = tensor.empty() : tensor<1x64xf32> + %7909 = "ttir.relu"(%7907, %7908) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7910 = tensor.empty() : tensor<1x64xf32> + %7911 = "ttir.relu"(%7909, %7910) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7912 = tensor.empty() : tensor<1x64xf32> + %7913 = "ttir.relu"(%7911, %7912) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7914 = tensor.empty() : tensor<1x64xf32> + %7915 = "ttir.relu"(%7913, %7914) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7916 = tensor.empty() : tensor<1x64xf32> + %7917 = "ttir.relu"(%7915, %7916) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7918 = tensor.empty() : tensor<1x64xf32> + %7919 = "ttir.relu"(%7917, %7918) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7920 = tensor.empty() : tensor<1x64xf32> + %7921 = "ttir.relu"(%7919, %7920) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7922 = tensor.empty() : tensor<1x64xf32> + %7923 = "ttir.relu"(%7921, %7922) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7924 = tensor.empty() : tensor<1x64xf32> + %7925 = "ttir.relu"(%7923, %7924) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7926 = tensor.empty() : tensor<1x64xf32> + %7927 = "ttir.relu"(%7925, %7926) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7928 = tensor.empty() : tensor<1x64xf32> + %7929 = "ttir.relu"(%7927, %7928) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7930 = tensor.empty() : tensor<1x64xf32> + %7931 = "ttir.relu"(%7929, %7930) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7932 = tensor.empty() : tensor<1x64xf32> + %7933 = "ttir.relu"(%7931, %7932) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7934 = tensor.empty() : tensor<1x64xf32> + %7935 = "ttir.relu"(%7933, %7934) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7936 = tensor.empty() : tensor<1x64xf32> + %7937 = "ttir.relu"(%7935, %7936) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7938 = tensor.empty() : tensor<1x64xf32> + %7939 = "ttir.relu"(%7937, %7938) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7940 = tensor.empty() : tensor<1x64xf32> + %7941 = "ttir.relu"(%7939, %7940) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7942 = tensor.empty() : tensor<1x64xf32> + %7943 = "ttir.relu"(%7941, %7942) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7944 = tensor.empty() : tensor<1x64xf32> + %7945 = "ttir.relu"(%7943, %7944) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7946 = tensor.empty() : tensor<1x64xf32> + %7947 = "ttir.relu"(%7945, %7946) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7948 = tensor.empty() : tensor<1x64xf32> + %7949 = "ttir.relu"(%7947, %7948) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7950 = tensor.empty() : tensor<1x64xf32> + %7951 = "ttir.relu"(%7949, %7950) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7952 = tensor.empty() : tensor<1x64xf32> + %7953 = "ttir.relu"(%7951, %7952) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7954 = tensor.empty() : tensor<1x64xf32> + %7955 = "ttir.relu"(%7953, %7954) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7956 = tensor.empty() : tensor<1x64xf32> + %7957 = "ttir.relu"(%7955, %7956) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7958 = tensor.empty() : tensor<1x64xf32> + %7959 = "ttir.relu"(%7957, %7958) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7960 = tensor.empty() : tensor<1x64xf32> + %7961 = "ttir.relu"(%7959, %7960) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7962 = tensor.empty() : tensor<1x64xf32> + %7963 = "ttir.relu"(%7961, %7962) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7964 = tensor.empty() : tensor<1x64xf32> + %7965 = "ttir.relu"(%7963, %7964) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7966 = tensor.empty() : tensor<1x64xf32> + %7967 = "ttir.relu"(%7965, %7966) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7968 = tensor.empty() : tensor<1x64xf32> + %7969 = "ttir.relu"(%7967, %7968) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7970 = tensor.empty() : tensor<1x64xf32> + %7971 = "ttir.relu"(%7969, %7970) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7972 = tensor.empty() : tensor<1x64xf32> + %7973 = "ttir.relu"(%7971, %7972) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7974 = tensor.empty() : tensor<1x64xf32> + %7975 = "ttir.relu"(%7973, %7974) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7976 = tensor.empty() : tensor<1x64xf32> + %7977 = "ttir.relu"(%7975, %7976) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7978 = tensor.empty() : tensor<1x64xf32> + %7979 = "ttir.relu"(%7977, %7978) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7980 = tensor.empty() : tensor<1x64xf32> + %7981 = "ttir.relu"(%7979, %7980) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7982 = tensor.empty() : tensor<1x64xf32> + %7983 = "ttir.relu"(%7981, %7982) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7984 = tensor.empty() : tensor<1x64xf32> + %7985 = "ttir.relu"(%7983, %7984) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7986 = tensor.empty() : tensor<1x64xf32> + %7987 = "ttir.relu"(%7985, %7986) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7988 = tensor.empty() : tensor<1x64xf32> + %7989 = "ttir.relu"(%7987, %7988) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7990 = tensor.empty() : tensor<1x64xf32> + %7991 = "ttir.relu"(%7989, %7990) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7992 = tensor.empty() : tensor<1x64xf32> + %7993 = "ttir.relu"(%7991, %7992) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7994 = tensor.empty() : tensor<1x64xf32> + %7995 = "ttir.relu"(%7993, %7994) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7996 = tensor.empty() : tensor<1x64xf32> + %7997 = "ttir.relu"(%7995, %7996) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %7998 = tensor.empty() : tensor<1x64xf32> + %7999 = "ttir.relu"(%7997, %7998) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8000 = tensor.empty() : tensor<1x64xf32> + %8001 = "ttir.relu"(%7999, %8000) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8002 = tensor.empty() : tensor<1x64xf32> + %8003 = "ttir.relu"(%8001, %8002) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8004 = tensor.empty() : tensor<1x64xf32> + %8005 = "ttir.relu"(%8003, %8004) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8006 = tensor.empty() : tensor<1x64xf32> + %8007 = "ttir.relu"(%8005, %8006) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8008 = tensor.empty() : tensor<1x64xf32> + %8009 = "ttir.relu"(%8007, %8008) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8010 = tensor.empty() : tensor<1x64xf32> + %8011 = "ttir.relu"(%8009, %8010) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8012 = tensor.empty() : tensor<1x64xf32> + %8013 = "ttir.relu"(%8011, %8012) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8014 = tensor.empty() : tensor<1x64xf32> + %8015 = "ttir.relu"(%8013, %8014) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8016 = tensor.empty() : tensor<1x64xf32> + %8017 = "ttir.relu"(%8015, %8016) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8018 = tensor.empty() : tensor<1x64xf32> + %8019 = "ttir.relu"(%8017, %8018) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8020 = tensor.empty() : tensor<1x64xf32> + %8021 = "ttir.relu"(%8019, %8020) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8022 = tensor.empty() : tensor<1x64xf32> + %8023 = "ttir.relu"(%8021, %8022) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8024 = tensor.empty() : tensor<1x64xf32> + %8025 = "ttir.relu"(%8023, %8024) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8026 = tensor.empty() : tensor<1x64xf32> + %8027 = "ttir.relu"(%8025, %8026) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8028 = tensor.empty() : tensor<1x64xf32> + %8029 = "ttir.relu"(%8027, %8028) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8030 = tensor.empty() : tensor<1x64xf32> + %8031 = "ttir.relu"(%8029, %8030) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8032 = tensor.empty() : tensor<1x64xf32> + %8033 = "ttir.relu"(%8031, %8032) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8034 = tensor.empty() : tensor<1x64xf32> + %8035 = "ttir.relu"(%8033, %8034) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8036 = tensor.empty() : tensor<1x64xf32> + %8037 = "ttir.relu"(%8035, %8036) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8038 = tensor.empty() : tensor<1x64xf32> + %8039 = "ttir.relu"(%8037, %8038) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8040 = tensor.empty() : tensor<1x64xf32> + %8041 = "ttir.relu"(%8039, %8040) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8042 = tensor.empty() : tensor<1x64xf32> + %8043 = "ttir.relu"(%8041, %8042) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8044 = tensor.empty() : tensor<1x64xf32> + %8045 = "ttir.relu"(%8043, %8044) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8046 = tensor.empty() : tensor<1x64xf32> + %8047 = "ttir.relu"(%8045, %8046) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8048 = tensor.empty() : tensor<1x64xf32> + %8049 = "ttir.relu"(%8047, %8048) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8050 = tensor.empty() : tensor<1x64xf32> + %8051 = "ttir.relu"(%8049, %8050) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8052 = tensor.empty() : tensor<1x64xf32> + %8053 = "ttir.relu"(%8051, %8052) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8054 = tensor.empty() : tensor<1x64xf32> + %8055 = "ttir.relu"(%8053, %8054) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8056 = tensor.empty() : tensor<1x64xf32> + %8057 = "ttir.relu"(%8055, %8056) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8058 = tensor.empty() : tensor<1x64xf32> + %8059 = "ttir.relu"(%8057, %8058) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8060 = tensor.empty() : tensor<1x64xf32> + %8061 = "ttir.relu"(%8059, %8060) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8062 = tensor.empty() : tensor<1x64xf32> + %8063 = "ttir.relu"(%8061, %8062) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8064 = tensor.empty() : tensor<1x64xf32> + %8065 = "ttir.relu"(%8063, %8064) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8066 = tensor.empty() : tensor<1x64xf32> + %8067 = "ttir.relu"(%8065, %8066) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8068 = tensor.empty() : tensor<1x64xf32> + %8069 = "ttir.relu"(%8067, %8068) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8070 = tensor.empty() : tensor<1x64xf32> + %8071 = "ttir.relu"(%8069, %8070) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8072 = tensor.empty() : tensor<1x64xf32> + %8073 = "ttir.relu"(%8071, %8072) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8074 = tensor.empty() : tensor<1x64xf32> + %8075 = "ttir.relu"(%8073, %8074) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8076 = tensor.empty() : tensor<1x64xf32> + %8077 = "ttir.relu"(%8075, %8076) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8078 = tensor.empty() : tensor<1x64xf32> + %8079 = "ttir.relu"(%8077, %8078) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8080 = tensor.empty() : tensor<1x64xf32> + %8081 = "ttir.relu"(%8079, %8080) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8082 = tensor.empty() : tensor<1x64xf32> + %8083 = "ttir.relu"(%8081, %8082) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8084 = tensor.empty() : tensor<1x64xf32> + %8085 = "ttir.relu"(%8083, %8084) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8086 = tensor.empty() : tensor<1x64xf32> + %8087 = "ttir.relu"(%8085, %8086) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8088 = tensor.empty() : tensor<1x64xf32> + %8089 = "ttir.relu"(%8087, %8088) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8090 = tensor.empty() : tensor<1x64xf32> + %8091 = "ttir.relu"(%8089, %8090) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8092 = tensor.empty() : tensor<1x64xf32> + %8093 = "ttir.relu"(%8091, %8092) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8094 = tensor.empty() : tensor<1x64xf32> + %8095 = "ttir.relu"(%8093, %8094) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8096 = tensor.empty() : tensor<1x64xf32> + %8097 = "ttir.relu"(%8095, %8096) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8098 = tensor.empty() : tensor<1x64xf32> + %8099 = "ttir.relu"(%8097, %8098) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8100 = tensor.empty() : tensor<1x64xf32> + %8101 = "ttir.relu"(%8099, %8100) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8102 = tensor.empty() : tensor<1x64xf32> + %8103 = "ttir.relu"(%8101, %8102) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8104 = tensor.empty() : tensor<1x64xf32> + %8105 = "ttir.relu"(%8103, %8104) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8106 = tensor.empty() : tensor<1x64xf32> + %8107 = "ttir.relu"(%8105, %8106) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8108 = tensor.empty() : tensor<1x64xf32> + %8109 = "ttir.relu"(%8107, %8108) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8110 = tensor.empty() : tensor<1x64xf32> + %8111 = "ttir.relu"(%8109, %8110) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8112 = tensor.empty() : tensor<1x64xf32> + %8113 = "ttir.relu"(%8111, %8112) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8114 = tensor.empty() : tensor<1x64xf32> + %8115 = "ttir.relu"(%8113, %8114) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8116 = tensor.empty() : tensor<1x64xf32> + %8117 = "ttir.relu"(%8115, %8116) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8118 = tensor.empty() : tensor<1x64xf32> + %8119 = "ttir.relu"(%8117, %8118) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8120 = tensor.empty() : tensor<1x64xf32> + %8121 = "ttir.relu"(%8119, %8120) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8122 = tensor.empty() : tensor<1x64xf32> + %8123 = "ttir.relu"(%8121, %8122) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8124 = tensor.empty() : tensor<1x64xf32> + %8125 = "ttir.relu"(%8123, %8124) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8126 = tensor.empty() : tensor<1x64xf32> + %8127 = "ttir.relu"(%8125, %8126) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8128 = tensor.empty() : tensor<1x64xf32> + %8129 = "ttir.relu"(%8127, %8128) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8130 = tensor.empty() : tensor<1x64xf32> + %8131 = "ttir.relu"(%8129, %8130) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8132 = tensor.empty() : tensor<1x64xf32> + %8133 = "ttir.relu"(%8131, %8132) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8134 = tensor.empty() : tensor<1x64xf32> + %8135 = "ttir.relu"(%8133, %8134) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8136 = tensor.empty() : tensor<1x64xf32> + %8137 = "ttir.relu"(%8135, %8136) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8138 = tensor.empty() : tensor<1x64xf32> + %8139 = "ttir.relu"(%8137, %8138) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8140 = tensor.empty() : tensor<1x64xf32> + %8141 = "ttir.relu"(%8139, %8140) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8142 = tensor.empty() : tensor<1x64xf32> + %8143 = "ttir.relu"(%8141, %8142) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8144 = tensor.empty() : tensor<1x64xf32> + %8145 = "ttir.relu"(%8143, %8144) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8146 = tensor.empty() : tensor<1x64xf32> + %8147 = "ttir.relu"(%8145, %8146) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8148 = tensor.empty() : tensor<1x64xf32> + %8149 = "ttir.relu"(%8147, %8148) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8150 = tensor.empty() : tensor<1x64xf32> + %8151 = "ttir.relu"(%8149, %8150) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8152 = tensor.empty() : tensor<1x64xf32> + %8153 = "ttir.relu"(%8151, %8152) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8154 = tensor.empty() : tensor<1x64xf32> + %8155 = "ttir.relu"(%8153, %8154) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8156 = tensor.empty() : tensor<1x64xf32> + %8157 = "ttir.relu"(%8155, %8156) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8158 = tensor.empty() : tensor<1x64xf32> + %8159 = "ttir.relu"(%8157, %8158) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8160 = tensor.empty() : tensor<1x64xf32> + %8161 = "ttir.relu"(%8159, %8160) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8162 = tensor.empty() : tensor<1x64xf32> + %8163 = "ttir.relu"(%8161, %8162) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8164 = tensor.empty() : tensor<1x64xf32> + %8165 = "ttir.relu"(%8163, %8164) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8166 = tensor.empty() : tensor<1x64xf32> + %8167 = "ttir.relu"(%8165, %8166) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8168 = tensor.empty() : tensor<1x64xf32> + %8169 = "ttir.relu"(%8167, %8168) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8170 = tensor.empty() : tensor<1x64xf32> + %8171 = "ttir.relu"(%8169, %8170) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8172 = tensor.empty() : tensor<1x64xf32> + %8173 = "ttir.relu"(%8171, %8172) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8174 = tensor.empty() : tensor<1x64xf32> + %8175 = "ttir.relu"(%8173, %8174) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8176 = tensor.empty() : tensor<1x64xf32> + %8177 = "ttir.relu"(%8175, %8176) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8178 = tensor.empty() : tensor<1x64xf32> + %8179 = "ttir.relu"(%8177, %8178) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8180 = tensor.empty() : tensor<1x64xf32> + %8181 = "ttir.relu"(%8179, %8180) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8182 = tensor.empty() : tensor<1x64xf32> + %8183 = "ttir.relu"(%8181, %8182) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8184 = tensor.empty() : tensor<1x64xf32> + %8185 = "ttir.relu"(%8183, %8184) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8186 = tensor.empty() : tensor<1x64xf32> + %8187 = "ttir.relu"(%8185, %8186) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8188 = tensor.empty() : tensor<1x64xf32> + %8189 = "ttir.relu"(%8187, %8188) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8190 = tensor.empty() : tensor<1x64xf32> + %8191 = "ttir.relu"(%8189, %8190) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8192 = tensor.empty() : tensor<1x64xf32> + %8193 = "ttir.relu"(%8191, %8192) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8194 = tensor.empty() : tensor<1x64xf32> + %8195 = "ttir.relu"(%8193, %8194) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8196 = tensor.empty() : tensor<1x64xf32> + %8197 = "ttir.relu"(%8195, %8196) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8198 = tensor.empty() : tensor<1x64xf32> + %8199 = "ttir.relu"(%8197, %8198) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8200 = tensor.empty() : tensor<1x64xf32> + %8201 = "ttir.relu"(%8199, %8200) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8202 = tensor.empty() : tensor<1x64xf32> + %8203 = "ttir.relu"(%8201, %8202) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8204 = tensor.empty() : tensor<1x64xf32> + %8205 = "ttir.relu"(%8203, %8204) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8206 = tensor.empty() : tensor<1x64xf32> + %8207 = "ttir.relu"(%8205, %8206) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8208 = tensor.empty() : tensor<1x64xf32> + %8209 = "ttir.relu"(%8207, %8208) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8210 = tensor.empty() : tensor<1x64xf32> + %8211 = "ttir.relu"(%8209, %8210) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8212 = tensor.empty() : tensor<1x64xf32> + %8213 = "ttir.relu"(%8211, %8212) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8214 = tensor.empty() : tensor<1x64xf32> + %8215 = "ttir.relu"(%8213, %8214) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8216 = tensor.empty() : tensor<1x64xf32> + %8217 = "ttir.relu"(%8215, %8216) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8218 = tensor.empty() : tensor<1x64xf32> + %8219 = "ttir.relu"(%8217, %8218) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8220 = tensor.empty() : tensor<1x64xf32> + %8221 = "ttir.relu"(%8219, %8220) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8222 = tensor.empty() : tensor<1x64xf32> + %8223 = "ttir.relu"(%8221, %8222) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8224 = tensor.empty() : tensor<1x64xf32> + %8225 = "ttir.relu"(%8223, %8224) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8226 = tensor.empty() : tensor<1x64xf32> + %8227 = "ttir.relu"(%8225, %8226) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8228 = tensor.empty() : tensor<1x64xf32> + %8229 = "ttir.relu"(%8227, %8228) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8230 = tensor.empty() : tensor<1x64xf32> + %8231 = "ttir.relu"(%8229, %8230) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8232 = tensor.empty() : tensor<1x64xf32> + %8233 = "ttir.relu"(%8231, %8232) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8234 = tensor.empty() : tensor<1x64xf32> + %8235 = "ttir.relu"(%8233, %8234) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8236 = tensor.empty() : tensor<1x64xf32> + %8237 = "ttir.relu"(%8235, %8236) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8238 = tensor.empty() : tensor<1x64xf32> + %8239 = "ttir.relu"(%8237, %8238) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8240 = tensor.empty() : tensor<1x64xf32> + %8241 = "ttir.relu"(%8239, %8240) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8242 = tensor.empty() : tensor<1x64xf32> + %8243 = "ttir.relu"(%8241, %8242) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8244 = tensor.empty() : tensor<1x64xf32> + %8245 = "ttir.relu"(%8243, %8244) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8246 = tensor.empty() : tensor<1x64xf32> + %8247 = "ttir.relu"(%8245, %8246) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8248 = tensor.empty() : tensor<1x64xf32> + %8249 = "ttir.relu"(%8247, %8248) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8250 = tensor.empty() : tensor<1x64xf32> + %8251 = "ttir.relu"(%8249, %8250) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8252 = tensor.empty() : tensor<1x64xf32> + %8253 = "ttir.relu"(%8251, %8252) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8254 = tensor.empty() : tensor<1x64xf32> + %8255 = "ttir.relu"(%8253, %8254) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8256 = tensor.empty() : tensor<1x64xf32> + %8257 = "ttir.relu"(%8255, %8256) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8258 = tensor.empty() : tensor<1x64xf32> + %8259 = "ttir.relu"(%8257, %8258) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8260 = tensor.empty() : tensor<1x64xf32> + %8261 = "ttir.relu"(%8259, %8260) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8262 = tensor.empty() : tensor<1x64xf32> + %8263 = "ttir.relu"(%8261, %8262) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8264 = tensor.empty() : tensor<1x64xf32> + %8265 = "ttir.relu"(%8263, %8264) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8266 = tensor.empty() : tensor<1x64xf32> + %8267 = "ttir.relu"(%8265, %8266) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8268 = tensor.empty() : tensor<1x64xf32> + %8269 = "ttir.relu"(%8267, %8268) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8270 = tensor.empty() : tensor<1x64xf32> + %8271 = "ttir.relu"(%8269, %8270) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8272 = tensor.empty() : tensor<1x64xf32> + %8273 = "ttir.relu"(%8271, %8272) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8274 = tensor.empty() : tensor<1x64xf32> + %8275 = "ttir.relu"(%8273, %8274) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8276 = tensor.empty() : tensor<1x64xf32> + %8277 = "ttir.relu"(%8275, %8276) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8278 = tensor.empty() : tensor<1x64xf32> + %8279 = "ttir.relu"(%8277, %8278) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8280 = tensor.empty() : tensor<1x64xf32> + %8281 = "ttir.relu"(%8279, %8280) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8282 = tensor.empty() : tensor<1x64xf32> + %8283 = "ttir.relu"(%8281, %8282) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8284 = tensor.empty() : tensor<1x64xf32> + %8285 = "ttir.relu"(%8283, %8284) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8286 = tensor.empty() : tensor<1x64xf32> + %8287 = "ttir.relu"(%8285, %8286) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8288 = tensor.empty() : tensor<1x64xf32> + %8289 = "ttir.relu"(%8287, %8288) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8290 = tensor.empty() : tensor<1x64xf32> + %8291 = "ttir.relu"(%8289, %8290) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8292 = tensor.empty() : tensor<1x64xf32> + %8293 = "ttir.relu"(%8291, %8292) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8294 = tensor.empty() : tensor<1x64xf32> + %8295 = "ttir.relu"(%8293, %8294) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8296 = tensor.empty() : tensor<1x64xf32> + %8297 = "ttir.relu"(%8295, %8296) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8298 = tensor.empty() : tensor<1x64xf32> + %8299 = "ttir.relu"(%8297, %8298) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8300 = tensor.empty() : tensor<1x64xf32> + %8301 = "ttir.relu"(%8299, %8300) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8302 = tensor.empty() : tensor<1x64xf32> + %8303 = "ttir.relu"(%8301, %8302) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8304 = tensor.empty() : tensor<1x64xf32> + %8305 = "ttir.relu"(%8303, %8304) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8306 = tensor.empty() : tensor<1x64xf32> + %8307 = "ttir.relu"(%8305, %8306) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8308 = tensor.empty() : tensor<1x64xf32> + %8309 = "ttir.relu"(%8307, %8308) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8310 = tensor.empty() : tensor<1x64xf32> + %8311 = "ttir.relu"(%8309, %8310) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8312 = tensor.empty() : tensor<1x64xf32> + %8313 = "ttir.relu"(%8311, %8312) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8314 = tensor.empty() : tensor<1x64xf32> + %8315 = "ttir.relu"(%8313, %8314) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8316 = tensor.empty() : tensor<1x64xf32> + %8317 = "ttir.relu"(%8315, %8316) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8318 = tensor.empty() : tensor<1x64xf32> + %8319 = "ttir.relu"(%8317, %8318) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8320 = tensor.empty() : tensor<1x64xf32> + %8321 = "ttir.relu"(%8319, %8320) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8322 = tensor.empty() : tensor<1x64xf32> + %8323 = "ttir.relu"(%8321, %8322) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8324 = tensor.empty() : tensor<1x64xf32> + %8325 = "ttir.relu"(%8323, %8324) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8326 = tensor.empty() : tensor<1x64xf32> + %8327 = "ttir.relu"(%8325, %8326) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8328 = tensor.empty() : tensor<1x64xf32> + %8329 = "ttir.relu"(%8327, %8328) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8330 = tensor.empty() : tensor<1x64xf32> + %8331 = "ttir.relu"(%8329, %8330) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8332 = tensor.empty() : tensor<1x64xf32> + %8333 = "ttir.relu"(%8331, %8332) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8334 = tensor.empty() : tensor<1x64xf32> + %8335 = "ttir.relu"(%8333, %8334) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8336 = tensor.empty() : tensor<1x64xf32> + %8337 = "ttir.relu"(%8335, %8336) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8338 = tensor.empty() : tensor<1x64xf32> + %8339 = "ttir.relu"(%8337, %8338) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8340 = tensor.empty() : tensor<1x64xf32> + %8341 = "ttir.relu"(%8339, %8340) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8342 = tensor.empty() : tensor<1x64xf32> + %8343 = "ttir.relu"(%8341, %8342) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8344 = tensor.empty() : tensor<1x64xf32> + %8345 = "ttir.relu"(%8343, %8344) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8346 = tensor.empty() : tensor<1x64xf32> + %8347 = "ttir.relu"(%8345, %8346) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8348 = tensor.empty() : tensor<1x64xf32> + %8349 = "ttir.relu"(%8347, %8348) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8350 = tensor.empty() : tensor<1x64xf32> + %8351 = "ttir.relu"(%8349, %8350) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8352 = tensor.empty() : tensor<1x64xf32> + %8353 = "ttir.relu"(%8351, %8352) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8354 = tensor.empty() : tensor<1x64xf32> + %8355 = "ttir.relu"(%8353, %8354) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8356 = tensor.empty() : tensor<1x64xf32> + %8357 = "ttir.relu"(%8355, %8356) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8358 = tensor.empty() : tensor<1x64xf32> + %8359 = "ttir.relu"(%8357, %8358) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8360 = tensor.empty() : tensor<1x64xf32> + %8361 = "ttir.relu"(%8359, %8360) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8362 = tensor.empty() : tensor<1x64xf32> + %8363 = "ttir.relu"(%8361, %8362) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8364 = tensor.empty() : tensor<1x64xf32> + %8365 = "ttir.relu"(%8363, %8364) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8366 = tensor.empty() : tensor<1x64xf32> + %8367 = "ttir.relu"(%8365, %8366) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8368 = tensor.empty() : tensor<1x64xf32> + %8369 = "ttir.relu"(%8367, %8368) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8370 = tensor.empty() : tensor<1x64xf32> + %8371 = "ttir.relu"(%8369, %8370) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8372 = tensor.empty() : tensor<1x64xf32> + %8373 = "ttir.relu"(%8371, %8372) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8374 = tensor.empty() : tensor<1x64xf32> + %8375 = "ttir.relu"(%8373, %8374) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8376 = tensor.empty() : tensor<1x64xf32> + %8377 = "ttir.relu"(%8375, %8376) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8378 = tensor.empty() : tensor<1x64xf32> + %8379 = "ttir.relu"(%8377, %8378) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8380 = tensor.empty() : tensor<1x64xf32> + %8381 = "ttir.relu"(%8379, %8380) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8382 = tensor.empty() : tensor<1x64xf32> + %8383 = "ttir.relu"(%8381, %8382) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8384 = tensor.empty() : tensor<1x64xf32> + %8385 = "ttir.relu"(%8383, %8384) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8386 = tensor.empty() : tensor<1x64xf32> + %8387 = "ttir.relu"(%8385, %8386) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8388 = tensor.empty() : tensor<1x64xf32> + %8389 = "ttir.relu"(%8387, %8388) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8390 = tensor.empty() : tensor<1x64xf32> + %8391 = "ttir.relu"(%8389, %8390) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8392 = tensor.empty() : tensor<1x64xf32> + %8393 = "ttir.relu"(%8391, %8392) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8394 = tensor.empty() : tensor<1x64xf32> + %8395 = "ttir.relu"(%8393, %8394) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8396 = tensor.empty() : tensor<1x64xf32> + %8397 = "ttir.relu"(%8395, %8396) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8398 = tensor.empty() : tensor<1x64xf32> + %8399 = "ttir.relu"(%8397, %8398) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8400 = tensor.empty() : tensor<1x64xf32> + %8401 = "ttir.relu"(%8399, %8400) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8402 = tensor.empty() : tensor<1x64xf32> + %8403 = "ttir.relu"(%8401, %8402) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8404 = tensor.empty() : tensor<1x64xf32> + %8405 = "ttir.relu"(%8403, %8404) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8406 = tensor.empty() : tensor<1x64xf32> + %8407 = "ttir.relu"(%8405, %8406) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8408 = tensor.empty() : tensor<1x64xf32> + %8409 = "ttir.relu"(%8407, %8408) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8410 = tensor.empty() : tensor<1x64xf32> + %8411 = "ttir.relu"(%8409, %8410) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8412 = tensor.empty() : tensor<1x64xf32> + %8413 = "ttir.relu"(%8411, %8412) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8414 = tensor.empty() : tensor<1x64xf32> + %8415 = "ttir.relu"(%8413, %8414) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8416 = tensor.empty() : tensor<1x64xf32> + %8417 = "ttir.relu"(%8415, %8416) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8418 = tensor.empty() : tensor<1x64xf32> + %8419 = "ttir.relu"(%8417, %8418) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8420 = tensor.empty() : tensor<1x64xf32> + %8421 = "ttir.relu"(%8419, %8420) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8422 = tensor.empty() : tensor<1x64xf32> + %8423 = "ttir.relu"(%8421, %8422) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8424 = tensor.empty() : tensor<1x64xf32> + %8425 = "ttir.relu"(%8423, %8424) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8426 = tensor.empty() : tensor<1x64xf32> + %8427 = "ttir.relu"(%8425, %8426) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8428 = tensor.empty() : tensor<1x64xf32> + %8429 = "ttir.relu"(%8427, %8428) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8430 = tensor.empty() : tensor<1x64xf32> + %8431 = "ttir.relu"(%8429, %8430) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8432 = tensor.empty() : tensor<1x64xf32> + %8433 = "ttir.relu"(%8431, %8432) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8434 = tensor.empty() : tensor<1x64xf32> + %8435 = "ttir.relu"(%8433, %8434) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8436 = tensor.empty() : tensor<1x64xf32> + %8437 = "ttir.relu"(%8435, %8436) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8438 = tensor.empty() : tensor<1x64xf32> + %8439 = "ttir.relu"(%8437, %8438) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8440 = tensor.empty() : tensor<1x64xf32> + %8441 = "ttir.relu"(%8439, %8440) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8442 = tensor.empty() : tensor<1x64xf32> + %8443 = "ttir.relu"(%8441, %8442) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8444 = tensor.empty() : tensor<1x64xf32> + %8445 = "ttir.relu"(%8443, %8444) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8446 = tensor.empty() : tensor<1x64xf32> + %8447 = "ttir.relu"(%8445, %8446) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8448 = tensor.empty() : tensor<1x64xf32> + %8449 = "ttir.relu"(%8447, %8448) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8450 = tensor.empty() : tensor<1x64xf32> + %8451 = "ttir.relu"(%8449, %8450) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8452 = tensor.empty() : tensor<1x64xf32> + %8453 = "ttir.relu"(%8451, %8452) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8454 = tensor.empty() : tensor<1x64xf32> + %8455 = "ttir.relu"(%8453, %8454) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8456 = tensor.empty() : tensor<1x64xf32> + %8457 = "ttir.relu"(%8455, %8456) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8458 = tensor.empty() : tensor<1x64xf32> + %8459 = "ttir.relu"(%8457, %8458) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8460 = tensor.empty() : tensor<1x64xf32> + %8461 = "ttir.relu"(%8459, %8460) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8462 = tensor.empty() : tensor<1x64xf32> + %8463 = "ttir.relu"(%8461, %8462) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8464 = tensor.empty() : tensor<1x64xf32> + %8465 = "ttir.relu"(%8463, %8464) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8466 = tensor.empty() : tensor<1x64xf32> + %8467 = "ttir.relu"(%8465, %8466) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8468 = tensor.empty() : tensor<1x64xf32> + %8469 = "ttir.relu"(%8467, %8468) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8470 = tensor.empty() : tensor<1x64xf32> + %8471 = "ttir.relu"(%8469, %8470) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8472 = tensor.empty() : tensor<1x64xf32> + %8473 = "ttir.relu"(%8471, %8472) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8474 = tensor.empty() : tensor<1x64xf32> + %8475 = "ttir.relu"(%8473, %8474) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8476 = tensor.empty() : tensor<1x64xf32> + %8477 = "ttir.relu"(%8475, %8476) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8478 = tensor.empty() : tensor<1x64xf32> + %8479 = "ttir.relu"(%8477, %8478) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8480 = tensor.empty() : tensor<1x64xf32> + %8481 = "ttir.relu"(%8479, %8480) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8482 = tensor.empty() : tensor<1x64xf32> + %8483 = "ttir.relu"(%8481, %8482) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8484 = tensor.empty() : tensor<1x64xf32> + %8485 = "ttir.relu"(%8483, %8484) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8486 = tensor.empty() : tensor<1x64xf32> + %8487 = "ttir.relu"(%8485, %8486) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8488 = tensor.empty() : tensor<1x64xf32> + %8489 = "ttir.relu"(%8487, %8488) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8490 = tensor.empty() : tensor<1x64xf32> + %8491 = "ttir.relu"(%8489, %8490) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8492 = tensor.empty() : tensor<1x64xf32> + %8493 = "ttir.relu"(%8491, %8492) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8494 = tensor.empty() : tensor<1x64xf32> + %8495 = "ttir.relu"(%8493, %8494) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8496 = tensor.empty() : tensor<1x64xf32> + %8497 = "ttir.relu"(%8495, %8496) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8498 = tensor.empty() : tensor<1x64xf32> + %8499 = "ttir.relu"(%8497, %8498) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8500 = tensor.empty() : tensor<1x64xf32> + %8501 = "ttir.relu"(%8499, %8500) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8502 = tensor.empty() : tensor<1x64xf32> + %8503 = "ttir.relu"(%8501, %8502) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8504 = tensor.empty() : tensor<1x64xf32> + %8505 = "ttir.relu"(%8503, %8504) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8506 = tensor.empty() : tensor<1x64xf32> + %8507 = "ttir.relu"(%8505, %8506) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8508 = tensor.empty() : tensor<1x64xf32> + %8509 = "ttir.relu"(%8507, %8508) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8510 = tensor.empty() : tensor<1x64xf32> + %8511 = "ttir.relu"(%8509, %8510) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8512 = tensor.empty() : tensor<1x64xf32> + %8513 = "ttir.relu"(%8511, %8512) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8514 = tensor.empty() : tensor<1x64xf32> + %8515 = "ttir.relu"(%8513, %8514) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8516 = tensor.empty() : tensor<1x64xf32> + %8517 = "ttir.relu"(%8515, %8516) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8518 = tensor.empty() : tensor<1x64xf32> + %8519 = "ttir.relu"(%8517, %8518) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8520 = tensor.empty() : tensor<1x64xf32> + %8521 = "ttir.relu"(%8519, %8520) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8522 = tensor.empty() : tensor<1x64xf32> + %8523 = "ttir.relu"(%8521, %8522) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8524 = tensor.empty() : tensor<1x64xf32> + %8525 = "ttir.relu"(%8523, %8524) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8526 = tensor.empty() : tensor<1x64xf32> + %8527 = "ttir.relu"(%8525, %8526) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8528 = tensor.empty() : tensor<1x64xf32> + %8529 = "ttir.relu"(%8527, %8528) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8530 = tensor.empty() : tensor<1x64xf32> + %8531 = "ttir.relu"(%8529, %8530) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8532 = tensor.empty() : tensor<1x64xf32> + %8533 = "ttir.relu"(%8531, %8532) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8534 = tensor.empty() : tensor<1x64xf32> + %8535 = "ttir.relu"(%8533, %8534) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8536 = tensor.empty() : tensor<1x64xf32> + %8537 = "ttir.relu"(%8535, %8536) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8538 = tensor.empty() : tensor<1x64xf32> + %8539 = "ttir.relu"(%8537, %8538) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8540 = tensor.empty() : tensor<1x64xf32> + %8541 = "ttir.relu"(%8539, %8540) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8542 = tensor.empty() : tensor<1x64xf32> + %8543 = "ttir.relu"(%8541, %8542) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8544 = tensor.empty() : tensor<1x64xf32> + %8545 = "ttir.relu"(%8543, %8544) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8546 = tensor.empty() : tensor<1x64xf32> + %8547 = "ttir.relu"(%8545, %8546) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8548 = tensor.empty() : tensor<1x64xf32> + %8549 = "ttir.relu"(%8547, %8548) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8550 = tensor.empty() : tensor<1x64xf32> + %8551 = "ttir.relu"(%8549, %8550) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8552 = tensor.empty() : tensor<1x64xf32> + %8553 = "ttir.relu"(%8551, %8552) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8554 = tensor.empty() : tensor<1x64xf32> + %8555 = "ttir.relu"(%8553, %8554) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8556 = tensor.empty() : tensor<1x64xf32> + %8557 = "ttir.relu"(%8555, %8556) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8558 = tensor.empty() : tensor<1x64xf32> + %8559 = "ttir.relu"(%8557, %8558) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8560 = tensor.empty() : tensor<1x64xf32> + %8561 = "ttir.relu"(%8559, %8560) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8562 = tensor.empty() : tensor<1x64xf32> + %8563 = "ttir.relu"(%8561, %8562) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8564 = tensor.empty() : tensor<1x64xf32> + %8565 = "ttir.relu"(%8563, %8564) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8566 = tensor.empty() : tensor<1x64xf32> + %8567 = "ttir.relu"(%8565, %8566) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8568 = tensor.empty() : tensor<1x64xf32> + %8569 = "ttir.relu"(%8567, %8568) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8570 = tensor.empty() : tensor<1x64xf32> + %8571 = "ttir.relu"(%8569, %8570) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8572 = tensor.empty() : tensor<1x64xf32> + %8573 = "ttir.relu"(%8571, %8572) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8574 = tensor.empty() : tensor<1x64xf32> + %8575 = "ttir.relu"(%8573, %8574) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8576 = tensor.empty() : tensor<1x64xf32> + %8577 = "ttir.relu"(%8575, %8576) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8578 = tensor.empty() : tensor<1x64xf32> + %8579 = "ttir.relu"(%8577, %8578) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8580 = tensor.empty() : tensor<1x64xf32> + %8581 = "ttir.relu"(%8579, %8580) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8582 = tensor.empty() : tensor<1x64xf32> + %8583 = "ttir.relu"(%8581, %8582) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8584 = tensor.empty() : tensor<1x64xf32> + %8585 = "ttir.relu"(%8583, %8584) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8586 = tensor.empty() : tensor<1x64xf32> + %8587 = "ttir.relu"(%8585, %8586) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8588 = tensor.empty() : tensor<1x64xf32> + %8589 = "ttir.relu"(%8587, %8588) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8590 = tensor.empty() : tensor<1x64xf32> + %8591 = "ttir.relu"(%8589, %8590) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8592 = tensor.empty() : tensor<1x64xf32> + %8593 = "ttir.relu"(%8591, %8592) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8594 = tensor.empty() : tensor<1x64xf32> + %8595 = "ttir.relu"(%8593, %8594) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8596 = tensor.empty() : tensor<1x64xf32> + %8597 = "ttir.relu"(%8595, %8596) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8598 = tensor.empty() : tensor<1x64xf32> + %8599 = "ttir.relu"(%8597, %8598) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8600 = tensor.empty() : tensor<1x64xf32> + %8601 = "ttir.relu"(%8599, %8600) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8602 = tensor.empty() : tensor<1x64xf32> + %8603 = "ttir.relu"(%8601, %8602) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8604 = tensor.empty() : tensor<1x64xf32> + %8605 = "ttir.relu"(%8603, %8604) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8606 = tensor.empty() : tensor<1x64xf32> + %8607 = "ttir.relu"(%8605, %8606) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8608 = tensor.empty() : tensor<1x64xf32> + %8609 = "ttir.relu"(%8607, %8608) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8610 = tensor.empty() : tensor<1x64xf32> + %8611 = "ttir.relu"(%8609, %8610) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8612 = tensor.empty() : tensor<1x64xf32> + %8613 = "ttir.relu"(%8611, %8612) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8614 = tensor.empty() : tensor<1x64xf32> + %8615 = "ttir.relu"(%8613, %8614) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8616 = tensor.empty() : tensor<1x64xf32> + %8617 = "ttir.relu"(%8615, %8616) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8618 = tensor.empty() : tensor<1x64xf32> + %8619 = "ttir.relu"(%8617, %8618) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8620 = tensor.empty() : tensor<1x64xf32> + %8621 = "ttir.relu"(%8619, %8620) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8622 = tensor.empty() : tensor<1x64xf32> + %8623 = "ttir.relu"(%8621, %8622) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8624 = tensor.empty() : tensor<1x64xf32> + %8625 = "ttir.relu"(%8623, %8624) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8626 = tensor.empty() : tensor<1x64xf32> + %8627 = "ttir.relu"(%8625, %8626) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8628 = tensor.empty() : tensor<1x64xf32> + %8629 = "ttir.relu"(%8627, %8628) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8630 = tensor.empty() : tensor<1x64xf32> + %8631 = "ttir.relu"(%8629, %8630) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8632 = tensor.empty() : tensor<1x64xf32> + %8633 = "ttir.relu"(%8631, %8632) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8634 = tensor.empty() : tensor<1x64xf32> + %8635 = "ttir.relu"(%8633, %8634) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8636 = tensor.empty() : tensor<1x64xf32> + %8637 = "ttir.relu"(%8635, %8636) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8638 = tensor.empty() : tensor<1x64xf32> + %8639 = "ttir.relu"(%8637, %8638) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8640 = tensor.empty() : tensor<1x64xf32> + %8641 = "ttir.relu"(%8639, %8640) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8642 = tensor.empty() : tensor<1x64xf32> + %8643 = "ttir.relu"(%8641, %8642) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8644 = tensor.empty() : tensor<1x64xf32> + %8645 = "ttir.relu"(%8643, %8644) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8646 = tensor.empty() : tensor<1x64xf32> + %8647 = "ttir.relu"(%8645, %8646) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8648 = tensor.empty() : tensor<1x64xf32> + %8649 = "ttir.relu"(%8647, %8648) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8650 = tensor.empty() : tensor<1x64xf32> + %8651 = "ttir.relu"(%8649, %8650) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8652 = tensor.empty() : tensor<1x64xf32> + %8653 = "ttir.relu"(%8651, %8652) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8654 = tensor.empty() : tensor<1x64xf32> + %8655 = "ttir.relu"(%8653, %8654) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8656 = tensor.empty() : tensor<1x64xf32> + %8657 = "ttir.relu"(%8655, %8656) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8658 = tensor.empty() : tensor<1x64xf32> + %8659 = "ttir.relu"(%8657, %8658) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8660 = tensor.empty() : tensor<1x64xf32> + %8661 = "ttir.relu"(%8659, %8660) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8662 = tensor.empty() : tensor<1x64xf32> + %8663 = "ttir.relu"(%8661, %8662) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8664 = tensor.empty() : tensor<1x64xf32> + %8665 = "ttir.relu"(%8663, %8664) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8666 = tensor.empty() : tensor<1x64xf32> + %8667 = "ttir.relu"(%8665, %8666) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8668 = tensor.empty() : tensor<1x64xf32> + %8669 = "ttir.relu"(%8667, %8668) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8670 = tensor.empty() : tensor<1x64xf32> + %8671 = "ttir.relu"(%8669, %8670) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8672 = tensor.empty() : tensor<1x64xf32> + %8673 = "ttir.relu"(%8671, %8672) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8674 = tensor.empty() : tensor<1x64xf32> + %8675 = "ttir.relu"(%8673, %8674) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8676 = tensor.empty() : tensor<1x64xf32> + %8677 = "ttir.relu"(%8675, %8676) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8678 = tensor.empty() : tensor<1x64xf32> + %8679 = "ttir.relu"(%8677, %8678) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8680 = tensor.empty() : tensor<1x64xf32> + %8681 = "ttir.relu"(%8679, %8680) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8682 = tensor.empty() : tensor<1x64xf32> + %8683 = "ttir.relu"(%8681, %8682) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8684 = tensor.empty() : tensor<1x64xf32> + %8685 = "ttir.relu"(%8683, %8684) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8686 = tensor.empty() : tensor<1x64xf32> + %8687 = "ttir.relu"(%8685, %8686) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8688 = tensor.empty() : tensor<1x64xf32> + %8689 = "ttir.relu"(%8687, %8688) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8690 = tensor.empty() : tensor<1x64xf32> + %8691 = "ttir.relu"(%8689, %8690) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8692 = tensor.empty() : tensor<1x64xf32> + %8693 = "ttir.relu"(%8691, %8692) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8694 = tensor.empty() : tensor<1x64xf32> + %8695 = "ttir.relu"(%8693, %8694) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8696 = tensor.empty() : tensor<1x64xf32> + %8697 = "ttir.relu"(%8695, %8696) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8698 = tensor.empty() : tensor<1x64xf32> + %8699 = "ttir.relu"(%8697, %8698) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8700 = tensor.empty() : tensor<1x64xf32> + %8701 = "ttir.relu"(%8699, %8700) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8702 = tensor.empty() : tensor<1x64xf32> + %8703 = "ttir.relu"(%8701, %8702) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8704 = tensor.empty() : tensor<1x64xf32> + %8705 = "ttir.relu"(%8703, %8704) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8706 = tensor.empty() : tensor<1x64xf32> + %8707 = "ttir.relu"(%8705, %8706) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8708 = tensor.empty() : tensor<1x64xf32> + %8709 = "ttir.relu"(%8707, %8708) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8710 = tensor.empty() : tensor<1x64xf32> + %8711 = "ttir.relu"(%8709, %8710) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8712 = tensor.empty() : tensor<1x64xf32> + %8713 = "ttir.relu"(%8711, %8712) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8714 = tensor.empty() : tensor<1x64xf32> + %8715 = "ttir.relu"(%8713, %8714) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8716 = tensor.empty() : tensor<1x64xf32> + %8717 = "ttir.relu"(%8715, %8716) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8718 = tensor.empty() : tensor<1x64xf32> + %8719 = "ttir.relu"(%8717, %8718) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8720 = tensor.empty() : tensor<1x64xf32> + %8721 = "ttir.relu"(%8719, %8720) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8722 = tensor.empty() : tensor<1x64xf32> + %8723 = "ttir.relu"(%8721, %8722) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8724 = tensor.empty() : tensor<1x64xf32> + %8725 = "ttir.relu"(%8723, %8724) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8726 = tensor.empty() : tensor<1x64xf32> + %8727 = "ttir.relu"(%8725, %8726) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8728 = tensor.empty() : tensor<1x64xf32> + %8729 = "ttir.relu"(%8727, %8728) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8730 = tensor.empty() : tensor<1x64xf32> + %8731 = "ttir.relu"(%8729, %8730) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8732 = tensor.empty() : tensor<1x64xf32> + %8733 = "ttir.relu"(%8731, %8732) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8734 = tensor.empty() : tensor<1x64xf32> + %8735 = "ttir.relu"(%8733, %8734) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8736 = tensor.empty() : tensor<1x64xf32> + %8737 = "ttir.relu"(%8735, %8736) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8738 = tensor.empty() : tensor<1x64xf32> + %8739 = "ttir.relu"(%8737, %8738) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8740 = tensor.empty() : tensor<1x64xf32> + %8741 = "ttir.relu"(%8739, %8740) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8742 = tensor.empty() : tensor<1x64xf32> + %8743 = "ttir.relu"(%8741, %8742) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8744 = tensor.empty() : tensor<1x64xf32> + %8745 = "ttir.relu"(%8743, %8744) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8746 = tensor.empty() : tensor<1x64xf32> + %8747 = "ttir.relu"(%8745, %8746) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8748 = tensor.empty() : tensor<1x64xf32> + %8749 = "ttir.relu"(%8747, %8748) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8750 = tensor.empty() : tensor<1x64xf32> + %8751 = "ttir.relu"(%8749, %8750) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8752 = tensor.empty() : tensor<1x64xf32> + %8753 = "ttir.relu"(%8751, %8752) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8754 = tensor.empty() : tensor<1x64xf32> + %8755 = "ttir.relu"(%8753, %8754) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8756 = tensor.empty() : tensor<1x64xf32> + %8757 = "ttir.relu"(%8755, %8756) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8758 = tensor.empty() : tensor<1x64xf32> + %8759 = "ttir.relu"(%8757, %8758) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8760 = tensor.empty() : tensor<1x64xf32> + %8761 = "ttir.relu"(%8759, %8760) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8762 = tensor.empty() : tensor<1x64xf32> + %8763 = "ttir.relu"(%8761, %8762) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8764 = tensor.empty() : tensor<1x64xf32> + %8765 = "ttir.relu"(%8763, %8764) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8766 = tensor.empty() : tensor<1x64xf32> + %8767 = "ttir.relu"(%8765, %8766) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8768 = tensor.empty() : tensor<1x64xf32> + %8769 = "ttir.relu"(%8767, %8768) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8770 = tensor.empty() : tensor<1x64xf32> + %8771 = "ttir.relu"(%8769, %8770) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8772 = tensor.empty() : tensor<1x64xf32> + %8773 = "ttir.relu"(%8771, %8772) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8774 = tensor.empty() : tensor<1x64xf32> + %8775 = "ttir.relu"(%8773, %8774) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8776 = tensor.empty() : tensor<1x64xf32> + %8777 = "ttir.relu"(%8775, %8776) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8778 = tensor.empty() : tensor<1x64xf32> + %8779 = "ttir.relu"(%8777, %8778) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8780 = tensor.empty() : tensor<1x64xf32> + %8781 = "ttir.relu"(%8779, %8780) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8782 = tensor.empty() : tensor<1x64xf32> + %8783 = "ttir.relu"(%8781, %8782) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8784 = tensor.empty() : tensor<1x64xf32> + %8785 = "ttir.relu"(%8783, %8784) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8786 = tensor.empty() : tensor<1x64xf32> + %8787 = "ttir.relu"(%8785, %8786) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8788 = tensor.empty() : tensor<1x64xf32> + %8789 = "ttir.relu"(%8787, %8788) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8790 = tensor.empty() : tensor<1x64xf32> + %8791 = "ttir.relu"(%8789, %8790) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8792 = tensor.empty() : tensor<1x64xf32> + %8793 = "ttir.relu"(%8791, %8792) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8794 = tensor.empty() : tensor<1x64xf32> + %8795 = "ttir.relu"(%8793, %8794) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8796 = tensor.empty() : tensor<1x64xf32> + %8797 = "ttir.relu"(%8795, %8796) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8798 = tensor.empty() : tensor<1x64xf32> + %8799 = "ttir.relu"(%8797, %8798) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8800 = tensor.empty() : tensor<1x64xf32> + %8801 = "ttir.relu"(%8799, %8800) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8802 = tensor.empty() : tensor<1x64xf32> + %8803 = "ttir.relu"(%8801, %8802) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8804 = tensor.empty() : tensor<1x64xf32> + %8805 = "ttir.relu"(%8803, %8804) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8806 = tensor.empty() : tensor<1x64xf32> + %8807 = "ttir.relu"(%8805, %8806) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8808 = tensor.empty() : tensor<1x64xf32> + %8809 = "ttir.relu"(%8807, %8808) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8810 = tensor.empty() : tensor<1x64xf32> + %8811 = "ttir.relu"(%8809, %8810) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8812 = tensor.empty() : tensor<1x64xf32> + %8813 = "ttir.relu"(%8811, %8812) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8814 = tensor.empty() : tensor<1x64xf32> + %8815 = "ttir.relu"(%8813, %8814) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8816 = tensor.empty() : tensor<1x64xf32> + %8817 = "ttir.relu"(%8815, %8816) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8818 = tensor.empty() : tensor<1x64xf32> + %8819 = "ttir.relu"(%8817, %8818) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8820 = tensor.empty() : tensor<1x64xf32> + %8821 = "ttir.relu"(%8819, %8820) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8822 = tensor.empty() : tensor<1x64xf32> + %8823 = "ttir.relu"(%8821, %8822) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8824 = tensor.empty() : tensor<1x64xf32> + %8825 = "ttir.relu"(%8823, %8824) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8826 = tensor.empty() : tensor<1x64xf32> + %8827 = "ttir.relu"(%8825, %8826) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8828 = tensor.empty() : tensor<1x64xf32> + %8829 = "ttir.relu"(%8827, %8828) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8830 = tensor.empty() : tensor<1x64xf32> + %8831 = "ttir.relu"(%8829, %8830) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8832 = tensor.empty() : tensor<1x64xf32> + %8833 = "ttir.relu"(%8831, %8832) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8834 = tensor.empty() : tensor<1x64xf32> + %8835 = "ttir.relu"(%8833, %8834) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8836 = tensor.empty() : tensor<1x64xf32> + %8837 = "ttir.relu"(%8835, %8836) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8838 = tensor.empty() : tensor<1x64xf32> + %8839 = "ttir.relu"(%8837, %8838) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8840 = tensor.empty() : tensor<1x64xf32> + %8841 = "ttir.relu"(%8839, %8840) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8842 = tensor.empty() : tensor<1x64xf32> + %8843 = "ttir.relu"(%8841, %8842) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8844 = tensor.empty() : tensor<1x64xf32> + %8845 = "ttir.relu"(%8843, %8844) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8846 = tensor.empty() : tensor<1x64xf32> + %8847 = "ttir.relu"(%8845, %8846) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8848 = tensor.empty() : tensor<1x64xf32> + %8849 = "ttir.relu"(%8847, %8848) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8850 = tensor.empty() : tensor<1x64xf32> + %8851 = "ttir.relu"(%8849, %8850) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8852 = tensor.empty() : tensor<1x64xf32> + %8853 = "ttir.relu"(%8851, %8852) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8854 = tensor.empty() : tensor<1x64xf32> + %8855 = "ttir.relu"(%8853, %8854) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8856 = tensor.empty() : tensor<1x64xf32> + %8857 = "ttir.relu"(%8855, %8856) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8858 = tensor.empty() : tensor<1x64xf32> + %8859 = "ttir.relu"(%8857, %8858) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8860 = tensor.empty() : tensor<1x64xf32> + %8861 = "ttir.relu"(%8859, %8860) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8862 = tensor.empty() : tensor<1x64xf32> + %8863 = "ttir.relu"(%8861, %8862) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8864 = tensor.empty() : tensor<1x64xf32> + %8865 = "ttir.relu"(%8863, %8864) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8866 = tensor.empty() : tensor<1x64xf32> + %8867 = "ttir.relu"(%8865, %8866) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8868 = tensor.empty() : tensor<1x64xf32> + %8869 = "ttir.relu"(%8867, %8868) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8870 = tensor.empty() : tensor<1x64xf32> + %8871 = "ttir.relu"(%8869, %8870) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8872 = tensor.empty() : tensor<1x64xf32> + %8873 = "ttir.relu"(%8871, %8872) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8874 = tensor.empty() : tensor<1x64xf32> + %8875 = "ttir.relu"(%8873, %8874) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8876 = tensor.empty() : tensor<1x64xf32> + %8877 = "ttir.relu"(%8875, %8876) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8878 = tensor.empty() : tensor<1x64xf32> + %8879 = "ttir.relu"(%8877, %8878) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8880 = tensor.empty() : tensor<1x64xf32> + %8881 = "ttir.relu"(%8879, %8880) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8882 = tensor.empty() : tensor<1x64xf32> + %8883 = "ttir.relu"(%8881, %8882) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8884 = tensor.empty() : tensor<1x64xf32> + %8885 = "ttir.relu"(%8883, %8884) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8886 = tensor.empty() : tensor<1x64xf32> + %8887 = "ttir.relu"(%8885, %8886) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8888 = tensor.empty() : tensor<1x64xf32> + %8889 = "ttir.relu"(%8887, %8888) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8890 = tensor.empty() : tensor<1x64xf32> + %8891 = "ttir.relu"(%8889, %8890) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8892 = tensor.empty() : tensor<1x64xf32> + %8893 = "ttir.relu"(%8891, %8892) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8894 = tensor.empty() : tensor<1x64xf32> + %8895 = "ttir.relu"(%8893, %8894) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8896 = tensor.empty() : tensor<1x64xf32> + %8897 = "ttir.relu"(%8895, %8896) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8898 = tensor.empty() : tensor<1x64xf32> + %8899 = "ttir.relu"(%8897, %8898) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8900 = tensor.empty() : tensor<1x64xf32> + %8901 = "ttir.relu"(%8899, %8900) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8902 = tensor.empty() : tensor<1x64xf32> + %8903 = "ttir.relu"(%8901, %8902) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8904 = tensor.empty() : tensor<1x64xf32> + %8905 = "ttir.relu"(%8903, %8904) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8906 = tensor.empty() : tensor<1x64xf32> + %8907 = "ttir.relu"(%8905, %8906) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8908 = tensor.empty() : tensor<1x64xf32> + %8909 = "ttir.relu"(%8907, %8908) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8910 = tensor.empty() : tensor<1x64xf32> + %8911 = "ttir.relu"(%8909, %8910) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8912 = tensor.empty() : tensor<1x64xf32> + %8913 = "ttir.relu"(%8911, %8912) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8914 = tensor.empty() : tensor<1x64xf32> + %8915 = "ttir.relu"(%8913, %8914) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8916 = tensor.empty() : tensor<1x64xf32> + %8917 = "ttir.relu"(%8915, %8916) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8918 = tensor.empty() : tensor<1x64xf32> + %8919 = "ttir.relu"(%8917, %8918) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8920 = tensor.empty() : tensor<1x64xf32> + %8921 = "ttir.relu"(%8919, %8920) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8922 = tensor.empty() : tensor<1x64xf32> + %8923 = "ttir.relu"(%8921, %8922) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8924 = tensor.empty() : tensor<1x64xf32> + %8925 = "ttir.relu"(%8923, %8924) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8926 = tensor.empty() : tensor<1x64xf32> + %8927 = "ttir.relu"(%8925, %8926) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8928 = tensor.empty() : tensor<1x64xf32> + %8929 = "ttir.relu"(%8927, %8928) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8930 = tensor.empty() : tensor<1x64xf32> + %8931 = "ttir.relu"(%8929, %8930) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8932 = tensor.empty() : tensor<1x64xf32> + %8933 = "ttir.relu"(%8931, %8932) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8934 = tensor.empty() : tensor<1x64xf32> + %8935 = "ttir.relu"(%8933, %8934) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8936 = tensor.empty() : tensor<1x64xf32> + %8937 = "ttir.relu"(%8935, %8936) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8938 = tensor.empty() : tensor<1x64xf32> + %8939 = "ttir.relu"(%8937, %8938) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8940 = tensor.empty() : tensor<1x64xf32> + %8941 = "ttir.relu"(%8939, %8940) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8942 = tensor.empty() : tensor<1x64xf32> + %8943 = "ttir.relu"(%8941, %8942) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8944 = tensor.empty() : tensor<1x64xf32> + %8945 = "ttir.relu"(%8943, %8944) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8946 = tensor.empty() : tensor<1x64xf32> + %8947 = "ttir.relu"(%8945, %8946) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8948 = tensor.empty() : tensor<1x64xf32> + %8949 = "ttir.relu"(%8947, %8948) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8950 = tensor.empty() : tensor<1x64xf32> + %8951 = "ttir.relu"(%8949, %8950) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8952 = tensor.empty() : tensor<1x64xf32> + %8953 = "ttir.relu"(%8951, %8952) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8954 = tensor.empty() : tensor<1x64xf32> + %8955 = "ttir.relu"(%8953, %8954) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8956 = tensor.empty() : tensor<1x64xf32> + %8957 = "ttir.relu"(%8955, %8956) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8958 = tensor.empty() : tensor<1x64xf32> + %8959 = "ttir.relu"(%8957, %8958) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8960 = tensor.empty() : tensor<1x64xf32> + %8961 = "ttir.relu"(%8959, %8960) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8962 = tensor.empty() : tensor<1x64xf32> + %8963 = "ttir.relu"(%8961, %8962) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8964 = tensor.empty() : tensor<1x64xf32> + %8965 = "ttir.relu"(%8963, %8964) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8966 = tensor.empty() : tensor<1x64xf32> + %8967 = "ttir.relu"(%8965, %8966) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8968 = tensor.empty() : tensor<1x64xf32> + %8969 = "ttir.relu"(%8967, %8968) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8970 = tensor.empty() : tensor<1x64xf32> + %8971 = "ttir.relu"(%8969, %8970) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8972 = tensor.empty() : tensor<1x64xf32> + %8973 = "ttir.relu"(%8971, %8972) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8974 = tensor.empty() : tensor<1x64xf32> + %8975 = "ttir.relu"(%8973, %8974) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8976 = tensor.empty() : tensor<1x64xf32> + %8977 = "ttir.relu"(%8975, %8976) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8978 = tensor.empty() : tensor<1x64xf32> + %8979 = "ttir.relu"(%8977, %8978) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8980 = tensor.empty() : tensor<1x64xf32> + %8981 = "ttir.relu"(%8979, %8980) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8982 = tensor.empty() : tensor<1x64xf32> + %8983 = "ttir.relu"(%8981, %8982) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8984 = tensor.empty() : tensor<1x64xf32> + %8985 = "ttir.relu"(%8983, %8984) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8986 = tensor.empty() : tensor<1x64xf32> + %8987 = "ttir.relu"(%8985, %8986) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8988 = tensor.empty() : tensor<1x64xf32> + %8989 = "ttir.relu"(%8987, %8988) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8990 = tensor.empty() : tensor<1x64xf32> + %8991 = "ttir.relu"(%8989, %8990) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8992 = tensor.empty() : tensor<1x64xf32> + %8993 = "ttir.relu"(%8991, %8992) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8994 = tensor.empty() : tensor<1x64xf32> + %8995 = "ttir.relu"(%8993, %8994) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8996 = tensor.empty() : tensor<1x64xf32> + %8997 = "ttir.relu"(%8995, %8996) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8998 = tensor.empty() : tensor<1x64xf32> + %8999 = "ttir.relu"(%8997, %8998) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9000 = tensor.empty() : tensor<1x64xf32> + %9001 = "ttir.relu"(%8999, %9000) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9002 = tensor.empty() : tensor<1x64xf32> + %9003 = "ttir.relu"(%9001, %9002) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9004 = tensor.empty() : tensor<1x64xf32> + %9005 = "ttir.relu"(%9003, %9004) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9006 = tensor.empty() : tensor<1x64xf32> + %9007 = "ttir.relu"(%9005, %9006) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9008 = tensor.empty() : tensor<1x64xf32> + %9009 = "ttir.relu"(%9007, %9008) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9010 = tensor.empty() : tensor<1x64xf32> + %9011 = "ttir.relu"(%9009, %9010) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9012 = tensor.empty() : tensor<1x64xf32> + %9013 = "ttir.relu"(%9011, %9012) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9014 = tensor.empty() : tensor<1x64xf32> + %9015 = "ttir.relu"(%9013, %9014) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9016 = tensor.empty() : tensor<1x64xf32> + %9017 = "ttir.relu"(%9015, %9016) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9018 = tensor.empty() : tensor<1x64xf32> + %9019 = "ttir.relu"(%9017, %9018) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9020 = tensor.empty() : tensor<1x64xf32> + %9021 = "ttir.relu"(%9019, %9020) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9022 = tensor.empty() : tensor<1x64xf32> + %9023 = "ttir.relu"(%9021, %9022) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9024 = tensor.empty() : tensor<1x64xf32> + %9025 = "ttir.relu"(%9023, %9024) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9026 = tensor.empty() : tensor<1x64xf32> + %9027 = "ttir.relu"(%9025, %9026) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9028 = tensor.empty() : tensor<1x64xf32> + %9029 = "ttir.relu"(%9027, %9028) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9030 = tensor.empty() : tensor<1x64xf32> + %9031 = "ttir.relu"(%9029, %9030) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9032 = tensor.empty() : tensor<1x64xf32> + %9033 = "ttir.relu"(%9031, %9032) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9034 = tensor.empty() : tensor<1x64xf32> + %9035 = "ttir.relu"(%9033, %9034) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9036 = tensor.empty() : tensor<1x64xf32> + %9037 = "ttir.relu"(%9035, %9036) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9038 = tensor.empty() : tensor<1x64xf32> + %9039 = "ttir.relu"(%9037, %9038) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9040 = tensor.empty() : tensor<1x64xf32> + %9041 = "ttir.relu"(%9039, %9040) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9042 = tensor.empty() : tensor<1x64xf32> + %9043 = "ttir.relu"(%9041, %9042) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9044 = tensor.empty() : tensor<1x64xf32> + %9045 = "ttir.relu"(%9043, %9044) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9046 = tensor.empty() : tensor<1x64xf32> + %9047 = "ttir.relu"(%9045, %9046) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9048 = tensor.empty() : tensor<1x64xf32> + %9049 = "ttir.relu"(%9047, %9048) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9050 = tensor.empty() : tensor<1x64xf32> + %9051 = "ttir.relu"(%9049, %9050) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9052 = tensor.empty() : tensor<1x64xf32> + %9053 = "ttir.relu"(%9051, %9052) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9054 = tensor.empty() : tensor<1x64xf32> + %9055 = "ttir.relu"(%9053, %9054) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9056 = tensor.empty() : tensor<1x64xf32> + %9057 = "ttir.relu"(%9055, %9056) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9058 = tensor.empty() : tensor<1x64xf32> + %9059 = "ttir.relu"(%9057, %9058) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9060 = tensor.empty() : tensor<1x64xf32> + %9061 = "ttir.relu"(%9059, %9060) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9062 = tensor.empty() : tensor<1x64xf32> + %9063 = "ttir.relu"(%9061, %9062) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9064 = tensor.empty() : tensor<1x64xf32> + %9065 = "ttir.relu"(%9063, %9064) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9066 = tensor.empty() : tensor<1x64xf32> + %9067 = "ttir.relu"(%9065, %9066) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9068 = tensor.empty() : tensor<1x64xf32> + %9069 = "ttir.relu"(%9067, %9068) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9070 = tensor.empty() : tensor<1x64xf32> + %9071 = "ttir.relu"(%9069, %9070) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9072 = tensor.empty() : tensor<1x64xf32> + %9073 = "ttir.relu"(%9071, %9072) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9074 = tensor.empty() : tensor<1x64xf32> + %9075 = "ttir.relu"(%9073, %9074) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9076 = tensor.empty() : tensor<1x64xf32> + %9077 = "ttir.relu"(%9075, %9076) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9078 = tensor.empty() : tensor<1x64xf32> + %9079 = "ttir.relu"(%9077, %9078) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9080 = tensor.empty() : tensor<1x64xf32> + %9081 = "ttir.relu"(%9079, %9080) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9082 = tensor.empty() : tensor<1x64xf32> + %9083 = "ttir.relu"(%9081, %9082) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9084 = tensor.empty() : tensor<1x64xf32> + %9085 = "ttir.relu"(%9083, %9084) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9086 = tensor.empty() : tensor<1x64xf32> + %9087 = "ttir.relu"(%9085, %9086) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9088 = tensor.empty() : tensor<1x64xf32> + %9089 = "ttir.relu"(%9087, %9088) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9090 = tensor.empty() : tensor<1x64xf32> + %9091 = "ttir.relu"(%9089, %9090) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9092 = tensor.empty() : tensor<1x64xf32> + %9093 = "ttir.relu"(%9091, %9092) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9094 = tensor.empty() : tensor<1x64xf32> + %9095 = "ttir.relu"(%9093, %9094) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9096 = tensor.empty() : tensor<1x64xf32> + %9097 = "ttir.relu"(%9095, %9096) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9098 = tensor.empty() : tensor<1x64xf32> + %9099 = "ttir.relu"(%9097, %9098) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9100 = tensor.empty() : tensor<1x64xf32> + %9101 = "ttir.relu"(%9099, %9100) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9102 = tensor.empty() : tensor<1x64xf32> + %9103 = "ttir.relu"(%9101, %9102) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9104 = tensor.empty() : tensor<1x64xf32> + %9105 = "ttir.relu"(%9103, %9104) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9106 = tensor.empty() : tensor<1x64xf32> + %9107 = "ttir.relu"(%9105, %9106) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9108 = tensor.empty() : tensor<1x64xf32> + %9109 = "ttir.relu"(%9107, %9108) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9110 = tensor.empty() : tensor<1x64xf32> + %9111 = "ttir.relu"(%9109, %9110) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9112 = tensor.empty() : tensor<1x64xf32> + %9113 = "ttir.relu"(%9111, %9112) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9114 = tensor.empty() : tensor<1x64xf32> + %9115 = "ttir.relu"(%9113, %9114) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9116 = tensor.empty() : tensor<1x64xf32> + %9117 = "ttir.relu"(%9115, %9116) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9118 = tensor.empty() : tensor<1x64xf32> + %9119 = "ttir.relu"(%9117, %9118) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9120 = tensor.empty() : tensor<1x64xf32> + %9121 = "ttir.relu"(%9119, %9120) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9122 = tensor.empty() : tensor<1x64xf32> + %9123 = "ttir.relu"(%9121, %9122) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9124 = tensor.empty() : tensor<1x64xf32> + %9125 = "ttir.relu"(%9123, %9124) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9126 = tensor.empty() : tensor<1x64xf32> + %9127 = "ttir.relu"(%9125, %9126) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9128 = tensor.empty() : tensor<1x64xf32> + %9129 = "ttir.relu"(%9127, %9128) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9130 = tensor.empty() : tensor<1x64xf32> + %9131 = "ttir.relu"(%9129, %9130) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9132 = tensor.empty() : tensor<1x64xf32> + %9133 = "ttir.relu"(%9131, %9132) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9134 = tensor.empty() : tensor<1x64xf32> + %9135 = "ttir.relu"(%9133, %9134) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9136 = tensor.empty() : tensor<1x64xf32> + %9137 = "ttir.relu"(%9135, %9136) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9138 = tensor.empty() : tensor<1x64xf32> + %9139 = "ttir.relu"(%9137, %9138) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9140 = tensor.empty() : tensor<1x64xf32> + %9141 = "ttir.relu"(%9139, %9140) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9142 = tensor.empty() : tensor<1x64xf32> + %9143 = "ttir.relu"(%9141, %9142) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9144 = tensor.empty() : tensor<1x64xf32> + %9145 = "ttir.relu"(%9143, %9144) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9146 = tensor.empty() : tensor<1x64xf32> + %9147 = "ttir.relu"(%9145, %9146) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9148 = tensor.empty() : tensor<1x64xf32> + %9149 = "ttir.relu"(%9147, %9148) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9150 = tensor.empty() : tensor<1x64xf32> + %9151 = "ttir.relu"(%9149, %9150) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9152 = tensor.empty() : tensor<1x64xf32> + %9153 = "ttir.relu"(%9151, %9152) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9154 = tensor.empty() : tensor<1x64xf32> + %9155 = "ttir.relu"(%9153, %9154) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9156 = tensor.empty() : tensor<1x64xf32> + %9157 = "ttir.relu"(%9155, %9156) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9158 = tensor.empty() : tensor<1x64xf32> + %9159 = "ttir.relu"(%9157, %9158) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9160 = tensor.empty() : tensor<1x64xf32> + %9161 = "ttir.relu"(%9159, %9160) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9162 = tensor.empty() : tensor<1x64xf32> + %9163 = "ttir.relu"(%9161, %9162) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9164 = tensor.empty() : tensor<1x64xf32> + %9165 = "ttir.relu"(%9163, %9164) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9166 = tensor.empty() : tensor<1x64xf32> + %9167 = "ttir.relu"(%9165, %9166) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9168 = tensor.empty() : tensor<1x64xf32> + %9169 = "ttir.relu"(%9167, %9168) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9170 = tensor.empty() : tensor<1x64xf32> + %9171 = "ttir.relu"(%9169, %9170) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9172 = tensor.empty() : tensor<1x64xf32> + %9173 = "ttir.relu"(%9171, %9172) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9174 = tensor.empty() : tensor<1x64xf32> + %9175 = "ttir.relu"(%9173, %9174) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9176 = tensor.empty() : tensor<1x64xf32> + %9177 = "ttir.relu"(%9175, %9176) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9178 = tensor.empty() : tensor<1x64xf32> + %9179 = "ttir.relu"(%9177, %9178) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9180 = tensor.empty() : tensor<1x64xf32> + %9181 = "ttir.relu"(%9179, %9180) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9182 = tensor.empty() : tensor<1x64xf32> + %9183 = "ttir.relu"(%9181, %9182) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9184 = tensor.empty() : tensor<1x64xf32> + %9185 = "ttir.relu"(%9183, %9184) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9186 = tensor.empty() : tensor<1x64xf32> + %9187 = "ttir.relu"(%9185, %9186) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9188 = tensor.empty() : tensor<1x64xf32> + %9189 = "ttir.relu"(%9187, %9188) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9190 = tensor.empty() : tensor<1x64xf32> + %9191 = "ttir.relu"(%9189, %9190) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9192 = tensor.empty() : tensor<1x64xf32> + %9193 = "ttir.relu"(%9191, %9192) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9194 = tensor.empty() : tensor<1x64xf32> + %9195 = "ttir.relu"(%9193, %9194) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9196 = tensor.empty() : tensor<1x64xf32> + %9197 = "ttir.relu"(%9195, %9196) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9198 = tensor.empty() : tensor<1x64xf32> + %9199 = "ttir.relu"(%9197, %9198) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9200 = tensor.empty() : tensor<1x64xf32> + %9201 = "ttir.relu"(%9199, %9200) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9202 = tensor.empty() : tensor<1x64xf32> + %9203 = "ttir.relu"(%9201, %9202) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9204 = tensor.empty() : tensor<1x64xf32> + %9205 = "ttir.relu"(%9203, %9204) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9206 = tensor.empty() : tensor<1x64xf32> + %9207 = "ttir.relu"(%9205, %9206) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9208 = tensor.empty() : tensor<1x64xf32> + %9209 = "ttir.relu"(%9207, %9208) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9210 = tensor.empty() : tensor<1x64xf32> + %9211 = "ttir.relu"(%9209, %9210) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9212 = tensor.empty() : tensor<1x64xf32> + %9213 = "ttir.relu"(%9211, %9212) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9214 = tensor.empty() : tensor<1x64xf32> + %9215 = "ttir.relu"(%9213, %9214) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9216 = tensor.empty() : tensor<1x64xf32> + %9217 = "ttir.relu"(%9215, %9216) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9218 = tensor.empty() : tensor<1x64xf32> + %9219 = "ttir.relu"(%9217, %9218) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9220 = tensor.empty() : tensor<1x64xf32> + %9221 = "ttir.relu"(%9219, %9220) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9222 = tensor.empty() : tensor<1x64xf32> + %9223 = "ttir.relu"(%9221, %9222) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9224 = tensor.empty() : tensor<1x64xf32> + %9225 = "ttir.relu"(%9223, %9224) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9226 = tensor.empty() : tensor<1x64xf32> + %9227 = "ttir.relu"(%9225, %9226) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9228 = tensor.empty() : tensor<1x64xf32> + %9229 = "ttir.relu"(%9227, %9228) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9230 = tensor.empty() : tensor<1x64xf32> + %9231 = "ttir.relu"(%9229, %9230) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9232 = tensor.empty() : tensor<1x64xf32> + %9233 = "ttir.relu"(%9231, %9232) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9234 = tensor.empty() : tensor<1x64xf32> + %9235 = "ttir.relu"(%9233, %9234) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9236 = tensor.empty() : tensor<1x64xf32> + %9237 = "ttir.relu"(%9235, %9236) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9238 = tensor.empty() : tensor<1x64xf32> + %9239 = "ttir.relu"(%9237, %9238) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9240 = tensor.empty() : tensor<1x64xf32> + %9241 = "ttir.relu"(%9239, %9240) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9242 = tensor.empty() : tensor<1x64xf32> + %9243 = "ttir.relu"(%9241, %9242) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9244 = tensor.empty() : tensor<1x64xf32> + %9245 = "ttir.relu"(%9243, %9244) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9246 = tensor.empty() : tensor<1x64xf32> + %9247 = "ttir.relu"(%9245, %9246) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9248 = tensor.empty() : tensor<1x64xf32> + %9249 = "ttir.relu"(%9247, %9248) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9250 = tensor.empty() : tensor<1x64xf32> + %9251 = "ttir.relu"(%9249, %9250) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9252 = tensor.empty() : tensor<1x64xf32> + %9253 = "ttir.relu"(%9251, %9252) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9254 = tensor.empty() : tensor<1x64xf32> + %9255 = "ttir.relu"(%9253, %9254) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9256 = tensor.empty() : tensor<1x64xf32> + %9257 = "ttir.relu"(%9255, %9256) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9258 = tensor.empty() : tensor<1x64xf32> + %9259 = "ttir.relu"(%9257, %9258) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9260 = tensor.empty() : tensor<1x64xf32> + %9261 = "ttir.relu"(%9259, %9260) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9262 = tensor.empty() : tensor<1x64xf32> + %9263 = "ttir.relu"(%9261, %9262) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9264 = tensor.empty() : tensor<1x64xf32> + %9265 = "ttir.relu"(%9263, %9264) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9266 = tensor.empty() : tensor<1x64xf32> + %9267 = "ttir.relu"(%9265, %9266) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9268 = tensor.empty() : tensor<1x64xf32> + %9269 = "ttir.relu"(%9267, %9268) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9270 = tensor.empty() : tensor<1x64xf32> + %9271 = "ttir.relu"(%9269, %9270) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9272 = tensor.empty() : tensor<1x64xf32> + %9273 = "ttir.relu"(%9271, %9272) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9274 = tensor.empty() : tensor<1x64xf32> + %9275 = "ttir.relu"(%9273, %9274) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9276 = tensor.empty() : tensor<1x64xf32> + %9277 = "ttir.relu"(%9275, %9276) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9278 = tensor.empty() : tensor<1x64xf32> + %9279 = "ttir.relu"(%9277, %9278) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9280 = tensor.empty() : tensor<1x64xf32> + %9281 = "ttir.relu"(%9279, %9280) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9282 = tensor.empty() : tensor<1x64xf32> + %9283 = "ttir.relu"(%9281, %9282) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9284 = tensor.empty() : tensor<1x64xf32> + %9285 = "ttir.relu"(%9283, %9284) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9286 = tensor.empty() : tensor<1x64xf32> + %9287 = "ttir.relu"(%9285, %9286) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9288 = tensor.empty() : tensor<1x64xf32> + %9289 = "ttir.relu"(%9287, %9288) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9290 = tensor.empty() : tensor<1x64xf32> + %9291 = "ttir.relu"(%9289, %9290) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9292 = tensor.empty() : tensor<1x64xf32> + %9293 = "ttir.relu"(%9291, %9292) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9294 = tensor.empty() : tensor<1x64xf32> + %9295 = "ttir.relu"(%9293, %9294) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9296 = tensor.empty() : tensor<1x64xf32> + %9297 = "ttir.relu"(%9295, %9296) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9298 = tensor.empty() : tensor<1x64xf32> + %9299 = "ttir.relu"(%9297, %9298) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9300 = tensor.empty() : tensor<1x64xf32> + %9301 = "ttir.relu"(%9299, %9300) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9302 = tensor.empty() : tensor<1x64xf32> + %9303 = "ttir.relu"(%9301, %9302) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9304 = tensor.empty() : tensor<1x64xf32> + %9305 = "ttir.relu"(%9303, %9304) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9306 = tensor.empty() : tensor<1x64xf32> + %9307 = "ttir.relu"(%9305, %9306) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9308 = tensor.empty() : tensor<1x64xf32> + %9309 = "ttir.relu"(%9307, %9308) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9310 = tensor.empty() : tensor<1x64xf32> + %9311 = "ttir.relu"(%9309, %9310) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9312 = tensor.empty() : tensor<1x64xf32> + %9313 = "ttir.relu"(%9311, %9312) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9314 = tensor.empty() : tensor<1x64xf32> + %9315 = "ttir.relu"(%9313, %9314) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9316 = tensor.empty() : tensor<1x64xf32> + %9317 = "ttir.relu"(%9315, %9316) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9318 = tensor.empty() : tensor<1x64xf32> + %9319 = "ttir.relu"(%9317, %9318) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9320 = tensor.empty() : tensor<1x64xf32> + %9321 = "ttir.relu"(%9319, %9320) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9322 = tensor.empty() : tensor<1x64xf32> + %9323 = "ttir.relu"(%9321, %9322) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9324 = tensor.empty() : tensor<1x64xf32> + %9325 = "ttir.relu"(%9323, %9324) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9326 = tensor.empty() : tensor<1x64xf32> + %9327 = "ttir.relu"(%9325, %9326) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9328 = tensor.empty() : tensor<1x64xf32> + %9329 = "ttir.relu"(%9327, %9328) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9330 = tensor.empty() : tensor<1x64xf32> + %9331 = "ttir.relu"(%9329, %9330) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9332 = tensor.empty() : tensor<1x64xf32> + %9333 = "ttir.relu"(%9331, %9332) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9334 = tensor.empty() : tensor<1x64xf32> + %9335 = "ttir.relu"(%9333, %9334) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9336 = tensor.empty() : tensor<1x64xf32> + %9337 = "ttir.relu"(%9335, %9336) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9338 = tensor.empty() : tensor<1x64xf32> + %9339 = "ttir.relu"(%9337, %9338) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9340 = tensor.empty() : tensor<1x64xf32> + %9341 = "ttir.relu"(%9339, %9340) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9342 = tensor.empty() : tensor<1x64xf32> + %9343 = "ttir.relu"(%9341, %9342) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9344 = tensor.empty() : tensor<1x64xf32> + %9345 = "ttir.relu"(%9343, %9344) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9346 = tensor.empty() : tensor<1x64xf32> + %9347 = "ttir.relu"(%9345, %9346) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9348 = tensor.empty() : tensor<1x64xf32> + %9349 = "ttir.relu"(%9347, %9348) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9350 = tensor.empty() : tensor<1x64xf32> + %9351 = "ttir.relu"(%9349, %9350) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9352 = tensor.empty() : tensor<1x64xf32> + %9353 = "ttir.relu"(%9351, %9352) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9354 = tensor.empty() : tensor<1x64xf32> + %9355 = "ttir.relu"(%9353, %9354) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9356 = tensor.empty() : tensor<1x64xf32> + %9357 = "ttir.relu"(%9355, %9356) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9358 = tensor.empty() : tensor<1x64xf32> + %9359 = "ttir.relu"(%9357, %9358) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9360 = tensor.empty() : tensor<1x64xf32> + %9361 = "ttir.relu"(%9359, %9360) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9362 = tensor.empty() : tensor<1x64xf32> + %9363 = "ttir.relu"(%9361, %9362) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9364 = tensor.empty() : tensor<1x64xf32> + %9365 = "ttir.relu"(%9363, %9364) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9366 = tensor.empty() : tensor<1x64xf32> + %9367 = "ttir.relu"(%9365, %9366) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9368 = tensor.empty() : tensor<1x64xf32> + %9369 = "ttir.relu"(%9367, %9368) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9370 = tensor.empty() : tensor<1x64xf32> + %9371 = "ttir.relu"(%9369, %9370) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9372 = tensor.empty() : tensor<1x64xf32> + %9373 = "ttir.relu"(%9371, %9372) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9374 = tensor.empty() : tensor<1x64xf32> + %9375 = "ttir.relu"(%9373, %9374) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9376 = tensor.empty() : tensor<1x64xf32> + %9377 = "ttir.relu"(%9375, %9376) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9378 = tensor.empty() : tensor<1x64xf32> + %9379 = "ttir.relu"(%9377, %9378) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9380 = tensor.empty() : tensor<1x64xf32> + %9381 = "ttir.relu"(%9379, %9380) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9382 = tensor.empty() : tensor<1x64xf32> + %9383 = "ttir.relu"(%9381, %9382) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9384 = tensor.empty() : tensor<1x64xf32> + %9385 = "ttir.relu"(%9383, %9384) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9386 = tensor.empty() : tensor<1x64xf32> + %9387 = "ttir.relu"(%9385, %9386) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9388 = tensor.empty() : tensor<1x64xf32> + %9389 = "ttir.relu"(%9387, %9388) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9390 = tensor.empty() : tensor<1x64xf32> + %9391 = "ttir.relu"(%9389, %9390) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9392 = tensor.empty() : tensor<1x64xf32> + %9393 = "ttir.relu"(%9391, %9392) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9394 = tensor.empty() : tensor<1x64xf32> + %9395 = "ttir.relu"(%9393, %9394) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9396 = tensor.empty() : tensor<1x64xf32> + %9397 = "ttir.relu"(%9395, %9396) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9398 = tensor.empty() : tensor<1x64xf32> + %9399 = "ttir.relu"(%9397, %9398) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9400 = tensor.empty() : tensor<1x64xf32> + %9401 = "ttir.relu"(%9399, %9400) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9402 = tensor.empty() : tensor<1x64xf32> + %9403 = "ttir.relu"(%9401, %9402) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9404 = tensor.empty() : tensor<1x64xf32> + %9405 = "ttir.relu"(%9403, %9404) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9406 = tensor.empty() : tensor<1x64xf32> + %9407 = "ttir.relu"(%9405, %9406) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9408 = tensor.empty() : tensor<1x64xf32> + %9409 = "ttir.relu"(%9407, %9408) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9410 = tensor.empty() : tensor<1x64xf32> + %9411 = "ttir.relu"(%9409, %9410) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9412 = tensor.empty() : tensor<1x64xf32> + %9413 = "ttir.relu"(%9411, %9412) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9414 = tensor.empty() : tensor<1x64xf32> + %9415 = "ttir.relu"(%9413, %9414) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9416 = tensor.empty() : tensor<1x64xf32> + %9417 = "ttir.relu"(%9415, %9416) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9418 = tensor.empty() : tensor<1x64xf32> + %9419 = "ttir.relu"(%9417, %9418) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9420 = tensor.empty() : tensor<1x64xf32> + %9421 = "ttir.relu"(%9419, %9420) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9422 = tensor.empty() : tensor<1x64xf32> + %9423 = "ttir.relu"(%9421, %9422) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9424 = tensor.empty() : tensor<1x64xf32> + %9425 = "ttir.relu"(%9423, %9424) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9426 = tensor.empty() : tensor<1x64xf32> + %9427 = "ttir.relu"(%9425, %9426) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9428 = tensor.empty() : tensor<1x64xf32> + %9429 = "ttir.relu"(%9427, %9428) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9430 = tensor.empty() : tensor<1x64xf32> + %9431 = "ttir.relu"(%9429, %9430) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9432 = tensor.empty() : tensor<1x64xf32> + %9433 = "ttir.relu"(%9431, %9432) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9434 = tensor.empty() : tensor<1x64xf32> + %9435 = "ttir.relu"(%9433, %9434) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9436 = tensor.empty() : tensor<1x64xf32> + %9437 = "ttir.relu"(%9435, %9436) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9438 = tensor.empty() : tensor<1x64xf32> + %9439 = "ttir.relu"(%9437, %9438) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9440 = tensor.empty() : tensor<1x64xf32> + %9441 = "ttir.relu"(%9439, %9440) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9442 = tensor.empty() : tensor<1x64xf32> + %9443 = "ttir.relu"(%9441, %9442) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9444 = tensor.empty() : tensor<1x64xf32> + %9445 = "ttir.relu"(%9443, %9444) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9446 = tensor.empty() : tensor<1x64xf32> + %9447 = "ttir.relu"(%9445, %9446) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9448 = tensor.empty() : tensor<1x64xf32> + %9449 = "ttir.relu"(%9447, %9448) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9450 = tensor.empty() : tensor<1x64xf32> + %9451 = "ttir.relu"(%9449, %9450) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9452 = tensor.empty() : tensor<1x64xf32> + %9453 = "ttir.relu"(%9451, %9452) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9454 = tensor.empty() : tensor<1x64xf32> + %9455 = "ttir.relu"(%9453, %9454) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9456 = tensor.empty() : tensor<1x64xf32> + %9457 = "ttir.relu"(%9455, %9456) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9458 = tensor.empty() : tensor<1x64xf32> + %9459 = "ttir.relu"(%9457, %9458) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9460 = tensor.empty() : tensor<1x64xf32> + %9461 = "ttir.relu"(%9459, %9460) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9462 = tensor.empty() : tensor<1x64xf32> + %9463 = "ttir.relu"(%9461, %9462) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9464 = tensor.empty() : tensor<1x64xf32> + %9465 = "ttir.relu"(%9463, %9464) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9466 = tensor.empty() : tensor<1x64xf32> + %9467 = "ttir.relu"(%9465, %9466) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9468 = tensor.empty() : tensor<1x64xf32> + %9469 = "ttir.relu"(%9467, %9468) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9470 = tensor.empty() : tensor<1x64xf32> + %9471 = "ttir.relu"(%9469, %9470) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9472 = tensor.empty() : tensor<1x64xf32> + %9473 = "ttir.relu"(%9471, %9472) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9474 = tensor.empty() : tensor<1x64xf32> + %9475 = "ttir.relu"(%9473, %9474) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9476 = tensor.empty() : tensor<1x64xf32> + %9477 = "ttir.relu"(%9475, %9476) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9478 = tensor.empty() : tensor<1x64xf32> + %9479 = "ttir.relu"(%9477, %9478) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9480 = tensor.empty() : tensor<1x64xf32> + %9481 = "ttir.relu"(%9479, %9480) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9482 = tensor.empty() : tensor<1x64xf32> + %9483 = "ttir.relu"(%9481, %9482) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9484 = tensor.empty() : tensor<1x64xf32> + %9485 = "ttir.relu"(%9483, %9484) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9486 = tensor.empty() : tensor<1x64xf32> + %9487 = "ttir.relu"(%9485, %9486) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9488 = tensor.empty() : tensor<1x64xf32> + %9489 = "ttir.relu"(%9487, %9488) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9490 = tensor.empty() : tensor<1x64xf32> + %9491 = "ttir.relu"(%9489, %9490) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9492 = tensor.empty() : tensor<1x64xf32> + %9493 = "ttir.relu"(%9491, %9492) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9494 = tensor.empty() : tensor<1x64xf32> + %9495 = "ttir.relu"(%9493, %9494) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9496 = tensor.empty() : tensor<1x64xf32> + %9497 = "ttir.relu"(%9495, %9496) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9498 = tensor.empty() : tensor<1x64xf32> + %9499 = "ttir.relu"(%9497, %9498) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9500 = tensor.empty() : tensor<1x64xf32> + %9501 = "ttir.relu"(%9499, %9500) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9502 = tensor.empty() : tensor<1x64xf32> + %9503 = "ttir.relu"(%9501, %9502) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9504 = tensor.empty() : tensor<1x64xf32> + %9505 = "ttir.relu"(%9503, %9504) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9506 = tensor.empty() : tensor<1x64xf32> + %9507 = "ttir.relu"(%9505, %9506) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9508 = tensor.empty() : tensor<1x64xf32> + %9509 = "ttir.relu"(%9507, %9508) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9510 = tensor.empty() : tensor<1x64xf32> + %9511 = "ttir.relu"(%9509, %9510) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9512 = tensor.empty() : tensor<1x64xf32> + %9513 = "ttir.relu"(%9511, %9512) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9514 = tensor.empty() : tensor<1x64xf32> + %9515 = "ttir.relu"(%9513, %9514) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9516 = tensor.empty() : tensor<1x64xf32> + %9517 = "ttir.relu"(%9515, %9516) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9518 = tensor.empty() : tensor<1x64xf32> + %9519 = "ttir.relu"(%9517, %9518) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9520 = tensor.empty() : tensor<1x64xf32> + %9521 = "ttir.relu"(%9519, %9520) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9522 = tensor.empty() : tensor<1x64xf32> + %9523 = "ttir.relu"(%9521, %9522) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9524 = tensor.empty() : tensor<1x64xf32> + %9525 = "ttir.relu"(%9523, %9524) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9526 = tensor.empty() : tensor<1x64xf32> + %9527 = "ttir.relu"(%9525, %9526) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9528 = tensor.empty() : tensor<1x64xf32> + %9529 = "ttir.relu"(%9527, %9528) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9530 = tensor.empty() : tensor<1x64xf32> + %9531 = "ttir.relu"(%9529, %9530) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9532 = tensor.empty() : tensor<1x64xf32> + %9533 = "ttir.relu"(%9531, %9532) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9534 = tensor.empty() : tensor<1x64xf32> + %9535 = "ttir.relu"(%9533, %9534) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9536 = tensor.empty() : tensor<1x64xf32> + %9537 = "ttir.relu"(%9535, %9536) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9538 = tensor.empty() : tensor<1x64xf32> + %9539 = "ttir.relu"(%9537, %9538) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9540 = tensor.empty() : tensor<1x64xf32> + %9541 = "ttir.relu"(%9539, %9540) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9542 = tensor.empty() : tensor<1x64xf32> + %9543 = "ttir.relu"(%9541, %9542) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9544 = tensor.empty() : tensor<1x64xf32> + %9545 = "ttir.relu"(%9543, %9544) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9546 = tensor.empty() : tensor<1x64xf32> + %9547 = "ttir.relu"(%9545, %9546) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9548 = tensor.empty() : tensor<1x64xf32> + %9549 = "ttir.relu"(%9547, %9548) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9550 = tensor.empty() : tensor<1x64xf32> + %9551 = "ttir.relu"(%9549, %9550) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9552 = tensor.empty() : tensor<1x64xf32> + %9553 = "ttir.relu"(%9551, %9552) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9554 = tensor.empty() : tensor<1x64xf32> + %9555 = "ttir.relu"(%9553, %9554) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9556 = tensor.empty() : tensor<1x64xf32> + %9557 = "ttir.relu"(%9555, %9556) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9558 = tensor.empty() : tensor<1x64xf32> + %9559 = "ttir.relu"(%9557, %9558) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9560 = tensor.empty() : tensor<1x64xf32> + %9561 = "ttir.relu"(%9559, %9560) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9562 = tensor.empty() : tensor<1x64xf32> + %9563 = "ttir.relu"(%9561, %9562) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9564 = tensor.empty() : tensor<1x64xf32> + %9565 = "ttir.relu"(%9563, %9564) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9566 = tensor.empty() : tensor<1x64xf32> + %9567 = "ttir.relu"(%9565, %9566) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9568 = tensor.empty() : tensor<1x64xf32> + %9569 = "ttir.relu"(%9567, %9568) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9570 = tensor.empty() : tensor<1x64xf32> + %9571 = "ttir.relu"(%9569, %9570) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9572 = tensor.empty() : tensor<1x64xf32> + %9573 = "ttir.relu"(%9571, %9572) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9574 = tensor.empty() : tensor<1x64xf32> + %9575 = "ttir.relu"(%9573, %9574) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9576 = tensor.empty() : tensor<1x64xf32> + %9577 = "ttir.relu"(%9575, %9576) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9578 = tensor.empty() : tensor<1x64xf32> + %9579 = "ttir.relu"(%9577, %9578) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9580 = tensor.empty() : tensor<1x64xf32> + %9581 = "ttir.relu"(%9579, %9580) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9582 = tensor.empty() : tensor<1x64xf32> + %9583 = "ttir.relu"(%9581, %9582) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9584 = tensor.empty() : tensor<1x64xf32> + %9585 = "ttir.relu"(%9583, %9584) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9586 = tensor.empty() : tensor<1x64xf32> + %9587 = "ttir.relu"(%9585, %9586) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9588 = tensor.empty() : tensor<1x64xf32> + %9589 = "ttir.relu"(%9587, %9588) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9590 = tensor.empty() : tensor<1x64xf32> + %9591 = "ttir.relu"(%9589, %9590) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9592 = tensor.empty() : tensor<1x64xf32> + %9593 = "ttir.relu"(%9591, %9592) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9594 = tensor.empty() : tensor<1x64xf32> + %9595 = "ttir.relu"(%9593, %9594) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9596 = tensor.empty() : tensor<1x64xf32> + %9597 = "ttir.relu"(%9595, %9596) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9598 = tensor.empty() : tensor<1x64xf32> + %9599 = "ttir.relu"(%9597, %9598) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9600 = tensor.empty() : tensor<1x64xf32> + %9601 = "ttir.relu"(%9599, %9600) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9602 = tensor.empty() : tensor<1x64xf32> + %9603 = "ttir.relu"(%9601, %9602) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9604 = tensor.empty() : tensor<1x64xf32> + %9605 = "ttir.relu"(%9603, %9604) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9606 = tensor.empty() : tensor<1x64xf32> + %9607 = "ttir.relu"(%9605, %9606) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9608 = tensor.empty() : tensor<1x64xf32> + %9609 = "ttir.relu"(%9607, %9608) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9610 = tensor.empty() : tensor<1x64xf32> + %9611 = "ttir.relu"(%9609, %9610) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9612 = tensor.empty() : tensor<1x64xf32> + %9613 = "ttir.relu"(%9611, %9612) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9614 = tensor.empty() : tensor<1x64xf32> + %9615 = "ttir.relu"(%9613, %9614) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9616 = tensor.empty() : tensor<1x64xf32> + %9617 = "ttir.relu"(%9615, %9616) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9618 = tensor.empty() : tensor<1x64xf32> + %9619 = "ttir.relu"(%9617, %9618) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9620 = tensor.empty() : tensor<1x64xf32> + %9621 = "ttir.relu"(%9619, %9620) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9622 = tensor.empty() : tensor<1x64xf32> + %9623 = "ttir.relu"(%9621, %9622) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9624 = tensor.empty() : tensor<1x64xf32> + %9625 = "ttir.relu"(%9623, %9624) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9626 = tensor.empty() : tensor<1x64xf32> + %9627 = "ttir.relu"(%9625, %9626) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9628 = tensor.empty() : tensor<1x64xf32> + %9629 = "ttir.relu"(%9627, %9628) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9630 = tensor.empty() : tensor<1x64xf32> + %9631 = "ttir.relu"(%9629, %9630) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9632 = tensor.empty() : tensor<1x64xf32> + %9633 = "ttir.relu"(%9631, %9632) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9634 = tensor.empty() : tensor<1x64xf32> + %9635 = "ttir.relu"(%9633, %9634) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9636 = tensor.empty() : tensor<1x64xf32> + %9637 = "ttir.relu"(%9635, %9636) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9638 = tensor.empty() : tensor<1x64xf32> + %9639 = "ttir.relu"(%9637, %9638) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9640 = tensor.empty() : tensor<1x64xf32> + %9641 = "ttir.relu"(%9639, %9640) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9642 = tensor.empty() : tensor<1x64xf32> + %9643 = "ttir.relu"(%9641, %9642) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9644 = tensor.empty() : tensor<1x64xf32> + %9645 = "ttir.relu"(%9643, %9644) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9646 = tensor.empty() : tensor<1x64xf32> + %9647 = "ttir.relu"(%9645, %9646) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9648 = tensor.empty() : tensor<1x64xf32> + %9649 = "ttir.relu"(%9647, %9648) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9650 = tensor.empty() : tensor<1x64xf32> + %9651 = "ttir.relu"(%9649, %9650) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9652 = tensor.empty() : tensor<1x64xf32> + %9653 = "ttir.relu"(%9651, %9652) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9654 = tensor.empty() : tensor<1x64xf32> + %9655 = "ttir.relu"(%9653, %9654) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9656 = tensor.empty() : tensor<1x64xf32> + %9657 = "ttir.relu"(%9655, %9656) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9658 = tensor.empty() : tensor<1x64xf32> + %9659 = "ttir.relu"(%9657, %9658) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9660 = tensor.empty() : tensor<1x64xf32> + %9661 = "ttir.relu"(%9659, %9660) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9662 = tensor.empty() : tensor<1x64xf32> + %9663 = "ttir.relu"(%9661, %9662) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9664 = tensor.empty() : tensor<1x64xf32> + %9665 = "ttir.relu"(%9663, %9664) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9666 = tensor.empty() : tensor<1x64xf32> + %9667 = "ttir.relu"(%9665, %9666) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9668 = tensor.empty() : tensor<1x64xf32> + %9669 = "ttir.relu"(%9667, %9668) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9670 = tensor.empty() : tensor<1x64xf32> + %9671 = "ttir.relu"(%9669, %9670) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9672 = tensor.empty() : tensor<1x64xf32> + %9673 = "ttir.relu"(%9671, %9672) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9674 = tensor.empty() : tensor<1x64xf32> + %9675 = "ttir.relu"(%9673, %9674) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9676 = tensor.empty() : tensor<1x64xf32> + %9677 = "ttir.relu"(%9675, %9676) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9678 = tensor.empty() : tensor<1x64xf32> + %9679 = "ttir.relu"(%9677, %9678) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9680 = tensor.empty() : tensor<1x64xf32> + %9681 = "ttir.relu"(%9679, %9680) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9682 = tensor.empty() : tensor<1x64xf32> + %9683 = "ttir.relu"(%9681, %9682) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9684 = tensor.empty() : tensor<1x64xf32> + %9685 = "ttir.relu"(%9683, %9684) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9686 = tensor.empty() : tensor<1x64xf32> + %9687 = "ttir.relu"(%9685, %9686) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9688 = tensor.empty() : tensor<1x64xf32> + %9689 = "ttir.relu"(%9687, %9688) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9690 = tensor.empty() : tensor<1x64xf32> + %9691 = "ttir.relu"(%9689, %9690) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9692 = tensor.empty() : tensor<1x64xf32> + %9693 = "ttir.relu"(%9691, %9692) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9694 = tensor.empty() : tensor<1x64xf32> + %9695 = "ttir.relu"(%9693, %9694) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9696 = tensor.empty() : tensor<1x64xf32> + %9697 = "ttir.relu"(%9695, %9696) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9698 = tensor.empty() : tensor<1x64xf32> + %9699 = "ttir.relu"(%9697, %9698) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9700 = tensor.empty() : tensor<1x64xf32> + %9701 = "ttir.relu"(%9699, %9700) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9702 = tensor.empty() : tensor<1x64xf32> + %9703 = "ttir.relu"(%9701, %9702) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9704 = tensor.empty() : tensor<1x64xf32> + %9705 = "ttir.relu"(%9703, %9704) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9706 = tensor.empty() : tensor<1x64xf32> + %9707 = "ttir.relu"(%9705, %9706) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9708 = tensor.empty() : tensor<1x64xf32> + %9709 = "ttir.relu"(%9707, %9708) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9710 = tensor.empty() : tensor<1x64xf32> + %9711 = "ttir.relu"(%9709, %9710) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9712 = tensor.empty() : tensor<1x64xf32> + %9713 = "ttir.relu"(%9711, %9712) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9714 = tensor.empty() : tensor<1x64xf32> + %9715 = "ttir.relu"(%9713, %9714) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9716 = tensor.empty() : tensor<1x64xf32> + %9717 = "ttir.relu"(%9715, %9716) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9718 = tensor.empty() : tensor<1x64xf32> + %9719 = "ttir.relu"(%9717, %9718) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9720 = tensor.empty() : tensor<1x64xf32> + %9721 = "ttir.relu"(%9719, %9720) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9722 = tensor.empty() : tensor<1x64xf32> + %9723 = "ttir.relu"(%9721, %9722) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9724 = tensor.empty() : tensor<1x64xf32> + %9725 = "ttir.relu"(%9723, %9724) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9726 = tensor.empty() : tensor<1x64xf32> + %9727 = "ttir.relu"(%9725, %9726) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9728 = tensor.empty() : tensor<1x64xf32> + %9729 = "ttir.relu"(%9727, %9728) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9730 = tensor.empty() : tensor<1x64xf32> + %9731 = "ttir.relu"(%9729, %9730) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9732 = tensor.empty() : tensor<1x64xf32> + %9733 = "ttir.relu"(%9731, %9732) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9734 = tensor.empty() : tensor<1x64xf32> + %9735 = "ttir.relu"(%9733, %9734) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9736 = tensor.empty() : tensor<1x64xf32> + %9737 = "ttir.relu"(%9735, %9736) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9738 = tensor.empty() : tensor<1x64xf32> + %9739 = "ttir.relu"(%9737, %9738) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9740 = tensor.empty() : tensor<1x64xf32> + %9741 = "ttir.relu"(%9739, %9740) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9742 = tensor.empty() : tensor<1x64xf32> + %9743 = "ttir.relu"(%9741, %9742) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9744 = tensor.empty() : tensor<1x64xf32> + %9745 = "ttir.relu"(%9743, %9744) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9746 = tensor.empty() : tensor<1x64xf32> + %9747 = "ttir.relu"(%9745, %9746) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9748 = tensor.empty() : tensor<1x64xf32> + %9749 = "ttir.relu"(%9747, %9748) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9750 = tensor.empty() : tensor<1x64xf32> + %9751 = "ttir.relu"(%9749, %9750) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9752 = tensor.empty() : tensor<1x64xf32> + %9753 = "ttir.relu"(%9751, %9752) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9754 = tensor.empty() : tensor<1x64xf32> + %9755 = "ttir.relu"(%9753, %9754) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9756 = tensor.empty() : tensor<1x64xf32> + %9757 = "ttir.relu"(%9755, %9756) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9758 = tensor.empty() : tensor<1x64xf32> + %9759 = "ttir.relu"(%9757, %9758) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9760 = tensor.empty() : tensor<1x64xf32> + %9761 = "ttir.relu"(%9759, %9760) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9762 = tensor.empty() : tensor<1x64xf32> + %9763 = "ttir.relu"(%9761, %9762) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9764 = tensor.empty() : tensor<1x64xf32> + %9765 = "ttir.relu"(%9763, %9764) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9766 = tensor.empty() : tensor<1x64xf32> + %9767 = "ttir.relu"(%9765, %9766) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9768 = tensor.empty() : tensor<1x64xf32> + %9769 = "ttir.relu"(%9767, %9768) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9770 = tensor.empty() : tensor<1x64xf32> + %9771 = "ttir.relu"(%9769, %9770) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9772 = tensor.empty() : tensor<1x64xf32> + %9773 = "ttir.relu"(%9771, %9772) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9774 = tensor.empty() : tensor<1x64xf32> + %9775 = "ttir.relu"(%9773, %9774) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9776 = tensor.empty() : tensor<1x64xf32> + %9777 = "ttir.relu"(%9775, %9776) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9778 = tensor.empty() : tensor<1x64xf32> + %9779 = "ttir.relu"(%9777, %9778) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9780 = tensor.empty() : tensor<1x64xf32> + %9781 = "ttir.relu"(%9779, %9780) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9782 = tensor.empty() : tensor<1x64xf32> + %9783 = "ttir.relu"(%9781, %9782) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9784 = tensor.empty() : tensor<1x64xf32> + %9785 = "ttir.relu"(%9783, %9784) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9786 = tensor.empty() : tensor<1x64xf32> + %9787 = "ttir.relu"(%9785, %9786) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9788 = tensor.empty() : tensor<1x64xf32> + %9789 = "ttir.relu"(%9787, %9788) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9790 = tensor.empty() : tensor<1x64xf32> + %9791 = "ttir.relu"(%9789, %9790) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9792 = tensor.empty() : tensor<1x64xf32> + %9793 = "ttir.relu"(%9791, %9792) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9794 = tensor.empty() : tensor<1x64xf32> + %9795 = "ttir.relu"(%9793, %9794) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9796 = tensor.empty() : tensor<1x64xf32> + %9797 = "ttir.relu"(%9795, %9796) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9798 = tensor.empty() : tensor<1x64xf32> + %9799 = "ttir.relu"(%9797, %9798) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9800 = tensor.empty() : tensor<1x64xf32> + %9801 = "ttir.relu"(%9799, %9800) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9802 = tensor.empty() : tensor<1x64xf32> + %9803 = "ttir.relu"(%9801, %9802) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9804 = tensor.empty() : tensor<1x64xf32> + %9805 = "ttir.relu"(%9803, %9804) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9806 = tensor.empty() : tensor<1x64xf32> + %9807 = "ttir.relu"(%9805, %9806) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9808 = tensor.empty() : tensor<1x64xf32> + %9809 = "ttir.relu"(%9807, %9808) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9810 = tensor.empty() : tensor<1x64xf32> + %9811 = "ttir.relu"(%9809, %9810) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9812 = tensor.empty() : tensor<1x64xf32> + %9813 = "ttir.relu"(%9811, %9812) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9814 = tensor.empty() : tensor<1x64xf32> + %9815 = "ttir.relu"(%9813, %9814) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9816 = tensor.empty() : tensor<1x64xf32> + %9817 = "ttir.relu"(%9815, %9816) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9818 = tensor.empty() : tensor<1x64xf32> + %9819 = "ttir.relu"(%9817, %9818) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9820 = tensor.empty() : tensor<1x64xf32> + %9821 = "ttir.relu"(%9819, %9820) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9822 = tensor.empty() : tensor<1x64xf32> + %9823 = "ttir.relu"(%9821, %9822) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9824 = tensor.empty() : tensor<1x64xf32> + %9825 = "ttir.relu"(%9823, %9824) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9826 = tensor.empty() : tensor<1x64xf32> + %9827 = "ttir.relu"(%9825, %9826) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9828 = tensor.empty() : tensor<1x64xf32> + %9829 = "ttir.relu"(%9827, %9828) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9830 = tensor.empty() : tensor<1x64xf32> + %9831 = "ttir.relu"(%9829, %9830) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9832 = tensor.empty() : tensor<1x64xf32> + %9833 = "ttir.relu"(%9831, %9832) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9834 = tensor.empty() : tensor<1x64xf32> + %9835 = "ttir.relu"(%9833, %9834) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9836 = tensor.empty() : tensor<1x64xf32> + %9837 = "ttir.relu"(%9835, %9836) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9838 = tensor.empty() : tensor<1x64xf32> + %9839 = "ttir.relu"(%9837, %9838) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9840 = tensor.empty() : tensor<1x64xf32> + %9841 = "ttir.relu"(%9839, %9840) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9842 = tensor.empty() : tensor<1x64xf32> + %9843 = "ttir.relu"(%9841, %9842) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9844 = tensor.empty() : tensor<1x64xf32> + %9845 = "ttir.relu"(%9843, %9844) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9846 = tensor.empty() : tensor<1x64xf32> + %9847 = "ttir.relu"(%9845, %9846) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9848 = tensor.empty() : tensor<1x64xf32> + %9849 = "ttir.relu"(%9847, %9848) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9850 = tensor.empty() : tensor<1x64xf32> + %9851 = "ttir.relu"(%9849, %9850) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9852 = tensor.empty() : tensor<1x64xf32> + %9853 = "ttir.relu"(%9851, %9852) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9854 = tensor.empty() : tensor<1x64xf32> + %9855 = "ttir.relu"(%9853, %9854) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9856 = tensor.empty() : tensor<1x64xf32> + %9857 = "ttir.relu"(%9855, %9856) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9858 = tensor.empty() : tensor<1x64xf32> + %9859 = "ttir.relu"(%9857, %9858) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9860 = tensor.empty() : tensor<1x64xf32> + %9861 = "ttir.relu"(%9859, %9860) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9862 = tensor.empty() : tensor<1x64xf32> + %9863 = "ttir.relu"(%9861, %9862) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9864 = tensor.empty() : tensor<1x64xf32> + %9865 = "ttir.relu"(%9863, %9864) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9866 = tensor.empty() : tensor<1x64xf32> + %9867 = "ttir.relu"(%9865, %9866) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9868 = tensor.empty() : tensor<1x64xf32> + %9869 = "ttir.relu"(%9867, %9868) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9870 = tensor.empty() : tensor<1x64xf32> + %9871 = "ttir.relu"(%9869, %9870) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9872 = tensor.empty() : tensor<1x64xf32> + %9873 = "ttir.relu"(%9871, %9872) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9874 = tensor.empty() : tensor<1x64xf32> + %9875 = "ttir.relu"(%9873, %9874) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9876 = tensor.empty() : tensor<1x64xf32> + %9877 = "ttir.relu"(%9875, %9876) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9878 = tensor.empty() : tensor<1x64xf32> + %9879 = "ttir.relu"(%9877, %9878) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9880 = tensor.empty() : tensor<1x64xf32> + %9881 = "ttir.relu"(%9879, %9880) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9882 = tensor.empty() : tensor<1x64xf32> + %9883 = "ttir.relu"(%9881, %9882) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9884 = tensor.empty() : tensor<1x64xf32> + %9885 = "ttir.relu"(%9883, %9884) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9886 = tensor.empty() : tensor<1x64xf32> + %9887 = "ttir.relu"(%9885, %9886) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9888 = tensor.empty() : tensor<1x64xf32> + %9889 = "ttir.relu"(%9887, %9888) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9890 = tensor.empty() : tensor<1x64xf32> + %9891 = "ttir.relu"(%9889, %9890) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9892 = tensor.empty() : tensor<1x64xf32> + %9893 = "ttir.relu"(%9891, %9892) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9894 = tensor.empty() : tensor<1x64xf32> + %9895 = "ttir.relu"(%9893, %9894) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9896 = tensor.empty() : tensor<1x64xf32> + %9897 = "ttir.relu"(%9895, %9896) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9898 = tensor.empty() : tensor<1x64xf32> + %9899 = "ttir.relu"(%9897, %9898) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9900 = tensor.empty() : tensor<1x64xf32> + %9901 = "ttir.relu"(%9899, %9900) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9902 = tensor.empty() : tensor<1x64xf32> + %9903 = "ttir.relu"(%9901, %9902) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9904 = tensor.empty() : tensor<1x64xf32> + %9905 = "ttir.relu"(%9903, %9904) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9906 = tensor.empty() : tensor<1x64xf32> + %9907 = "ttir.relu"(%9905, %9906) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9908 = tensor.empty() : tensor<1x64xf32> + %9909 = "ttir.relu"(%9907, %9908) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9910 = tensor.empty() : tensor<1x64xf32> + %9911 = "ttir.relu"(%9909, %9910) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9912 = tensor.empty() : tensor<1x64xf32> + %9913 = "ttir.relu"(%9911, %9912) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9914 = tensor.empty() : tensor<1x64xf32> + %9915 = "ttir.relu"(%9913, %9914) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9916 = tensor.empty() : tensor<1x64xf32> + %9917 = "ttir.relu"(%9915, %9916) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9918 = tensor.empty() : tensor<1x64xf32> + %9919 = "ttir.relu"(%9917, %9918) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9920 = tensor.empty() : tensor<1x64xf32> + %9921 = "ttir.relu"(%9919, %9920) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9922 = tensor.empty() : tensor<1x64xf32> + %9923 = "ttir.relu"(%9921, %9922) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9924 = tensor.empty() : tensor<1x64xf32> + %9925 = "ttir.relu"(%9923, %9924) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9926 = tensor.empty() : tensor<1x64xf32> + %9927 = "ttir.relu"(%9925, %9926) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9928 = tensor.empty() : tensor<1x64xf32> + %9929 = "ttir.relu"(%9927, %9928) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9930 = tensor.empty() : tensor<1x64xf32> + %9931 = "ttir.relu"(%9929, %9930) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9932 = tensor.empty() : tensor<1x64xf32> + %9933 = "ttir.relu"(%9931, %9932) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9934 = tensor.empty() : tensor<1x64xf32> + %9935 = "ttir.relu"(%9933, %9934) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9936 = tensor.empty() : tensor<1x64xf32> + %9937 = "ttir.relu"(%9935, %9936) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9938 = tensor.empty() : tensor<1x64xf32> + %9939 = "ttir.relu"(%9937, %9938) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9940 = tensor.empty() : tensor<1x64xf32> + %9941 = "ttir.relu"(%9939, %9940) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9942 = tensor.empty() : tensor<1x64xf32> + %9943 = "ttir.relu"(%9941, %9942) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9944 = tensor.empty() : tensor<1x64xf32> + %9945 = "ttir.relu"(%9943, %9944) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9946 = tensor.empty() : tensor<1x64xf32> + %9947 = "ttir.relu"(%9945, %9946) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9948 = tensor.empty() : tensor<1x64xf32> + %9949 = "ttir.relu"(%9947, %9948) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9950 = tensor.empty() : tensor<1x64xf32> + %9951 = "ttir.relu"(%9949, %9950) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9952 = tensor.empty() : tensor<1x64xf32> + %9953 = "ttir.relu"(%9951, %9952) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9954 = tensor.empty() : tensor<1x64xf32> + %9955 = "ttir.relu"(%9953, %9954) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9956 = tensor.empty() : tensor<1x64xf32> + %9957 = "ttir.relu"(%9955, %9956) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9958 = tensor.empty() : tensor<1x64xf32> + %9959 = "ttir.relu"(%9957, %9958) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9960 = tensor.empty() : tensor<1x64xf32> + %9961 = "ttir.relu"(%9959, %9960) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9962 = tensor.empty() : tensor<1x64xf32> + %9963 = "ttir.relu"(%9961, %9962) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9964 = tensor.empty() : tensor<1x64xf32> + %9965 = "ttir.relu"(%9963, %9964) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9966 = tensor.empty() : tensor<1x64xf32> + %9967 = "ttir.relu"(%9965, %9966) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9968 = tensor.empty() : tensor<1x64xf32> + %9969 = "ttir.relu"(%9967, %9968) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9970 = tensor.empty() : tensor<1x64xf32> + %9971 = "ttir.relu"(%9969, %9970) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9972 = tensor.empty() : tensor<1x64xf32> + %9973 = "ttir.relu"(%9971, %9972) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9974 = tensor.empty() : tensor<1x64xf32> + %9975 = "ttir.relu"(%9973, %9974) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9976 = tensor.empty() : tensor<1x64xf32> + %9977 = "ttir.relu"(%9975, %9976) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9978 = tensor.empty() : tensor<1x64xf32> + %9979 = "ttir.relu"(%9977, %9978) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9980 = tensor.empty() : tensor<1x64xf32> + %9981 = "ttir.relu"(%9979, %9980) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9982 = tensor.empty() : tensor<1x64xf32> + %9983 = "ttir.relu"(%9981, %9982) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9984 = tensor.empty() : tensor<1x64xf32> + %9985 = "ttir.relu"(%9983, %9984) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9986 = tensor.empty() : tensor<1x64xf32> + %9987 = "ttir.relu"(%9985, %9986) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9988 = tensor.empty() : tensor<1x64xf32> + %9989 = "ttir.relu"(%9987, %9988) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9990 = tensor.empty() : tensor<1x64xf32> + %9991 = "ttir.relu"(%9989, %9990) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9992 = tensor.empty() : tensor<1x64xf32> + %9993 = "ttir.relu"(%9991, %9992) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9994 = tensor.empty() : tensor<1x64xf32> + %9995 = "ttir.relu"(%9993, %9994) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9996 = tensor.empty() : tensor<1x64xf32> + %9997 = "ttir.relu"(%9995, %9996) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %9998 = tensor.empty() : tensor<1x64xf32> + %9999 = "ttir.relu"(%9997, %9998) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + return %9999 : tensor<1x64xf32> + } +} diff --git a/tools/explorer/test/models/test_1k_ops.mlir b/tools/explorer/test/models/test_1k_ops.mlir new file mode 100644 index 000000000..fc29c77c0 --- /dev/null +++ b/tools/explorer/test/models/test_1k_ops.mlir @@ -0,0 +1,1005 @@ +module @Test10k attributes {} { + func.func @forward(%arg0: tensor<1x64xf32> {ttir.name = "input_1"}) -> (tensor<1x64xf32> {ttir.name = "TEST10k"}) { + %0 = tensor.empty() : tensor<1x64xf32> + %1 = "ttir.relu"(%arg0, %0) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %2 = tensor.empty() : tensor<1x64xf32> + %3 = "ttir.relu"(%1, %2) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %4 = tensor.empty() : tensor<1x64xf32> + %5 = "ttir.relu"(%3, %4) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %6 = tensor.empty() : tensor<1x64xf32> + %7 = "ttir.relu"(%5, %6) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %8 = tensor.empty() : tensor<1x64xf32> + %9 = "ttir.relu"(%7, %8) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %10 = tensor.empty() : tensor<1x64xf32> + %11 = "ttir.relu"(%9, %10) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %12 = tensor.empty() : tensor<1x64xf32> + %13 = "ttir.relu"(%11, %12) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %14 = tensor.empty() : tensor<1x64xf32> + %15 = "ttir.relu"(%13, %14) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %16 = tensor.empty() : tensor<1x64xf32> + %17 = "ttir.relu"(%15, %16) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %18 = tensor.empty() : tensor<1x64xf32> + %19 = "ttir.relu"(%17, %18) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %20 = tensor.empty() : tensor<1x64xf32> + %21 = "ttir.relu"(%19, %20) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %22 = tensor.empty() : tensor<1x64xf32> + %23 = "ttir.relu"(%21, %22) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %24 = tensor.empty() : tensor<1x64xf32> + %25 = "ttir.relu"(%23, %24) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %26 = tensor.empty() : tensor<1x64xf32> + %27 = "ttir.relu"(%25, %26) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %28 = tensor.empty() : tensor<1x64xf32> + %29 = "ttir.relu"(%27, %28) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %30 = tensor.empty() : tensor<1x64xf32> + %31 = "ttir.relu"(%29, %30) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %32 = tensor.empty() : tensor<1x64xf32> + %33 = "ttir.relu"(%31, %32) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %34 = tensor.empty() : tensor<1x64xf32> + %35 = "ttir.relu"(%33, %34) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %36 = tensor.empty() : tensor<1x64xf32> + %37 = "ttir.relu"(%35, %36) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %38 = tensor.empty() : tensor<1x64xf32> + %39 = "ttir.relu"(%37, %38) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %40 = tensor.empty() : tensor<1x64xf32> + %41 = "ttir.relu"(%39, %40) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %42 = tensor.empty() : tensor<1x64xf32> + %43 = "ttir.relu"(%41, %42) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %44 = tensor.empty() : tensor<1x64xf32> + %45 = "ttir.relu"(%43, %44) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %46 = tensor.empty() : tensor<1x64xf32> + %47 = "ttir.relu"(%45, %46) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %48 = tensor.empty() : tensor<1x64xf32> + %49 = "ttir.relu"(%47, %48) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %50 = tensor.empty() : tensor<1x64xf32> + %51 = "ttir.relu"(%49, %50) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %52 = tensor.empty() : tensor<1x64xf32> + %53 = "ttir.relu"(%51, %52) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %54 = tensor.empty() : tensor<1x64xf32> + %55 = "ttir.relu"(%53, %54) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %56 = tensor.empty() : tensor<1x64xf32> + %57 = "ttir.relu"(%55, %56) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %58 = tensor.empty() : tensor<1x64xf32> + %59 = "ttir.relu"(%57, %58) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %60 = tensor.empty() : tensor<1x64xf32> + %61 = "ttir.relu"(%59, %60) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %62 = tensor.empty() : tensor<1x64xf32> + %63 = "ttir.relu"(%61, %62) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %64 = tensor.empty() : tensor<1x64xf32> + %65 = "ttir.relu"(%63, %64) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %66 = tensor.empty() : tensor<1x64xf32> + %67 = "ttir.relu"(%65, %66) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %68 = tensor.empty() : tensor<1x64xf32> + %69 = "ttir.relu"(%67, %68) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %70 = tensor.empty() : tensor<1x64xf32> + %71 = "ttir.relu"(%69, %70) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %72 = tensor.empty() : tensor<1x64xf32> + %73 = "ttir.relu"(%71, %72) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %74 = tensor.empty() : tensor<1x64xf32> + %75 = "ttir.relu"(%73, %74) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %76 = tensor.empty() : tensor<1x64xf32> + %77 = "ttir.relu"(%75, %76) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %78 = tensor.empty() : tensor<1x64xf32> + %79 = "ttir.relu"(%77, %78) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %80 = tensor.empty() : tensor<1x64xf32> + %81 = "ttir.relu"(%79, %80) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %82 = tensor.empty() : tensor<1x64xf32> + %83 = "ttir.relu"(%81, %82) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %84 = tensor.empty() : tensor<1x64xf32> + %85 = "ttir.relu"(%83, %84) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %86 = tensor.empty() : tensor<1x64xf32> + %87 = "ttir.relu"(%85, %86) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %88 = tensor.empty() : tensor<1x64xf32> + %89 = "ttir.relu"(%87, %88) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %90 = tensor.empty() : tensor<1x64xf32> + %91 = "ttir.relu"(%89, %90) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %92 = tensor.empty() : tensor<1x64xf32> + %93 = "ttir.relu"(%91, %92) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %94 = tensor.empty() : tensor<1x64xf32> + %95 = "ttir.relu"(%93, %94) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %96 = tensor.empty() : tensor<1x64xf32> + %97 = "ttir.relu"(%95, %96) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %98 = tensor.empty() : tensor<1x64xf32> + %99 = "ttir.relu"(%97, %98) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %100 = tensor.empty() : tensor<1x64xf32> + %101 = "ttir.relu"(%99, %100) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %102 = tensor.empty() : tensor<1x64xf32> + %103 = "ttir.relu"(%101, %102) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %104 = tensor.empty() : tensor<1x64xf32> + %105 = "ttir.relu"(%103, %104) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %106 = tensor.empty() : tensor<1x64xf32> + %107 = "ttir.relu"(%105, %106) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %108 = tensor.empty() : tensor<1x64xf32> + %109 = "ttir.relu"(%107, %108) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %110 = tensor.empty() : tensor<1x64xf32> + %111 = "ttir.relu"(%109, %110) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %112 = tensor.empty() : tensor<1x64xf32> + %113 = "ttir.relu"(%111, %112) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %114 = tensor.empty() : tensor<1x64xf32> + %115 = "ttir.relu"(%113, %114) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %116 = tensor.empty() : tensor<1x64xf32> + %117 = "ttir.relu"(%115, %116) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %118 = tensor.empty() : tensor<1x64xf32> + %119 = "ttir.relu"(%117, %118) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %120 = tensor.empty() : tensor<1x64xf32> + %121 = "ttir.relu"(%119, %120) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %122 = tensor.empty() : tensor<1x64xf32> + %123 = "ttir.relu"(%121, %122) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %124 = tensor.empty() : tensor<1x64xf32> + %125 = "ttir.relu"(%123, %124) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %126 = tensor.empty() : tensor<1x64xf32> + %127 = "ttir.relu"(%125, %126) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %128 = tensor.empty() : tensor<1x64xf32> + %129 = "ttir.relu"(%127, %128) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %130 = tensor.empty() : tensor<1x64xf32> + %131 = "ttir.relu"(%129, %130) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %132 = tensor.empty() : tensor<1x64xf32> + %133 = "ttir.relu"(%131, %132) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %134 = tensor.empty() : tensor<1x64xf32> + %135 = "ttir.relu"(%133, %134) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %136 = tensor.empty() : tensor<1x64xf32> + %137 = "ttir.relu"(%135, %136) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %138 = tensor.empty() : tensor<1x64xf32> + %139 = "ttir.relu"(%137, %138) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %140 = tensor.empty() : tensor<1x64xf32> + %141 = "ttir.relu"(%139, %140) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %142 = tensor.empty() : tensor<1x64xf32> + %143 = "ttir.relu"(%141, %142) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %144 = tensor.empty() : tensor<1x64xf32> + %145 = "ttir.relu"(%143, %144) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %146 = tensor.empty() : tensor<1x64xf32> + %147 = "ttir.relu"(%145, %146) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %148 = tensor.empty() : tensor<1x64xf32> + %149 = "ttir.relu"(%147, %148) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %150 = tensor.empty() : tensor<1x64xf32> + %151 = "ttir.relu"(%149, %150) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %152 = tensor.empty() : tensor<1x64xf32> + %153 = "ttir.relu"(%151, %152) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %154 = tensor.empty() : tensor<1x64xf32> + %155 = "ttir.relu"(%153, %154) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %156 = tensor.empty() : tensor<1x64xf32> + %157 = "ttir.relu"(%155, %156) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %158 = tensor.empty() : tensor<1x64xf32> + %159 = "ttir.relu"(%157, %158) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %160 = tensor.empty() : tensor<1x64xf32> + %161 = "ttir.relu"(%159, %160) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %162 = tensor.empty() : tensor<1x64xf32> + %163 = "ttir.relu"(%161, %162) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %164 = tensor.empty() : tensor<1x64xf32> + %165 = "ttir.relu"(%163, %164) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %166 = tensor.empty() : tensor<1x64xf32> + %167 = "ttir.relu"(%165, %166) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %168 = tensor.empty() : tensor<1x64xf32> + %169 = "ttir.relu"(%167, %168) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %170 = tensor.empty() : tensor<1x64xf32> + %171 = "ttir.relu"(%169, %170) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %172 = tensor.empty() : tensor<1x64xf32> + %173 = "ttir.relu"(%171, %172) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %174 = tensor.empty() : tensor<1x64xf32> + %175 = "ttir.relu"(%173, %174) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %176 = tensor.empty() : tensor<1x64xf32> + %177 = "ttir.relu"(%175, %176) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %178 = tensor.empty() : tensor<1x64xf32> + %179 = "ttir.relu"(%177, %178) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %180 = tensor.empty() : tensor<1x64xf32> + %181 = "ttir.relu"(%179, %180) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %182 = tensor.empty() : tensor<1x64xf32> + %183 = "ttir.relu"(%181, %182) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %184 = tensor.empty() : tensor<1x64xf32> + %185 = "ttir.relu"(%183, %184) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %186 = tensor.empty() : tensor<1x64xf32> + %187 = "ttir.relu"(%185, %186) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %188 = tensor.empty() : tensor<1x64xf32> + %189 = "ttir.relu"(%187, %188) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %190 = tensor.empty() : tensor<1x64xf32> + %191 = "ttir.relu"(%189, %190) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %192 = tensor.empty() : tensor<1x64xf32> + %193 = "ttir.relu"(%191, %192) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %194 = tensor.empty() : tensor<1x64xf32> + %195 = "ttir.relu"(%193, %194) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %196 = tensor.empty() : tensor<1x64xf32> + %197 = "ttir.relu"(%195, %196) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %198 = tensor.empty() : tensor<1x64xf32> + %199 = "ttir.relu"(%197, %198) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %200 = tensor.empty() : tensor<1x64xf32> + %201 = "ttir.relu"(%199, %200) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %202 = tensor.empty() : tensor<1x64xf32> + %203 = "ttir.relu"(%201, %202) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %204 = tensor.empty() : tensor<1x64xf32> + %205 = "ttir.relu"(%203, %204) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %206 = tensor.empty() : tensor<1x64xf32> + %207 = "ttir.relu"(%205, %206) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %208 = tensor.empty() : tensor<1x64xf32> + %209 = "ttir.relu"(%207, %208) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %210 = tensor.empty() : tensor<1x64xf32> + %211 = "ttir.relu"(%209, %210) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %212 = tensor.empty() : tensor<1x64xf32> + %213 = "ttir.relu"(%211, %212) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %214 = tensor.empty() : tensor<1x64xf32> + %215 = "ttir.relu"(%213, %214) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %216 = tensor.empty() : tensor<1x64xf32> + %217 = "ttir.relu"(%215, %216) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %218 = tensor.empty() : tensor<1x64xf32> + %219 = "ttir.relu"(%217, %218) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %220 = tensor.empty() : tensor<1x64xf32> + %221 = "ttir.relu"(%219, %220) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %222 = tensor.empty() : tensor<1x64xf32> + %223 = "ttir.relu"(%221, %222) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %224 = tensor.empty() : tensor<1x64xf32> + %225 = "ttir.relu"(%223, %224) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %226 = tensor.empty() : tensor<1x64xf32> + %227 = "ttir.relu"(%225, %226) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %228 = tensor.empty() : tensor<1x64xf32> + %229 = "ttir.relu"(%227, %228) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %230 = tensor.empty() : tensor<1x64xf32> + %231 = "ttir.relu"(%229, %230) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %232 = tensor.empty() : tensor<1x64xf32> + %233 = "ttir.relu"(%231, %232) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %234 = tensor.empty() : tensor<1x64xf32> + %235 = "ttir.relu"(%233, %234) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %236 = tensor.empty() : tensor<1x64xf32> + %237 = "ttir.relu"(%235, %236) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %238 = tensor.empty() : tensor<1x64xf32> + %239 = "ttir.relu"(%237, %238) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %240 = tensor.empty() : tensor<1x64xf32> + %241 = "ttir.relu"(%239, %240) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %242 = tensor.empty() : tensor<1x64xf32> + %243 = "ttir.relu"(%241, %242) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %244 = tensor.empty() : tensor<1x64xf32> + %245 = "ttir.relu"(%243, %244) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %246 = tensor.empty() : tensor<1x64xf32> + %247 = "ttir.relu"(%245, %246) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %248 = tensor.empty() : tensor<1x64xf32> + %249 = "ttir.relu"(%247, %248) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %250 = tensor.empty() : tensor<1x64xf32> + %251 = "ttir.relu"(%249, %250) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %252 = tensor.empty() : tensor<1x64xf32> + %253 = "ttir.relu"(%251, %252) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %254 = tensor.empty() : tensor<1x64xf32> + %255 = "ttir.relu"(%253, %254) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %256 = tensor.empty() : tensor<1x64xf32> + %257 = "ttir.relu"(%255, %256) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %258 = tensor.empty() : tensor<1x64xf32> + %259 = "ttir.relu"(%257, %258) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %260 = tensor.empty() : tensor<1x64xf32> + %261 = "ttir.relu"(%259, %260) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %262 = tensor.empty() : tensor<1x64xf32> + %263 = "ttir.relu"(%261, %262) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %264 = tensor.empty() : tensor<1x64xf32> + %265 = "ttir.relu"(%263, %264) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %266 = tensor.empty() : tensor<1x64xf32> + %267 = "ttir.relu"(%265, %266) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %268 = tensor.empty() : tensor<1x64xf32> + %269 = "ttir.relu"(%267, %268) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %270 = tensor.empty() : tensor<1x64xf32> + %271 = "ttir.relu"(%269, %270) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %272 = tensor.empty() : tensor<1x64xf32> + %273 = "ttir.relu"(%271, %272) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %274 = tensor.empty() : tensor<1x64xf32> + %275 = "ttir.relu"(%273, %274) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %276 = tensor.empty() : tensor<1x64xf32> + %277 = "ttir.relu"(%275, %276) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %278 = tensor.empty() : tensor<1x64xf32> + %279 = "ttir.relu"(%277, %278) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %280 = tensor.empty() : tensor<1x64xf32> + %281 = "ttir.relu"(%279, %280) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %282 = tensor.empty() : tensor<1x64xf32> + %283 = "ttir.relu"(%281, %282) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %284 = tensor.empty() : tensor<1x64xf32> + %285 = "ttir.relu"(%283, %284) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %286 = tensor.empty() : tensor<1x64xf32> + %287 = "ttir.relu"(%285, %286) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %288 = tensor.empty() : tensor<1x64xf32> + %289 = "ttir.relu"(%287, %288) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %290 = tensor.empty() : tensor<1x64xf32> + %291 = "ttir.relu"(%289, %290) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %292 = tensor.empty() : tensor<1x64xf32> + %293 = "ttir.relu"(%291, %292) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %294 = tensor.empty() : tensor<1x64xf32> + %295 = "ttir.relu"(%293, %294) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %296 = tensor.empty() : tensor<1x64xf32> + %297 = "ttir.relu"(%295, %296) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %298 = tensor.empty() : tensor<1x64xf32> + %299 = "ttir.relu"(%297, %298) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %300 = tensor.empty() : tensor<1x64xf32> + %301 = "ttir.relu"(%299, %300) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %302 = tensor.empty() : tensor<1x64xf32> + %303 = "ttir.relu"(%301, %302) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %304 = tensor.empty() : tensor<1x64xf32> + %305 = "ttir.relu"(%303, %304) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %306 = tensor.empty() : tensor<1x64xf32> + %307 = "ttir.relu"(%305, %306) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %308 = tensor.empty() : tensor<1x64xf32> + %309 = "ttir.relu"(%307, %308) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %310 = tensor.empty() : tensor<1x64xf32> + %311 = "ttir.relu"(%309, %310) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %312 = tensor.empty() : tensor<1x64xf32> + %313 = "ttir.relu"(%311, %312) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %314 = tensor.empty() : tensor<1x64xf32> + %315 = "ttir.relu"(%313, %314) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %316 = tensor.empty() : tensor<1x64xf32> + %317 = "ttir.relu"(%315, %316) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %318 = tensor.empty() : tensor<1x64xf32> + %319 = "ttir.relu"(%317, %318) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %320 = tensor.empty() : tensor<1x64xf32> + %321 = "ttir.relu"(%319, %320) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %322 = tensor.empty() : tensor<1x64xf32> + %323 = "ttir.relu"(%321, %322) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %324 = tensor.empty() : tensor<1x64xf32> + %325 = "ttir.relu"(%323, %324) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %326 = tensor.empty() : tensor<1x64xf32> + %327 = "ttir.relu"(%325, %326) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %328 = tensor.empty() : tensor<1x64xf32> + %329 = "ttir.relu"(%327, %328) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %330 = tensor.empty() : tensor<1x64xf32> + %331 = "ttir.relu"(%329, %330) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %332 = tensor.empty() : tensor<1x64xf32> + %333 = "ttir.relu"(%331, %332) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %334 = tensor.empty() : tensor<1x64xf32> + %335 = "ttir.relu"(%333, %334) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %336 = tensor.empty() : tensor<1x64xf32> + %337 = "ttir.relu"(%335, %336) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %338 = tensor.empty() : tensor<1x64xf32> + %339 = "ttir.relu"(%337, %338) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %340 = tensor.empty() : tensor<1x64xf32> + %341 = "ttir.relu"(%339, %340) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %342 = tensor.empty() : tensor<1x64xf32> + %343 = "ttir.relu"(%341, %342) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %344 = tensor.empty() : tensor<1x64xf32> + %345 = "ttir.relu"(%343, %344) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %346 = tensor.empty() : tensor<1x64xf32> + %347 = "ttir.relu"(%345, %346) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %348 = tensor.empty() : tensor<1x64xf32> + %349 = "ttir.relu"(%347, %348) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %350 = tensor.empty() : tensor<1x64xf32> + %351 = "ttir.relu"(%349, %350) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %352 = tensor.empty() : tensor<1x64xf32> + %353 = "ttir.relu"(%351, %352) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %354 = tensor.empty() : tensor<1x64xf32> + %355 = "ttir.relu"(%353, %354) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %356 = tensor.empty() : tensor<1x64xf32> + %357 = "ttir.relu"(%355, %356) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %358 = tensor.empty() : tensor<1x64xf32> + %359 = "ttir.relu"(%357, %358) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %360 = tensor.empty() : tensor<1x64xf32> + %361 = "ttir.relu"(%359, %360) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %362 = tensor.empty() : tensor<1x64xf32> + %363 = "ttir.relu"(%361, %362) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %364 = tensor.empty() : tensor<1x64xf32> + %365 = "ttir.relu"(%363, %364) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %366 = tensor.empty() : tensor<1x64xf32> + %367 = "ttir.relu"(%365, %366) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %368 = tensor.empty() : tensor<1x64xf32> + %369 = "ttir.relu"(%367, %368) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %370 = tensor.empty() : tensor<1x64xf32> + %371 = "ttir.relu"(%369, %370) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %372 = tensor.empty() : tensor<1x64xf32> + %373 = "ttir.relu"(%371, %372) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %374 = tensor.empty() : tensor<1x64xf32> + %375 = "ttir.relu"(%373, %374) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %376 = tensor.empty() : tensor<1x64xf32> + %377 = "ttir.relu"(%375, %376) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %378 = tensor.empty() : tensor<1x64xf32> + %379 = "ttir.relu"(%377, %378) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %380 = tensor.empty() : tensor<1x64xf32> + %381 = "ttir.relu"(%379, %380) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %382 = tensor.empty() : tensor<1x64xf32> + %383 = "ttir.relu"(%381, %382) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %384 = tensor.empty() : tensor<1x64xf32> + %385 = "ttir.relu"(%383, %384) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %386 = tensor.empty() : tensor<1x64xf32> + %387 = "ttir.relu"(%385, %386) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %388 = tensor.empty() : tensor<1x64xf32> + %389 = "ttir.relu"(%387, %388) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %390 = tensor.empty() : tensor<1x64xf32> + %391 = "ttir.relu"(%389, %390) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %392 = tensor.empty() : tensor<1x64xf32> + %393 = "ttir.relu"(%391, %392) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %394 = tensor.empty() : tensor<1x64xf32> + %395 = "ttir.relu"(%393, %394) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %396 = tensor.empty() : tensor<1x64xf32> + %397 = "ttir.relu"(%395, %396) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %398 = tensor.empty() : tensor<1x64xf32> + %399 = "ttir.relu"(%397, %398) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %400 = tensor.empty() : tensor<1x64xf32> + %401 = "ttir.relu"(%399, %400) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %402 = tensor.empty() : tensor<1x64xf32> + %403 = "ttir.relu"(%401, %402) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %404 = tensor.empty() : tensor<1x64xf32> + %405 = "ttir.relu"(%403, %404) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %406 = tensor.empty() : tensor<1x64xf32> + %407 = "ttir.relu"(%405, %406) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %408 = tensor.empty() : tensor<1x64xf32> + %409 = "ttir.relu"(%407, %408) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %410 = tensor.empty() : tensor<1x64xf32> + %411 = "ttir.relu"(%409, %410) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %412 = tensor.empty() : tensor<1x64xf32> + %413 = "ttir.relu"(%411, %412) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %414 = tensor.empty() : tensor<1x64xf32> + %415 = "ttir.relu"(%413, %414) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %416 = tensor.empty() : tensor<1x64xf32> + %417 = "ttir.relu"(%415, %416) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %418 = tensor.empty() : tensor<1x64xf32> + %419 = "ttir.relu"(%417, %418) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %420 = tensor.empty() : tensor<1x64xf32> + %421 = "ttir.relu"(%419, %420) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %422 = tensor.empty() : tensor<1x64xf32> + %423 = "ttir.relu"(%421, %422) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %424 = tensor.empty() : tensor<1x64xf32> + %425 = "ttir.relu"(%423, %424) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %426 = tensor.empty() : tensor<1x64xf32> + %427 = "ttir.relu"(%425, %426) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %428 = tensor.empty() : tensor<1x64xf32> + %429 = "ttir.relu"(%427, %428) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %430 = tensor.empty() : tensor<1x64xf32> + %431 = "ttir.relu"(%429, %430) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %432 = tensor.empty() : tensor<1x64xf32> + %433 = "ttir.relu"(%431, %432) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %434 = tensor.empty() : tensor<1x64xf32> + %435 = "ttir.relu"(%433, %434) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %436 = tensor.empty() : tensor<1x64xf32> + %437 = "ttir.relu"(%435, %436) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %438 = tensor.empty() : tensor<1x64xf32> + %439 = "ttir.relu"(%437, %438) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %440 = tensor.empty() : tensor<1x64xf32> + %441 = "ttir.relu"(%439, %440) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %442 = tensor.empty() : tensor<1x64xf32> + %443 = "ttir.relu"(%441, %442) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %444 = tensor.empty() : tensor<1x64xf32> + %445 = "ttir.relu"(%443, %444) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %446 = tensor.empty() : tensor<1x64xf32> + %447 = "ttir.relu"(%445, %446) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %448 = tensor.empty() : tensor<1x64xf32> + %449 = "ttir.relu"(%447, %448) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %450 = tensor.empty() : tensor<1x64xf32> + %451 = "ttir.relu"(%449, %450) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %452 = tensor.empty() : tensor<1x64xf32> + %453 = "ttir.relu"(%451, %452) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %454 = tensor.empty() : tensor<1x64xf32> + %455 = "ttir.relu"(%453, %454) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %456 = tensor.empty() : tensor<1x64xf32> + %457 = "ttir.relu"(%455, %456) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %458 = tensor.empty() : tensor<1x64xf32> + %459 = "ttir.relu"(%457, %458) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %460 = tensor.empty() : tensor<1x64xf32> + %461 = "ttir.relu"(%459, %460) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %462 = tensor.empty() : tensor<1x64xf32> + %463 = "ttir.relu"(%461, %462) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %464 = tensor.empty() : tensor<1x64xf32> + %465 = "ttir.relu"(%463, %464) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %466 = tensor.empty() : tensor<1x64xf32> + %467 = "ttir.relu"(%465, %466) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %468 = tensor.empty() : tensor<1x64xf32> + %469 = "ttir.relu"(%467, %468) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %470 = tensor.empty() : tensor<1x64xf32> + %471 = "ttir.relu"(%469, %470) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %472 = tensor.empty() : tensor<1x64xf32> + %473 = "ttir.relu"(%471, %472) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %474 = tensor.empty() : tensor<1x64xf32> + %475 = "ttir.relu"(%473, %474) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %476 = tensor.empty() : tensor<1x64xf32> + %477 = "ttir.relu"(%475, %476) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %478 = tensor.empty() : tensor<1x64xf32> + %479 = "ttir.relu"(%477, %478) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %480 = tensor.empty() : tensor<1x64xf32> + %481 = "ttir.relu"(%479, %480) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %482 = tensor.empty() : tensor<1x64xf32> + %483 = "ttir.relu"(%481, %482) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %484 = tensor.empty() : tensor<1x64xf32> + %485 = "ttir.relu"(%483, %484) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %486 = tensor.empty() : tensor<1x64xf32> + %487 = "ttir.relu"(%485, %486) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %488 = tensor.empty() : tensor<1x64xf32> + %489 = "ttir.relu"(%487, %488) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %490 = tensor.empty() : tensor<1x64xf32> + %491 = "ttir.relu"(%489, %490) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %492 = tensor.empty() : tensor<1x64xf32> + %493 = "ttir.relu"(%491, %492) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %494 = tensor.empty() : tensor<1x64xf32> + %495 = "ttir.relu"(%493, %494) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %496 = tensor.empty() : tensor<1x64xf32> + %497 = "ttir.relu"(%495, %496) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %498 = tensor.empty() : tensor<1x64xf32> + %499 = "ttir.relu"(%497, %498) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %500 = tensor.empty() : tensor<1x64xf32> + %501 = "ttir.relu"(%499, %500) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %502 = tensor.empty() : tensor<1x64xf32> + %503 = "ttir.relu"(%501, %502) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %504 = tensor.empty() : tensor<1x64xf32> + %505 = "ttir.relu"(%503, %504) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %506 = tensor.empty() : tensor<1x64xf32> + %507 = "ttir.relu"(%505, %506) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %508 = tensor.empty() : tensor<1x64xf32> + %509 = "ttir.relu"(%507, %508) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %510 = tensor.empty() : tensor<1x64xf32> + %511 = "ttir.relu"(%509, %510) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %512 = tensor.empty() : tensor<1x64xf32> + %513 = "ttir.relu"(%511, %512) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %514 = tensor.empty() : tensor<1x64xf32> + %515 = "ttir.relu"(%513, %514) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %516 = tensor.empty() : tensor<1x64xf32> + %517 = "ttir.relu"(%515, %516) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %518 = tensor.empty() : tensor<1x64xf32> + %519 = "ttir.relu"(%517, %518) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %520 = tensor.empty() : tensor<1x64xf32> + %521 = "ttir.relu"(%519, %520) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %522 = tensor.empty() : tensor<1x64xf32> + %523 = "ttir.relu"(%521, %522) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %524 = tensor.empty() : tensor<1x64xf32> + %525 = "ttir.relu"(%523, %524) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %526 = tensor.empty() : tensor<1x64xf32> + %527 = "ttir.relu"(%525, %526) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %528 = tensor.empty() : tensor<1x64xf32> + %529 = "ttir.relu"(%527, %528) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %530 = tensor.empty() : tensor<1x64xf32> + %531 = "ttir.relu"(%529, %530) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %532 = tensor.empty() : tensor<1x64xf32> + %533 = "ttir.relu"(%531, %532) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %534 = tensor.empty() : tensor<1x64xf32> + %535 = "ttir.relu"(%533, %534) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %536 = tensor.empty() : tensor<1x64xf32> + %537 = "ttir.relu"(%535, %536) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %538 = tensor.empty() : tensor<1x64xf32> + %539 = "ttir.relu"(%537, %538) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %540 = tensor.empty() : tensor<1x64xf32> + %541 = "ttir.relu"(%539, %540) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %542 = tensor.empty() : tensor<1x64xf32> + %543 = "ttir.relu"(%541, %542) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %544 = tensor.empty() : tensor<1x64xf32> + %545 = "ttir.relu"(%543, %544) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %546 = tensor.empty() : tensor<1x64xf32> + %547 = "ttir.relu"(%545, %546) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %548 = tensor.empty() : tensor<1x64xf32> + %549 = "ttir.relu"(%547, %548) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %550 = tensor.empty() : tensor<1x64xf32> + %551 = "ttir.relu"(%549, %550) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %552 = tensor.empty() : tensor<1x64xf32> + %553 = "ttir.relu"(%551, %552) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %554 = tensor.empty() : tensor<1x64xf32> + %555 = "ttir.relu"(%553, %554) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %556 = tensor.empty() : tensor<1x64xf32> + %557 = "ttir.relu"(%555, %556) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %558 = tensor.empty() : tensor<1x64xf32> + %559 = "ttir.relu"(%557, %558) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %560 = tensor.empty() : tensor<1x64xf32> + %561 = "ttir.relu"(%559, %560) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %562 = tensor.empty() : tensor<1x64xf32> + %563 = "ttir.relu"(%561, %562) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %564 = tensor.empty() : tensor<1x64xf32> + %565 = "ttir.relu"(%563, %564) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %566 = tensor.empty() : tensor<1x64xf32> + %567 = "ttir.relu"(%565, %566) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %568 = tensor.empty() : tensor<1x64xf32> + %569 = "ttir.relu"(%567, %568) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %570 = tensor.empty() : tensor<1x64xf32> + %571 = "ttir.relu"(%569, %570) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %572 = tensor.empty() : tensor<1x64xf32> + %573 = "ttir.relu"(%571, %572) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %574 = tensor.empty() : tensor<1x64xf32> + %575 = "ttir.relu"(%573, %574) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %576 = tensor.empty() : tensor<1x64xf32> + %577 = "ttir.relu"(%575, %576) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %578 = tensor.empty() : tensor<1x64xf32> + %579 = "ttir.relu"(%577, %578) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %580 = tensor.empty() : tensor<1x64xf32> + %581 = "ttir.relu"(%579, %580) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %582 = tensor.empty() : tensor<1x64xf32> + %583 = "ttir.relu"(%581, %582) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %584 = tensor.empty() : tensor<1x64xf32> + %585 = "ttir.relu"(%583, %584) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %586 = tensor.empty() : tensor<1x64xf32> + %587 = "ttir.relu"(%585, %586) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %588 = tensor.empty() : tensor<1x64xf32> + %589 = "ttir.relu"(%587, %588) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %590 = tensor.empty() : tensor<1x64xf32> + %591 = "ttir.relu"(%589, %590) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %592 = tensor.empty() : tensor<1x64xf32> + %593 = "ttir.relu"(%591, %592) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %594 = tensor.empty() : tensor<1x64xf32> + %595 = "ttir.relu"(%593, %594) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %596 = tensor.empty() : tensor<1x64xf32> + %597 = "ttir.relu"(%595, %596) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %598 = tensor.empty() : tensor<1x64xf32> + %599 = "ttir.relu"(%597, %598) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %600 = tensor.empty() : tensor<1x64xf32> + %601 = "ttir.relu"(%599, %600) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %602 = tensor.empty() : tensor<1x64xf32> + %603 = "ttir.relu"(%601, %602) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %604 = tensor.empty() : tensor<1x64xf32> + %605 = "ttir.relu"(%603, %604) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %606 = tensor.empty() : tensor<1x64xf32> + %607 = "ttir.relu"(%605, %606) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %608 = tensor.empty() : tensor<1x64xf32> + %609 = "ttir.relu"(%607, %608) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %610 = tensor.empty() : tensor<1x64xf32> + %611 = "ttir.relu"(%609, %610) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %612 = tensor.empty() : tensor<1x64xf32> + %613 = "ttir.relu"(%611, %612) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %614 = tensor.empty() : tensor<1x64xf32> + %615 = "ttir.relu"(%613, %614) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %616 = tensor.empty() : tensor<1x64xf32> + %617 = "ttir.relu"(%615, %616) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %618 = tensor.empty() : tensor<1x64xf32> + %619 = "ttir.relu"(%617, %618) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %620 = tensor.empty() : tensor<1x64xf32> + %621 = "ttir.relu"(%619, %620) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %622 = tensor.empty() : tensor<1x64xf32> + %623 = "ttir.relu"(%621, %622) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %624 = tensor.empty() : tensor<1x64xf32> + %625 = "ttir.relu"(%623, %624) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %626 = tensor.empty() : tensor<1x64xf32> + %627 = "ttir.relu"(%625, %626) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %628 = tensor.empty() : tensor<1x64xf32> + %629 = "ttir.relu"(%627, %628) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %630 = tensor.empty() : tensor<1x64xf32> + %631 = "ttir.relu"(%629, %630) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %632 = tensor.empty() : tensor<1x64xf32> + %633 = "ttir.relu"(%631, %632) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %634 = tensor.empty() : tensor<1x64xf32> + %635 = "ttir.relu"(%633, %634) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %636 = tensor.empty() : tensor<1x64xf32> + %637 = "ttir.relu"(%635, %636) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %638 = tensor.empty() : tensor<1x64xf32> + %639 = "ttir.relu"(%637, %638) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %640 = tensor.empty() : tensor<1x64xf32> + %641 = "ttir.relu"(%639, %640) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %642 = tensor.empty() : tensor<1x64xf32> + %643 = "ttir.relu"(%641, %642) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %644 = tensor.empty() : tensor<1x64xf32> + %645 = "ttir.relu"(%643, %644) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %646 = tensor.empty() : tensor<1x64xf32> + %647 = "ttir.relu"(%645, %646) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %648 = tensor.empty() : tensor<1x64xf32> + %649 = "ttir.relu"(%647, %648) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %650 = tensor.empty() : tensor<1x64xf32> + %651 = "ttir.relu"(%649, %650) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %652 = tensor.empty() : tensor<1x64xf32> + %653 = "ttir.relu"(%651, %652) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %654 = tensor.empty() : tensor<1x64xf32> + %655 = "ttir.relu"(%653, %654) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %656 = tensor.empty() : tensor<1x64xf32> + %657 = "ttir.relu"(%655, %656) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %658 = tensor.empty() : tensor<1x64xf32> + %659 = "ttir.relu"(%657, %658) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %660 = tensor.empty() : tensor<1x64xf32> + %661 = "ttir.relu"(%659, %660) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %662 = tensor.empty() : tensor<1x64xf32> + %663 = "ttir.relu"(%661, %662) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %664 = tensor.empty() : tensor<1x64xf32> + %665 = "ttir.relu"(%663, %664) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %666 = tensor.empty() : tensor<1x64xf32> + %667 = "ttir.relu"(%665, %666) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %668 = tensor.empty() : tensor<1x64xf32> + %669 = "ttir.relu"(%667, %668) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %670 = tensor.empty() : tensor<1x64xf32> + %671 = "ttir.relu"(%669, %670) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %672 = tensor.empty() : tensor<1x64xf32> + %673 = "ttir.relu"(%671, %672) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %674 = tensor.empty() : tensor<1x64xf32> + %675 = "ttir.relu"(%673, %674) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %676 = tensor.empty() : tensor<1x64xf32> + %677 = "ttir.relu"(%675, %676) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %678 = tensor.empty() : tensor<1x64xf32> + %679 = "ttir.relu"(%677, %678) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %680 = tensor.empty() : tensor<1x64xf32> + %681 = "ttir.relu"(%679, %680) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %682 = tensor.empty() : tensor<1x64xf32> + %683 = "ttir.relu"(%681, %682) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %684 = tensor.empty() : tensor<1x64xf32> + %685 = "ttir.relu"(%683, %684) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %686 = tensor.empty() : tensor<1x64xf32> + %687 = "ttir.relu"(%685, %686) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %688 = tensor.empty() : tensor<1x64xf32> + %689 = "ttir.relu"(%687, %688) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %690 = tensor.empty() : tensor<1x64xf32> + %691 = "ttir.relu"(%689, %690) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %692 = tensor.empty() : tensor<1x64xf32> + %693 = "ttir.relu"(%691, %692) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %694 = tensor.empty() : tensor<1x64xf32> + %695 = "ttir.relu"(%693, %694) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %696 = tensor.empty() : tensor<1x64xf32> + %697 = "ttir.relu"(%695, %696) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %698 = tensor.empty() : tensor<1x64xf32> + %699 = "ttir.relu"(%697, %698) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %700 = tensor.empty() : tensor<1x64xf32> + %701 = "ttir.relu"(%699, %700) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %702 = tensor.empty() : tensor<1x64xf32> + %703 = "ttir.relu"(%701, %702) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %704 = tensor.empty() : tensor<1x64xf32> + %705 = "ttir.relu"(%703, %704) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %706 = tensor.empty() : tensor<1x64xf32> + %707 = "ttir.relu"(%705, %706) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %708 = tensor.empty() : tensor<1x64xf32> + %709 = "ttir.relu"(%707, %708) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %710 = tensor.empty() : tensor<1x64xf32> + %711 = "ttir.relu"(%709, %710) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %712 = tensor.empty() : tensor<1x64xf32> + %713 = "ttir.relu"(%711, %712) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %714 = tensor.empty() : tensor<1x64xf32> + %715 = "ttir.relu"(%713, %714) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %716 = tensor.empty() : tensor<1x64xf32> + %717 = "ttir.relu"(%715, %716) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %718 = tensor.empty() : tensor<1x64xf32> + %719 = "ttir.relu"(%717, %718) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %720 = tensor.empty() : tensor<1x64xf32> + %721 = "ttir.relu"(%719, %720) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %722 = tensor.empty() : tensor<1x64xf32> + %723 = "ttir.relu"(%721, %722) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %724 = tensor.empty() : tensor<1x64xf32> + %725 = "ttir.relu"(%723, %724) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %726 = tensor.empty() : tensor<1x64xf32> + %727 = "ttir.relu"(%725, %726) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %728 = tensor.empty() : tensor<1x64xf32> + %729 = "ttir.relu"(%727, %728) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %730 = tensor.empty() : tensor<1x64xf32> + %731 = "ttir.relu"(%729, %730) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %732 = tensor.empty() : tensor<1x64xf32> + %733 = "ttir.relu"(%731, %732) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %734 = tensor.empty() : tensor<1x64xf32> + %735 = "ttir.relu"(%733, %734) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %736 = tensor.empty() : tensor<1x64xf32> + %737 = "ttir.relu"(%735, %736) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %738 = tensor.empty() : tensor<1x64xf32> + %739 = "ttir.relu"(%737, %738) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %740 = tensor.empty() : tensor<1x64xf32> + %741 = "ttir.relu"(%739, %740) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %742 = tensor.empty() : tensor<1x64xf32> + %743 = "ttir.relu"(%741, %742) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %744 = tensor.empty() : tensor<1x64xf32> + %745 = "ttir.relu"(%743, %744) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %746 = tensor.empty() : tensor<1x64xf32> + %747 = "ttir.relu"(%745, %746) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %748 = tensor.empty() : tensor<1x64xf32> + %749 = "ttir.relu"(%747, %748) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %750 = tensor.empty() : tensor<1x64xf32> + %751 = "ttir.relu"(%749, %750) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %752 = tensor.empty() : tensor<1x64xf32> + %753 = "ttir.relu"(%751, %752) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %754 = tensor.empty() : tensor<1x64xf32> + %755 = "ttir.relu"(%753, %754) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %756 = tensor.empty() : tensor<1x64xf32> + %757 = "ttir.relu"(%755, %756) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %758 = tensor.empty() : tensor<1x64xf32> + %759 = "ttir.relu"(%757, %758) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %760 = tensor.empty() : tensor<1x64xf32> + %761 = "ttir.relu"(%759, %760) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %762 = tensor.empty() : tensor<1x64xf32> + %763 = "ttir.relu"(%761, %762) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %764 = tensor.empty() : tensor<1x64xf32> + %765 = "ttir.relu"(%763, %764) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %766 = tensor.empty() : tensor<1x64xf32> + %767 = "ttir.relu"(%765, %766) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %768 = tensor.empty() : tensor<1x64xf32> + %769 = "ttir.relu"(%767, %768) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %770 = tensor.empty() : tensor<1x64xf32> + %771 = "ttir.relu"(%769, %770) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %772 = tensor.empty() : tensor<1x64xf32> + %773 = "ttir.relu"(%771, %772) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %774 = tensor.empty() : tensor<1x64xf32> + %775 = "ttir.relu"(%773, %774) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %776 = tensor.empty() : tensor<1x64xf32> + %777 = "ttir.relu"(%775, %776) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %778 = tensor.empty() : tensor<1x64xf32> + %779 = "ttir.relu"(%777, %778) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %780 = tensor.empty() : tensor<1x64xf32> + %781 = "ttir.relu"(%779, %780) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %782 = tensor.empty() : tensor<1x64xf32> + %783 = "ttir.relu"(%781, %782) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %784 = tensor.empty() : tensor<1x64xf32> + %785 = "ttir.relu"(%783, %784) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %786 = tensor.empty() : tensor<1x64xf32> + %787 = "ttir.relu"(%785, %786) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %788 = tensor.empty() : tensor<1x64xf32> + %789 = "ttir.relu"(%787, %788) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %790 = tensor.empty() : tensor<1x64xf32> + %791 = "ttir.relu"(%789, %790) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %792 = tensor.empty() : tensor<1x64xf32> + %793 = "ttir.relu"(%791, %792) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %794 = tensor.empty() : tensor<1x64xf32> + %795 = "ttir.relu"(%793, %794) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %796 = tensor.empty() : tensor<1x64xf32> + %797 = "ttir.relu"(%795, %796) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %798 = tensor.empty() : tensor<1x64xf32> + %799 = "ttir.relu"(%797, %798) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %800 = tensor.empty() : tensor<1x64xf32> + %801 = "ttir.relu"(%799, %800) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %802 = tensor.empty() : tensor<1x64xf32> + %803 = "ttir.relu"(%801, %802) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %804 = tensor.empty() : tensor<1x64xf32> + %805 = "ttir.relu"(%803, %804) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %806 = tensor.empty() : tensor<1x64xf32> + %807 = "ttir.relu"(%805, %806) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %808 = tensor.empty() : tensor<1x64xf32> + %809 = "ttir.relu"(%807, %808) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %810 = tensor.empty() : tensor<1x64xf32> + %811 = "ttir.relu"(%809, %810) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %812 = tensor.empty() : tensor<1x64xf32> + %813 = "ttir.relu"(%811, %812) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %814 = tensor.empty() : tensor<1x64xf32> + %815 = "ttir.relu"(%813, %814) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %816 = tensor.empty() : tensor<1x64xf32> + %817 = "ttir.relu"(%815, %816) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %818 = tensor.empty() : tensor<1x64xf32> + %819 = "ttir.relu"(%817, %818) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %820 = tensor.empty() : tensor<1x64xf32> + %821 = "ttir.relu"(%819, %820) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %822 = tensor.empty() : tensor<1x64xf32> + %823 = "ttir.relu"(%821, %822) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %824 = tensor.empty() : tensor<1x64xf32> + %825 = "ttir.relu"(%823, %824) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %826 = tensor.empty() : tensor<1x64xf32> + %827 = "ttir.relu"(%825, %826) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %828 = tensor.empty() : tensor<1x64xf32> + %829 = "ttir.relu"(%827, %828) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %830 = tensor.empty() : tensor<1x64xf32> + %831 = "ttir.relu"(%829, %830) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %832 = tensor.empty() : tensor<1x64xf32> + %833 = "ttir.relu"(%831, %832) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %834 = tensor.empty() : tensor<1x64xf32> + %835 = "ttir.relu"(%833, %834) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %836 = tensor.empty() : tensor<1x64xf32> + %837 = "ttir.relu"(%835, %836) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %838 = tensor.empty() : tensor<1x64xf32> + %839 = "ttir.relu"(%837, %838) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %840 = tensor.empty() : tensor<1x64xf32> + %841 = "ttir.relu"(%839, %840) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %842 = tensor.empty() : tensor<1x64xf32> + %843 = "ttir.relu"(%841, %842) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %844 = tensor.empty() : tensor<1x64xf32> + %845 = "ttir.relu"(%843, %844) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %846 = tensor.empty() : tensor<1x64xf32> + %847 = "ttir.relu"(%845, %846) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %848 = tensor.empty() : tensor<1x64xf32> + %849 = "ttir.relu"(%847, %848) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %850 = tensor.empty() : tensor<1x64xf32> + %851 = "ttir.relu"(%849, %850) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %852 = tensor.empty() : tensor<1x64xf32> + %853 = "ttir.relu"(%851, %852) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %854 = tensor.empty() : tensor<1x64xf32> + %855 = "ttir.relu"(%853, %854) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %856 = tensor.empty() : tensor<1x64xf32> + %857 = "ttir.relu"(%855, %856) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %858 = tensor.empty() : tensor<1x64xf32> + %859 = "ttir.relu"(%857, %858) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %860 = tensor.empty() : tensor<1x64xf32> + %861 = "ttir.relu"(%859, %860) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %862 = tensor.empty() : tensor<1x64xf32> + %863 = "ttir.relu"(%861, %862) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %864 = tensor.empty() : tensor<1x64xf32> + %865 = "ttir.relu"(%863, %864) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %866 = tensor.empty() : tensor<1x64xf32> + %867 = "ttir.relu"(%865, %866) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %868 = tensor.empty() : tensor<1x64xf32> + %869 = "ttir.relu"(%867, %868) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %870 = tensor.empty() : tensor<1x64xf32> + %871 = "ttir.relu"(%869, %870) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %872 = tensor.empty() : tensor<1x64xf32> + %873 = "ttir.relu"(%871, %872) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %874 = tensor.empty() : tensor<1x64xf32> + %875 = "ttir.relu"(%873, %874) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %876 = tensor.empty() : tensor<1x64xf32> + %877 = "ttir.relu"(%875, %876) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %878 = tensor.empty() : tensor<1x64xf32> + %879 = "ttir.relu"(%877, %878) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %880 = tensor.empty() : tensor<1x64xf32> + %881 = "ttir.relu"(%879, %880) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %882 = tensor.empty() : tensor<1x64xf32> + %883 = "ttir.relu"(%881, %882) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %884 = tensor.empty() : tensor<1x64xf32> + %885 = "ttir.relu"(%883, %884) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %886 = tensor.empty() : tensor<1x64xf32> + %887 = "ttir.relu"(%885, %886) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %888 = tensor.empty() : tensor<1x64xf32> + %889 = "ttir.relu"(%887, %888) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %890 = tensor.empty() : tensor<1x64xf32> + %891 = "ttir.relu"(%889, %890) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %892 = tensor.empty() : tensor<1x64xf32> + %893 = "ttir.relu"(%891, %892) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %894 = tensor.empty() : tensor<1x64xf32> + %895 = "ttir.relu"(%893, %894) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %896 = tensor.empty() : tensor<1x64xf32> + %897 = "ttir.relu"(%895, %896) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %898 = tensor.empty() : tensor<1x64xf32> + %899 = "ttir.relu"(%897, %898) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %900 = tensor.empty() : tensor<1x64xf32> + %901 = "ttir.relu"(%899, %900) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %902 = tensor.empty() : tensor<1x64xf32> + %903 = "ttir.relu"(%901, %902) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %904 = tensor.empty() : tensor<1x64xf32> + %905 = "ttir.relu"(%903, %904) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %906 = tensor.empty() : tensor<1x64xf32> + %907 = "ttir.relu"(%905, %906) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %908 = tensor.empty() : tensor<1x64xf32> + %909 = "ttir.relu"(%907, %908) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %910 = tensor.empty() : tensor<1x64xf32> + %911 = "ttir.relu"(%909, %910) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %912 = tensor.empty() : tensor<1x64xf32> + %913 = "ttir.relu"(%911, %912) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %914 = tensor.empty() : tensor<1x64xf32> + %915 = "ttir.relu"(%913, %914) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %916 = tensor.empty() : tensor<1x64xf32> + %917 = "ttir.relu"(%915, %916) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %918 = tensor.empty() : tensor<1x64xf32> + %919 = "ttir.relu"(%917, %918) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %920 = tensor.empty() : tensor<1x64xf32> + %921 = "ttir.relu"(%919, %920) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %922 = tensor.empty() : tensor<1x64xf32> + %923 = "ttir.relu"(%921, %922) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %924 = tensor.empty() : tensor<1x64xf32> + %925 = "ttir.relu"(%923, %924) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %926 = tensor.empty() : tensor<1x64xf32> + %927 = "ttir.relu"(%925, %926) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %928 = tensor.empty() : tensor<1x64xf32> + %929 = "ttir.relu"(%927, %928) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %930 = tensor.empty() : tensor<1x64xf32> + %931 = "ttir.relu"(%929, %930) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %932 = tensor.empty() : tensor<1x64xf32> + %933 = "ttir.relu"(%931, %932) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %934 = tensor.empty() : tensor<1x64xf32> + %935 = "ttir.relu"(%933, %934) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %936 = tensor.empty() : tensor<1x64xf32> + %937 = "ttir.relu"(%935, %936) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %938 = tensor.empty() : tensor<1x64xf32> + %939 = "ttir.relu"(%937, %938) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %940 = tensor.empty() : tensor<1x64xf32> + %941 = "ttir.relu"(%939, %940) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %942 = tensor.empty() : tensor<1x64xf32> + %943 = "ttir.relu"(%941, %942) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %944 = tensor.empty() : tensor<1x64xf32> + %945 = "ttir.relu"(%943, %944) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %946 = tensor.empty() : tensor<1x64xf32> + %947 = "ttir.relu"(%945, %946) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %948 = tensor.empty() : tensor<1x64xf32> + %949 = "ttir.relu"(%947, %948) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %950 = tensor.empty() : tensor<1x64xf32> + %951 = "ttir.relu"(%949, %950) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %952 = tensor.empty() : tensor<1x64xf32> + %953 = "ttir.relu"(%951, %952) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %954 = tensor.empty() : tensor<1x64xf32> + %955 = "ttir.relu"(%953, %954) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %956 = tensor.empty() : tensor<1x64xf32> + %957 = "ttir.relu"(%955, %956) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %958 = tensor.empty() : tensor<1x64xf32> + %959 = "ttir.relu"(%957, %958) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %960 = tensor.empty() : tensor<1x64xf32> + %961 = "ttir.relu"(%959, %960) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %962 = tensor.empty() : tensor<1x64xf32> + %963 = "ttir.relu"(%961, %962) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %964 = tensor.empty() : tensor<1x64xf32> + %965 = "ttir.relu"(%963, %964) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %966 = tensor.empty() : tensor<1x64xf32> + %967 = "ttir.relu"(%965, %966) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %968 = tensor.empty() : tensor<1x64xf32> + %969 = "ttir.relu"(%967, %968) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %970 = tensor.empty() : tensor<1x64xf32> + %971 = "ttir.relu"(%969, %970) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %972 = tensor.empty() : tensor<1x64xf32> + %973 = "ttir.relu"(%971, %972) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %974 = tensor.empty() : tensor<1x64xf32> + %975 = "ttir.relu"(%973, %974) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %976 = tensor.empty() : tensor<1x64xf32> + %977 = "ttir.relu"(%975, %976) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %978 = tensor.empty() : tensor<1x64xf32> + %979 = "ttir.relu"(%977, %978) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %980 = tensor.empty() : tensor<1x64xf32> + %981 = "ttir.relu"(%979, %980) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %982 = tensor.empty() : tensor<1x64xf32> + %983 = "ttir.relu"(%981, %982) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %984 = tensor.empty() : tensor<1x64xf32> + %985 = "ttir.relu"(%983, %984) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %986 = tensor.empty() : tensor<1x64xf32> + %987 = "ttir.relu"(%985, %986) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %988 = tensor.empty() : tensor<1x64xf32> + %989 = "ttir.relu"(%987, %988) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %990 = tensor.empty() : tensor<1x64xf32> + %991 = "ttir.relu"(%989, %990) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %992 = tensor.empty() : tensor<1x64xf32> + %993 = "ttir.relu"(%991, %992) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %994 = tensor.empty() : tensor<1x64xf32> + %995 = "ttir.relu"(%993, %994) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %996 = tensor.empty() : tensor<1x64xf32> + %997 = "ttir.relu"(%995, %996) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + %998 = tensor.empty() : tensor<1x64xf32> + %999 = "ttir.relu"(%997, %998) <{operandSegmentSizes = array, operand_constraints = [#tt.operand_constraint, #tt.operand_constraint]}> : (tensor<1x64xf32>, tensor<1x64xf32>) -> tensor<1x64xf32> + return %999 : tensor<1x64xf32> + } +} diff --git a/tools/explorer/test/run_tests.py b/tools/explorer/test/run_tests.py index ceff14ae0..de1003f00 100644 --- a/tools/explorer/test/run_tests.py +++ b/tools/explorer/test/run_tests.py @@ -13,11 +13,19 @@ PORT = 8002 COMMAND_URL = "http://" + HOST + ":" + str(PORT) + "/apipost/v1/send_command" TEST_LOAD_MODEL_PATHS = [ - "test/ttmlir/Dialect/TTNN/optimizer/mnist_sharding.mlir", - "tools/explorer/test/models/*.mlir", + "tools/explorer/test/models/forward_and_backward.mlir", + "tools/explorer/test/models/test_1k_ops.mlir", + "tools/explorer/test/models/linear_autoencoder.mlir", + "tools/explorer/test/models/resnet_ttir.mlir", + "tools/explorer/test/models/llama_attention_no_rot_emb_ttir.mlir", + "tools/explorer/test/models/open_llama_3b_single_layer.mlir", + ] +MNIST_SHARDING_TILED_PATH = ( + "test/ttmlir/Silicon/TTNN/optimizer/mnist_sharding_tiled.mlir" +) TEST_EXECUTE_MODEL_PATHS = [ - "test/ttmlir/Silicon/TTNN/optimizer/mnist_sharding_tiled.mlir", + MNIST_SHARDING_TILED_PATH, ] @@ -97,14 +105,14 @@ def test_execute_model(model_path): def test_execute_mnist_l1_interleaved(): execute_command( - "test/ttmlir/Silicon/TTNN/optimizer/mnist_sharding_tiled.mlir", + MNIST_SHARDING_TILED_PATH, {"optimizationPolicy": "L1 Interleaved"}, ) def test_execute_mnist_optimizer_disabled(): execute_command( - "test/ttmlir/Silicon/TTNN/optimizer/mnist_sharding_tiled.mlir", + MNIST_SHARDING_TILED_PATH, {"optimizationPolicy": "Optimizer Disabled"}, )