diff --git "a/coreml/text-generation/float32_model.mlpackage/Data/com.apple.CoreML/model.mlmodel" "b/coreml/text-generation/float32_model.mlpackage/Data/com.apple.CoreML/model.mlmodel" new file mode 100644--- /dev/null +++ "b/coreml/text-generation/float32_model.mlpackage/Data/com.apple.CoreML/model.mlmodel" @@ -0,0 +1,27064 @@ +� +L + input_ids2Indices of input sequence tokens in the vocabulary * +��� +w +attention_maskXMask to avoid performing attention on padding token indices (1 = not masked, 0 = masked) * +���R8 +logits&Classification scores (before softmax)*���� +gpt2-medium (text-generation)�3 +#com.github.apple.coremltools.source torch==2.0.1�- +$com.github.apple.coremltools.version7.0b1�, +co.huggingface.exporters.name gpt2-medium�0 +co.huggingface.exporters.tasktext-generation�8 +%co.huggingface.exporters.architectureGPT2LMHeadModel�- +"co.huggingface.exporters.frameworkpytorch�- +"co.huggingface.exporters.precisionfloat32����� +main�� + + input_ids + + +� +% +attention_mask + + +�CoreML5�� +CoreML5��logits� +const5 +model_transformer_wte_weight +  +ш +�*2 +name* +" + " +model_transformer_wte_weight*B +val; +  +ш +�*" +@model_path/weights/weight.bin@� +const0 +model_transformer_h_0_ln_1_bias +   +�*5 +name- +% +#"! +model_transformer_h_0_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const2 +!model_transformer_h_0_ln_1_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_0_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_0_attn_c_attn_bias +   +�*< +name4 +, +*"( +&model_transformer_h_0_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin��b� +const7 +&model_transformer_h_0_attn_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_0_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin�•b� +const0 +model_transformer_h_0_ln_2_bias +   +�*5 +name- +% +#"! +model_transformer_h_0_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin��b� +const2 +!model_transformer_h_0_ln_2_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_0_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const4 +#model_transformer_h_0_mlp_c_fc_bias +   +� *9 +name1 +) +'"% +#model_transformer_h_0_mlp_c_fc_bias*= +val6 +   +� *% +@model_path/weights/weight.bin���b� +const6 +%model_transformer_h_0_mlp_c_proj_bias +   +�*; +name3 ++ +)"' +%model_transformer_h_0_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const0 +model_transformer_h_1_ln_1_bias +   +�*5 +name- +% +#"! +model_transformer_h_1_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin�ŗb� +const2 +!model_transformer_h_1_ln_1_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_1_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin��b� +const7 +&model_transformer_h_1_attn_c_attn_bias +   +�*< +name4 +, +*"( +&model_transformer_h_1_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_1_attn_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_1_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin��b� +const0 +model_transformer_h_1_ln_2_bias +   +�*5 +name- +% +#"! +model_transformer_h_1_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const2 +!model_transformer_h_1_ln_2_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_1_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const4 +#model_transformer_h_1_mlp_c_fc_bias +   +� *9 +name1 +) +'"% +#model_transformer_h_1_mlp_c_fc_bias*= +val6 +   +� *% +@model_path/weights/weight.bin�șb� +const6 +%model_transformer_h_1_mlp_c_proj_bias +   +�*; +name3 ++ +)"' +%model_transformer_h_1_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin�Țb� +const0 +model_transformer_h_2_ln_1_bias +   +�*5 +name- +% +#"! +model_transformer_h_2_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin��b� +const2 +!model_transformer_h_2_ln_1_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_2_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_2_attn_c_attn_bias +   +�*< +name4 +, +*"( +&model_transformer_h_2_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_2_attn_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_2_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const0 +model_transformer_h_2_ln_2_bias +   +�*5 +name- +% +#"! +model_transformer_h_2_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const2 +!model_transformer_h_2_ln_2_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_2_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin�˜b� +const4 +#model_transformer_h_2_mlp_c_fc_bias +   +� *9 +name1 +) +'"% +#model_transformer_h_2_mlp_c_fc_bias*= +val6 +   +� *% +@model_path/weights/weight.bin��b� +const6 +%model_transformer_h_2_mlp_c_proj_bias +   +�*; +name3 ++ +)"' +%model_transformer_h_2_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin��b� +const0 +model_transformer_h_3_ln_1_bias +   +�*5 +name- +% +#"! +model_transformer_h_3_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const2 +!model_transformer_h_3_ln_1_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_3_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_3_attn_c_attn_bias +   +�*< +name4 +, +*"( +&model_transformer_h_3_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin�Ξb� +const7 +&model_transformer_h_3_attn_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_3_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const0 +model_transformer_h_3_ln_2_bias +   +�*5 +name- +% +#"! +model_transformer_h_3_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin�ϟb� +const2 +!model_transformer_h_3_ln_2_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_3_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin��b� +const4 +#model_transformer_h_3_mlp_c_fc_bias +   +� *9 +name1 +) +'"% +#model_transformer_h_3_mlp_c_fc_bias*= +val6 +   +� *% +@model_path/weights/weight.bin���b� +const6 +%model_transformer_h_3_mlp_c_proj_bias +   +�*; +name3 ++ +)"' +%model_transformer_h_3_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const0 +model_transformer_h_4_ln_1_bias +   +�*5 +name- +% +#"! +model_transformer_h_4_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const2 +!model_transformer_h_4_ln_1_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_4_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin�ѡb� +const7 +&model_transformer_h_4_attn_c_attn_bias +   +�*< +name4 +, +*"( +&model_transformer_h_4_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin��b� +const7 +&model_transformer_h_4_attn_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_4_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin�Ңb� +const0 +model_transformer_h_4_ln_2_bias +   +�*5 +name- +% +#"! +model_transformer_h_4_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin��b� +const2 +!model_transformer_h_4_ln_2_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_4_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const4 +#model_transformer_h_4_mlp_c_fc_bias +   +� *9 +name1 +) +'"% +#model_transformer_h_4_mlp_c_fc_bias*= +val6 +   +� *% +@model_path/weights/weight.bin���b� +const6 +%model_transformer_h_4_mlp_c_proj_bias +   +�*; +name3 ++ +)"' +%model_transformer_h_4_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const0 +model_transformer_h_5_ln_1_bias +   +�*5 +name- +% +#"! +model_transformer_h_5_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin�դb� +const2 +!model_transformer_h_5_ln_1_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_5_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_5_attn_c_attn_bias +   +�*< +name4 +, +*"( +&model_transformer_h_5_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_5_attn_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_5_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const0 +model_transformer_h_5_ln_2_bias +   +�*5 +name- +% +#"! +model_transformer_h_5_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const2 +!model_transformer_h_5_ln_2_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_5_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const4 +#model_transformer_h_5_mlp_c_fc_bias +   +� *9 +name1 +) +'"% +#model_transformer_h_5_mlp_c_fc_bias*= +val6 +   +� *% +@model_path/weights/weight.bin�ئb� +const6 +%model_transformer_h_5_mlp_c_proj_bias +   +�*; +name3 ++ +)"' +%model_transformer_h_5_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin�اb� +const0 +model_transformer_h_6_ln_1_bias +   +�*5 +name- +% +#"! +model_transformer_h_6_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const2 +!model_transformer_h_6_ln_1_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_6_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_6_attn_c_attn_bias +   +�*< +name4 +, +*"( +&model_transformer_h_6_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_6_attn_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_6_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const0 +model_transformer_h_6_ln_2_bias +   +�*5 +name- +% +#"! +model_transformer_h_6_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const2 +!model_transformer_h_6_ln_2_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_6_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin�۩b� +const4 +#model_transformer_h_6_mlp_c_fc_bias +   +� *9 +name1 +) +'"% +#model_transformer_h_6_mlp_c_fc_bias*= +val6 +   +� *% +@model_path/weights/weight.bin���b� +const6 +%model_transformer_h_6_mlp_c_proj_bias +   +�*; +name3 ++ +)"' +%model_transformer_h_6_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const0 +model_transformer_h_7_ln_1_bias +   +�*5 +name- +% +#"! +model_transformer_h_7_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const2 +!model_transformer_h_7_ln_1_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_7_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_7_attn_c_attn_bias +   +�*< +name4 +, +*"( +&model_transformer_h_7_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin�ޫb� +const7 +&model_transformer_h_7_attn_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_7_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const0 +model_transformer_h_7_ln_2_bias +   +�*5 +name- +% +#"! +model_transformer_h_7_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin�߬b� +const2 +!model_transformer_h_7_ln_2_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_7_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const4 +#model_transformer_h_7_mlp_c_fc_bias +   +� *9 +name1 +) +'"% +#model_transformer_h_7_mlp_c_fc_bias*= +val6 +   +� *% +@model_path/weights/weight.bin���b� +const6 +%model_transformer_h_7_mlp_c_proj_bias +   +�*; +name3 ++ +)"' +%model_transformer_h_7_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const0 +model_transformer_h_8_ln_1_bias +   +�*5 +name- +% +#"! +model_transformer_h_8_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const2 +!model_transformer_h_8_ln_1_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_8_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin��b� +const7 +&model_transformer_h_8_attn_c_attn_bias +   +�*< +name4 +, +*"( +&model_transformer_h_8_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_8_attn_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_8_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin��b� +const0 +model_transformer_h_8_ln_2_bias +   +�*5 +name- +% +#"! +model_transformer_h_8_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const2 +!model_transformer_h_8_ln_2_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_8_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const4 +#model_transformer_h_8_mlp_c_fc_bias +   +� *9 +name1 +) +'"% +#model_transformer_h_8_mlp_c_fc_bias*= +val6 +   +� *% +@model_path/weights/weight.bin�İb� +const6 +%model_transformer_h_8_mlp_c_proj_bias +   +�*; +name3 ++ +)"' +%model_transformer_h_8_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin�ıb� +const0 +model_transformer_h_9_ln_1_bias +   +�*5 +name- +% +#"! +model_transformer_h_9_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin��b� +const2 +!model_transformer_h_9_ln_1_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_9_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_9_attn_c_attn_bias +   +�*< +name4 +, +*"( +&model_transformer_h_9_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_9_attn_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_9_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const0 +model_transformer_h_9_ln_2_bias +   +�*5 +name- +% +#"! +model_transformer_h_9_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const2 +!model_transformer_h_9_ln_2_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_9_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin�dzb� +const4 +#model_transformer_h_9_mlp_c_fc_bias +   +� *9 +name1 +) +'"% +#model_transformer_h_9_mlp_c_fc_bias*= +val6 +   +� *% +@model_path/weights/weight.bin��b� +const6 +%model_transformer_h_9_mlp_c_proj_bias +   +�*; +name3 ++ +)"' +%model_transformer_h_9_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin��b� +const1 + model_transformer_h_10_ln_1_bias +   +�*6 +name. +& +$"" + model_transformer_h_10_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const3 +"model_transformer_h_10_ln_1_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_10_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const8 +'model_transformer_h_10_attn_c_attn_bias +   +�*= +name5 +- ++") +'model_transformer_h_10_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin�ʵb� +const8 +'model_transformer_h_10_attn_c_proj_bias +   +�*= +name5 +- ++") +'model_transformer_h_10_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const1 + model_transformer_h_10_ln_2_bias +   +�*6 +name. +& +$"" + model_transformer_h_10_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin�˶b� +const3 +"model_transformer_h_10_ln_2_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_10_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin��b� +const5 +$model_transformer_h_10_mlp_c_fc_bias +   +� *: +name2 +* +("& +$model_transformer_h_10_mlp_c_fc_bias*= +val6 +   +� *% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_10_mlp_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_10_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const1 + model_transformer_h_11_ln_1_bias +   +�*6 +name. +& +$"" + model_transformer_h_11_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const3 +"model_transformer_h_11_ln_1_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_11_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin�͸b� +const8 +'model_transformer_h_11_attn_c_attn_bias +   +�*= +name5 +- ++") +'model_transformer_h_11_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin��b� +const8 +'model_transformer_h_11_attn_c_proj_bias +   +�*= +name5 +- ++") +'model_transformer_h_11_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin�ιb� +const1 + model_transformer_h_11_ln_2_bias +   +�*6 +name. +& +$"" + model_transformer_h_11_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin��b� +const3 +"model_transformer_h_11_ln_2_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_11_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const5 +$model_transformer_h_11_mlp_c_fc_bias +   +� *: +name2 +* +("& +$model_transformer_h_11_mlp_c_fc_bias*= +val6 +   +� *% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_11_mlp_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_11_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const1 + model_transformer_h_12_ln_1_bias +   +�*6 +name. +& +$"" + model_transformer_h_12_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin�ѻb� +const3 +"model_transformer_h_12_ln_1_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_12_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin��b� +const8 +'model_transformer_h_12_attn_c_attn_bias +   +�*= +name5 +- ++") +'model_transformer_h_12_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const8 +'model_transformer_h_12_attn_c_proj_bias +   +�*= +name5 +- ++") +'model_transformer_h_12_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin��b� +const1 + model_transformer_h_12_ln_2_bias +   +�*6 +name. +& +$"" + model_transformer_h_12_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const3 +"model_transformer_h_12_ln_2_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_12_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const5 +$model_transformer_h_12_mlp_c_fc_bias +   +� *: +name2 +* +("& +$model_transformer_h_12_mlp_c_fc_bias*= +val6 +   +� *% +@model_path/weights/weight.bin�Խb� +const7 +&model_transformer_h_12_mlp_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_12_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin�Ծb� +const1 + model_transformer_h_13_ln_1_bias +   +�*6 +name. +& +$"" + model_transformer_h_13_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const3 +"model_transformer_h_13_ln_1_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_13_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const8 +'model_transformer_h_13_attn_c_attn_bias +   +�*= +name5 +- ++") +'model_transformer_h_13_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const8 +'model_transformer_h_13_attn_c_proj_bias +   +�*= +name5 +- ++") +'model_transformer_h_13_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const1 + model_transformer_h_13_ln_2_bias +   +�*6 +name. +& +$"" + model_transformer_h_13_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const3 +"model_transformer_h_13_ln_2_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_13_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const5 +$model_transformer_h_13_mlp_c_fc_bias +   +� *: +name2 +* +("& +$model_transformer_h_13_mlp_c_fc_bias*= +val6 +   +� *% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_13_mlp_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_13_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const1 + model_transformer_h_14_ln_1_bias +   +�*6 +name. +& +$"" + model_transformer_h_14_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const3 +"model_transformer_h_14_ln_1_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_14_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const8 +'model_transformer_h_14_attn_c_attn_bias +   +�*= +name5 +- ++") +'model_transformer_h_14_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const8 +'model_transformer_h_14_attn_c_proj_bias +   +�*= +name5 +- ++") +'model_transformer_h_14_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const1 + model_transformer_h_14_ln_2_bias +   +�*6 +name. +& +$"" + model_transformer_h_14_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const3 +"model_transformer_h_14_ln_2_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_14_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const5 +$model_transformer_h_14_mlp_c_fc_bias +   +� *: +name2 +* +("& +$model_transformer_h_14_mlp_c_fc_bias*= +val6 +   +� *% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_14_mlp_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_14_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const1 + model_transformer_h_15_ln_1_bias +   +�*6 +name. +& +$"" + model_transformer_h_15_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const3 +"model_transformer_h_15_ln_1_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_15_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const8 +'model_transformer_h_15_attn_c_attn_bias +   +�*= +name5 +- ++") +'model_transformer_h_15_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const8 +'model_transformer_h_15_attn_c_proj_bias +   +�*= +name5 +- ++") +'model_transformer_h_15_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const1 + model_transformer_h_15_ln_2_bias +   +�*6 +name. +& +$"" + model_transformer_h_15_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const3 +"model_transformer_h_15_ln_2_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_15_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const5 +$model_transformer_h_15_mlp_c_fc_bias +   +� *: +name2 +* +("& +$model_transformer_h_15_mlp_c_fc_bias*= +val6 +   +� *% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_15_mlp_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_15_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const1 + model_transformer_h_16_ln_1_bias +   +�*6 +name. +& +$"" + model_transformer_h_16_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const3 +"model_transformer_h_16_ln_1_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_16_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const8 +'model_transformer_h_16_attn_c_attn_bias +   +�*= +name5 +- ++") +'model_transformer_h_16_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const8 +'model_transformer_h_16_attn_c_proj_bias +   +�*= +name5 +- ++") +'model_transformer_h_16_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const1 + model_transformer_h_16_ln_2_bias +   +�*6 +name. +& +$"" + model_transformer_h_16_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const3 +"model_transformer_h_16_ln_2_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_16_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const5 +$model_transformer_h_16_mlp_c_fc_bias +   +� *: +name2 +* +("& +$model_transformer_h_16_mlp_c_fc_bias*= +val6 +   +� *% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_16_mlp_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_16_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const1 + model_transformer_h_17_ln_1_bias +   +�*6 +name. +& +$"" + model_transformer_h_17_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const3 +"model_transformer_h_17_ln_1_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_17_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const8 +'model_transformer_h_17_attn_c_attn_bias +   +�*= +name5 +- ++") +'model_transformer_h_17_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const8 +'model_transformer_h_17_attn_c_proj_bias +   +�*= +name5 +- ++") +'model_transformer_h_17_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const1 + model_transformer_h_17_ln_2_bias +   +�*6 +name. +& +$"" + model_transformer_h_17_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const3 +"model_transformer_h_17_ln_2_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_17_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const5 +$model_transformer_h_17_mlp_c_fc_bias +   +� *: +name2 +* +("& +$model_transformer_h_17_mlp_c_fc_bias*= +val6 +   +� *% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_17_mlp_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_17_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const1 + model_transformer_h_18_ln_1_bias +   +�*6 +name. +& +$"" + model_transformer_h_18_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const3 +"model_transformer_h_18_ln_1_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_18_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const8 +'model_transformer_h_18_attn_c_attn_bias +   +�*= +name5 +- ++") +'model_transformer_h_18_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const8 +'model_transformer_h_18_attn_c_proj_bias +   +�*= +name5 +- ++") +'model_transformer_h_18_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const1 + model_transformer_h_18_ln_2_bias +   +�*6 +name. +& +$"" + model_transformer_h_18_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const3 +"model_transformer_h_18_ln_2_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_18_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const5 +$model_transformer_h_18_mlp_c_fc_bias +   +� *: +name2 +* +("& +$model_transformer_h_18_mlp_c_fc_bias*= +val6 +   +� *% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_18_mlp_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_18_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const1 + model_transformer_h_19_ln_1_bias +   +�*6 +name. +& +$"" + model_transformer_h_19_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const3 +"model_transformer_h_19_ln_1_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_19_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const8 +'model_transformer_h_19_attn_c_attn_bias +   +�*= +name5 +- ++") +'model_transformer_h_19_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const8 +'model_transformer_h_19_attn_c_proj_bias +   +�*= +name5 +- ++") +'model_transformer_h_19_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const1 + model_transformer_h_19_ln_2_bias +   +�*6 +name. +& +$"" + model_transformer_h_19_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const3 +"model_transformer_h_19_ln_2_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_19_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const5 +$model_transformer_h_19_mlp_c_fc_bias +   +� *: +name2 +* +("& +$model_transformer_h_19_mlp_c_fc_bias*= +val6 +   +� *% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_19_mlp_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_19_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const1 + model_transformer_h_20_ln_1_bias +   +�*6 +name. +& +$"" + model_transformer_h_20_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const3 +"model_transformer_h_20_ln_1_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_20_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const8 +'model_transformer_h_20_attn_c_attn_bias +   +�*= +name5 +- ++") +'model_transformer_h_20_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const8 +'model_transformer_h_20_attn_c_proj_bias +   +�*= +name5 +- ++") +'model_transformer_h_20_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const1 + model_transformer_h_20_ln_2_bias +   +�*6 +name. +& +$"" + model_transformer_h_20_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const3 +"model_transformer_h_20_ln_2_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_20_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const5 +$model_transformer_h_20_mlp_c_fc_bias +   +� *: +name2 +* +("& +$model_transformer_h_20_mlp_c_fc_bias*= +val6 +   +� *% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_20_mlp_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_20_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const1 + model_transformer_h_21_ln_1_bias +   +�*6 +name. +& +$"" + model_transformer_h_21_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const3 +"model_transformer_h_21_ln_1_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_21_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const8 +'model_transformer_h_21_attn_c_attn_bias +   +�*= +name5 +- ++") +'model_transformer_h_21_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const8 +'model_transformer_h_21_attn_c_proj_bias +   +�*= +name5 +- ++") +'model_transformer_h_21_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const1 + model_transformer_h_21_ln_2_bias +   +�*6 +name. +& +$"" + model_transformer_h_21_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const3 +"model_transformer_h_21_ln_2_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_21_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const5 +$model_transformer_h_21_mlp_c_fc_bias +   +� *: +name2 +* +("& +$model_transformer_h_21_mlp_c_fc_bias*= +val6 +   +� *% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_21_mlp_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_21_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const1 + model_transformer_h_22_ln_1_bias +   +�*6 +name. +& +$"" + model_transformer_h_22_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const3 +"model_transformer_h_22_ln_1_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_22_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const8 +'model_transformer_h_22_attn_c_attn_bias +   +�*= +name5 +- ++") +'model_transformer_h_22_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const8 +'model_transformer_h_22_attn_c_proj_bias +   +�*= +name5 +- ++") +'model_transformer_h_22_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const1 + model_transformer_h_22_ln_2_bias +   +�*6 +name. +& +$"" + model_transformer_h_22_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const3 +"model_transformer_h_22_ln_2_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_22_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const5 +$model_transformer_h_22_mlp_c_fc_bias +   +� *: +name2 +* +("& +$model_transformer_h_22_mlp_c_fc_bias*= +val6 +   +� *% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_22_mlp_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_22_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const1 + model_transformer_h_23_ln_1_bias +   +�*6 +name. +& +$"" + model_transformer_h_23_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const3 +"model_transformer_h_23_ln_1_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_23_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const8 +'model_transformer_h_23_attn_c_attn_bias +   +�*= +name5 +- ++") +'model_transformer_h_23_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const8 +'model_transformer_h_23_attn_c_proj_bias +   +�*= +name5 +- ++") +'model_transformer_h_23_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const1 + model_transformer_h_23_ln_2_bias +   +�*6 +name. +& +$"" + model_transformer_h_23_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const3 +"model_transformer_h_23_ln_2_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_23_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const5 +$model_transformer_h_23_mlp_c_fc_bias +   +� *: +name2 +* +("& +$model_transformer_h_23_mlp_c_fc_bias*= +val6 +   +� *% +@model_path/weights/weight.bin���b� +const7 +&model_transformer_h_23_mlp_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_23_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const, +model_transformer_ln_f_bias +   +�*1 +name) +! +" +model_transformer_ln_f_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���b� +const. +model_transformer_ln_f_weight +   +�*3 +name+ +# +!" +model_transformer_ln_f_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���bS +const +var_6 +* +name + + +" +op_6* +val + + + +���������O +const +var_15 + * +name + + " +op_15* +val +  + + + +�?O +const +var_26 + * +name + + " +op_26* +val +  + + + +��'7j +const + var_97_axes_0 + + +*" +name + +" + op_97_axes_0* +val + + + + +� + expand_dims +x + +attention_mask +axes + + var_97_axes_0# +var_97 + + + +�* +name + + " +op_97j +const + var_98_axes_0 + + +*" +name + +" + op_98_axes_0* +val + + + + +� + expand_dims +x + + +var_97 +axes + + var_98_axes_0) +var_98 + + + + +�* +name + + " +op_98a +const +var_100_dtype_0 +*$ +name + +" +op_100_dtype_0* +val + + +" +fp32� +cast +x + + +var_98 +dtype + +var_100_dtype_0+ +cast_390 +  + + + +�* +name + + " + +cast_390s +sub +x + + +var_15 +y + + +cast_390* +var_101 +  + + + +�* +name + + +" +op_101Q +const +var_102 + * +name + + +" +op_102* +val +  + + + +���� +mul +x + +var_101 +y + +var_1023 +attention_mask_1 +  + + + +�*$ +name + +" +attention_maski +const +inputs_embeds_axis_0 +** +name" + +" +inputs_embeds_axis_0* +val + + +� +gather% +x + +model_transformer_wte_weight +indices + + input_ids +axis + +inputs_embeds_axis_0+ + inputs_embeds +  + +� +�*# +name + +" + inputs_embeds� +const- +position_embeds +  + +� +�*% +name + +" +position_embeds*J +valC +  + +� +�*% +@model_path/weights/weight.bin���b} +add +x + + inputs_embeds +y + +position_embeds% +input_3 +  + +� +�* +name + + " +input_3n +const + +x_1_axes_0 + + +* +name + +" + +x_1_axes_0*' +val  + + + + + +���������� + +layer_norm +x + +input_3 +axes + + +x_1_axes_0. +gamma% +# +!model_transformer_h_0_ln_1_weight+ +beta# +! +model_transformer_h_0_ln_1_bias +epsilon + + +var_26! +x_1 +  + +� +�* +name + +" +x_1i +const +var_127 + + +* +name + + +" +op_127*) +val" + + + + + ����������l +reshape +x + +x_1 +shape + +var_127 +var_128 +  +� +�* +name + + +" +op_128� +constK +3model_transformer_h_0_attn_c_attn_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_0_attn_c_attn_weight_transposed*D +val= +  +� +�*% +@model_path/weights/weight.bin�ƒc� +linear +x + +var_128A +weight7 +5 +3model_transformer_h_0_attn_c_attn_weight_transposed2 +bias* +( +&model_transformer_h_0_attn_c_attn_bias +x_3 +  +� +�* +name + +" +x_3b +const +var_130 + + +* +name + + +" +op_130*" +val + + + +  +��r +reshape +x + +x_3 +shape + +var_130% +var_131 +  + +� +�* +name + + +" +op_131b +const +tile_0 + + +* +name + + +" +tile_0*# +val + + + + + +���\ +const +var_132_axis_0 +*# +name + +" + op_132_axis_0* +val + + +� +split +x + +var_131 + split_sizes + + +tile_0 +axis + +var_132_axis_0' + var_132_0 +  + +� +�' + var_132_1 +  + +� +�' + var_132_2 +  + +� +�* +name + + +" +op_132b +const +var_138 + + +* +name + + +" +op_138*" +val + + + +  +�@� +reshape +x + + var_132_0 +shape + +var_138+ +tensor_3 +  + +� + +@* +name + + " + +tensor_3b +const +var_144 + + +* +name + + +" +op_144*" +val + + + +  +�@� +reshape +x + + var_132_1 +shape + +var_144+ +tensor_7 +  + +� + +@* +name + + " + +tensor_7b +const +var_150 + + +* +name + + +" +op_150*" +val + + + +  +�@� +reshape +x + + var_132_2 +shape + +var_150, + tensor_11 +  + +� + +@* +name + + " + tensor_11a +const +var_152 + + +* +name + + +" +op_152*! +val + + + + + +y +const$ +attn_weights_1_transpose_x_0 +*2 +name* +" + " +attn_weights_1_transpose_x_0* +val + + +y +const$ +attn_weights_1_transpose_y_0 +*2 +name* +" + " +attn_weights_1_transpose_y_0* +val + + +z +const# +transpose_72_perm_0 + + +*) +name! + +" +transpose_72_perm_0*! +val + + + + + +z +const# +transpose_73_perm_0 + + +*) +name! + +" +transpose_73_perm_0*! +val + + + + + +� + transpose +x + + +tensor_7 +perm + +transpose_73_perm_00 + transpose_213 +  + + +@ +�*# +name + +" + transpose_213� + transpose +x + + +tensor_3 +perm + +transpose_72_perm_00 + transpose_214 +  + + +� +@*# +name + +" + transpose_214� +matmul +x + + transpose_214 +y + + transpose_213/ + transpose_x + +attn_weights_1_transpose_x_0/ + transpose_y + +attn_weights_1_transpose_y_02 +attn_weights_1 +  + + +� +�*$ +name + +" +attn_weights_1| +const$ +_inversed_attn_weights_3_y_0 + *2 +name* +" + " +_inversed_attn_weights_3_y_0* +val +  + + + +>� +mul +x + +attn_weights_1% +y + +_inversed_attn_weights_3_y_0< +_inversed_attn_weights_3 +  + + +� +�*. +name& + +" +_inversed_attn_weights_3�� +constA +causal_mask_1_after_broadcast + + + +� +�*3 +name+ +# +!" +causal_mask_1_after_broadcast*�� +val�� + + + +� +��� +���� +��� +const@ +mask_value_1_after_broadcast +  + + +� +�*2 +name* +" + " +mask_value_1_after_broadcast*P +valI +  + + +� +�*% +@model_path/weights/weight.bin�Ãi� +select) +cond! + +causal_mask_1_after_broadcast! +a + +_inversed_attn_weights_3% +b + +mask_value_1_after_broadcast2 +attn_weights_7 +  + + +� +�*$ +name + +" +attn_weights_7� +add +x + +attn_weights_7 +y + +attention_mask_1+ +input_5 +  + + +� +�* +name + + " +input_5� +softmax +x + +input_5 +axis + +var_62 +attn_weights_9 +  + + +� +�*$ +name + +" +attn_weights_9o +const +tensor_13_transpose_x_0 +*- +name% + +" +tensor_13_transpose_x_0* +val + + +o +const +tensor_13_transpose_y_0 +*- +name% + +" +tensor_13_transpose_y_0* +val + + +� + transpose +x + + tensor_11 +perm + +var_1520 + transpose_215 +  + + +� +@*# +name + +" + transpose_215� +matmul +x + +attn_weights_9 +y + + transpose_215* + transpose_x + +tensor_13_transpose_x_0* + transpose_y + +tensor_13_transpose_y_0, + tensor_13 +  + + +� +@* +name + + " + tensor_13a +const +var_183 + + +* +name + + +" +op_183*! +val + + + + + +i +const +var_195 + + +* +name + + +" +op_195*) +val" + + + + + ����������� + transpose +x + + tensor_13 +perm + +var_1830 + transpose_212 +  + +� + +@*# +name + +" + transpose_212v +reshape +x + + transpose_212 +shape + +var_195 +var_196 +  +� +�* +name + + +" +op_196� +constK +3model_transformer_h_0_attn_c_proj_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_0_attn_c_proj_weight_transposed*D +val= +  +� +�*% +@model_path/weights/weight.bin���i� +linear +x + +var_196A +weight7 +5 +3model_transformer_h_0_attn_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_0_attn_c_proj_bias +x_7 +  +� +�* +name + +" +x_7b +const +var_198 + + +* +name + + +" +op_198*" +val + + + +  +��s +reshape +x + +x_7 +shape + +var_198% +input_9 +  + +� +�* +name + + " +input_9q +add +x + +input_9 +y + +input_3& +input_11 +  + +� +�* +name + + " + +input_11n +const + +x_9_axes_0 + + +* +name + +" + +x_9_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + +input_11 +axes + + +x_9_axes_0. +gamma% +# +!model_transformer_h_0_ln_2_weight+ +beta# +! +model_transformer_h_0_ln_2_bias +epsilon + + +var_26! +x_9 +  + +� +�* +name + +" +x_9i +const +var_213 + + +* +name + + +" +op_213*) +val" + + + + + ����������l +reshape +x + +x_9 +shape + +var_213 +var_214 +  +� +�* +name + + +" +op_214� +constH +0model_transformer_h_0_mlp_c_fc_weight_transposed +  +�  +�*F +name> +6 +4"2 +0model_transformer_h_0_mlp_c_fc_weight_transposed*D +val= +  +�  +�*% +@model_path/weights/weight.bin���k� +linear +x + +var_214> +weight4 +2 +0model_transformer_h_0_mlp_c_fc_weight_transposed/ +bias' +% +#model_transformer_h_0_mlp_c_fc_bias +x_11 +  +� +� * +name + + +" +x_11b +const +var_216 + + +* +name + + +" +op_216*" +val + + + +  +�� v +reshape +x + +x_11 +shape + +var_216& +input_13 +  + +� +� * +name + + " + +input_13h +const + x_13_mode_0 +*! +name + +" + x_13_mode_0*' +val  + +" +TANH_APPROXIMATIONr +gelu +x + + +input_13 +mode + + x_13_mode_0" +x_13 +  + +� +� * +name + + +" +x_13i +const +var_235 + + +* +name + + +" +op_235*) +val" + + + + + ���������� m +reshape +x + +x_13 +shape + +var_235 +var_236 +  +� +� * +name + + +" +op_236� +constJ +2model_transformer_h_0_mlp_c_proj_weight_transposed +  +� +� *H +name@ +8 +6"4 +2model_transformer_h_0_mlp_c_proj_weight_transposed*D +val= +  +� +� *% +@model_path/weights/weight.bin���s� +linear +x + +var_236@ +weight6 +4 +2model_transformer_h_0_mlp_c_proj_weight_transposed1 +bias) +' +%model_transformer_h_0_mlp_c_proj_bias +x_15 +  +� +�* +name + + +" +x_15b +const +var_238 + + +* +name + + +" +op_238*" +val + + + +  +��v +reshape +x + +x_15 +shape + +var_238& +input_15 +  + +� +�* +name + + " + +input_15s +add +x + + +input_11 +y + + +input_15& +input_17 +  + +� +�* +name + + " + +input_17p +const + x_17_axes_0 + + +*! +name + +" + x_17_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + +input_17 +axes + + x_17_axes_0. +gamma% +# +!model_transformer_h_1_ln_1_weight+ +beta# +! +model_transformer_h_1_ln_1_bias +epsilon + + +var_26" +x_17 +  + +� +�* +name + + +" +x_17i +const +var_258 + + +* +name + + +" +op_258*) +val" + + + + + ����������m +reshape +x + +x_17 +shape + +var_258 +var_259 +  +� +�* +name + + +" +op_259� +constK +3model_transformer_h_1_attn_c_attn_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_1_attn_c_attn_weight_transposed*D +val= +  +� +�*% +@model_path/weights/weight.bin���{� +linear +x + +var_259A +weight7 +5 +3model_transformer_h_1_attn_c_attn_weight_transposed2 +bias* +( +&model_transformer_h_1_attn_c_attn_bias +x_19 +  +� +�* +name + + +" +x_19b +const +var_261 + + +* +name + + +" +op_261*" +val + + + +  +��s +reshape +x + +x_19 +shape + +var_261% +var_262 +  + +� +�* +name + + +" +op_262b +const +tile_1 + + +* +name + + +" +tile_1*# +val + + + + + +���\ +const +var_263_axis_0 +*# +name + +" + op_263_axis_0* +val + + +� +split +x + +var_262 + split_sizes + + +tile_1 +axis + +var_263_axis_0' + var_263_0 +  + +� +�' + var_263_1 +  + +� +�' + var_263_2 +  + +� +�* +name + + +" +op_263b +const +var_269 + + +* +name + + +" +op_269*" +val + + + +  +�@� +reshape +x + + var_263_0 +shape + +var_269, + tensor_19 +  + +� + +@* +name + + " + tensor_19b +const +var_275 + + +* +name + + +" +op_275*" +val + + + +  +�@� +reshape +x + + var_263_1 +shape + +var_275, + tensor_23 +  + +� + +@* +name + + " + tensor_23b +const +var_281 + + +* +name + + +" +op_281*" +val + + + +  +�@� +reshape +x + + var_263_2 +shape + +var_281, + tensor_27 +  + +� + +@* +name + + " + tensor_27a +const +var_283 + + +* +name + + +" +op_283*! +val + + + + + +{ +const% +attn_weights_13_transpose_x_0 +*3 +name+ +# +!" +attn_weights_13_transpose_x_0* +val + + +{ +const% +attn_weights_13_transpose_y_0 +*3 +name+ +# +!" +attn_weights_13_transpose_y_0* +val + + +z +const# +transpose_74_perm_0 + + +*) +name! + +" +transpose_74_perm_0*! +val + + + + + +z +const# +transpose_75_perm_0 + + +*) +name! + +" +transpose_75_perm_0*! +val + + + + + +� + transpose +x + + tensor_23 +perm + +transpose_75_perm_00 + transpose_209 +  + + +@ +�*# +name + +" + transpose_209� + transpose +x + + tensor_19 +perm + +transpose_74_perm_00 + transpose_210 +  + + +� +@*# +name + +" + transpose_210� +matmul +x + + transpose_210 +y + + transpose_2090 + transpose_x! + +attn_weights_13_transpose_x_00 + transpose_y! + +attn_weights_13_transpose_y_03 +attn_weights_13 +  + + +� +�*% +name + +" +attn_weights_13~ +const% +_inversed_attn_weights_15_y_0 + *3 +name+ +# +!" +_inversed_attn_weights_15_y_0* +val +  + + + +>� +mul +x + +attn_weights_13& +y! + +_inversed_attn_weights_15_y_0= +_inversed_attn_weights_15 +  + + +� +�*/ +name' + +" +_inversed_attn_weights_15� +select) +cond! + +causal_mask_1_after_broadcast" +a + +_inversed_attn_weights_15% +b + +mask_value_1_after_broadcast3 +attn_weights_19 +  + + +� +�*% +name + +" +attn_weights_19� +add +x + +attn_weights_19 +y + +attention_mask_1, +input_19 +  + + +� +�* +name + + " + +input_19� +softmax +x + + +input_19 +axis + +var_63 +attn_weights_21 +  + + +� +�*% +name + +" +attn_weights_21o +const +tensor_29_transpose_x_0 +*- +name% + +" +tensor_29_transpose_x_0* +val + + +o +const +tensor_29_transpose_y_0 +*- +name% + +" +tensor_29_transpose_y_0* +val + + +� + transpose +x + + tensor_27 +perm + +var_2830 + transpose_211 +  + + +� +@*# +name + +" + transpose_211� +matmul +x + +attn_weights_21 +y + + transpose_211* + transpose_x + +tensor_29_transpose_x_0* + transpose_y + +tensor_29_transpose_y_0, + tensor_29 +  + + +� +@* +name + + " + tensor_29a +const +var_314 + + +* +name + + +" +op_314*! +val + + + + + +i +const +var_326 + + +* +name + + +" +op_326*) +val" + + + + + ����������� + transpose +x + + tensor_29 +perm + +var_3140 + transpose_208 +  + +� + +@*# +name + +" + transpose_208v +reshape +x + + transpose_208 +shape + +var_326 +var_327 +  +� +�* +name + + +" +op_327� +constK +3model_transformer_h_1_attn_c_proj_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_1_attn_c_proj_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin��Á� +linear +x + +var_327A +weight7 +5 +3model_transformer_h_1_attn_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_1_attn_c_proj_bias +x_23 +  +� +�* +name + + +" +x_23b +const +var_329 + + +* +name + + +" +op_329*" +val + + + +  +��v +reshape +x + +x_23 +shape + +var_329& +input_23 +  + +� +�* +name + + " + +input_23s +add +x + + +input_23 +y + + +input_17& +input_25 +  + +� +�* +name + + " + +input_25p +const + x_25_axes_0 + + +*! +name + +" + x_25_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + +input_25 +axes + + x_25_axes_0. +gamma% +# +!model_transformer_h_1_ln_2_weight+ +beta# +! +model_transformer_h_1_ln_2_bias +epsilon + + +var_26" +x_25 +  + +� +�* +name + + +" +x_25i +const +var_344 + + +* +name + + +" +op_344*) +val" + + + + + ����������m +reshape +x + +x_25 +shape + +var_344 +var_345 +  +� +�* +name + + +" +op_345� +constH +0model_transformer_h_1_mlp_c_fc_weight_transposed +  +�  +�*F +name> +6 +4"2 +0model_transformer_h_1_mlp_c_fc_weight_transposed*E +val> +  +�  +�*& +@model_path/weights/weight.bin��Ã� +linear +x + +var_345> +weight4 +2 +0model_transformer_h_1_mlp_c_fc_weight_transposed/ +bias' +% +#model_transformer_h_1_mlp_c_fc_bias +x_27 +  +� +� * +name + + +" +x_27b +const +var_347 + + +* +name + + +" +op_347*" +val + + + +  +�� v +reshape +x + +x_27 +shape + +var_347& +input_27 +  + +� +� * +name + + " + +input_27h +const + x_29_mode_0 +*! +name + +" + x_29_mode_0*' +val  + +" +TANH_APPROXIMATIONr +gelu +x + + +input_27 +mode + + x_29_mode_0" +x_29 +  + +� +� * +name + + +" +x_29i +const +var_366 + + +* +name + + +" +op_366*) +val" + + + + + ���������� m +reshape +x + +x_29 +shape + +var_366 +var_367 +  +� +� * +name + + +" +op_367� +constJ +2model_transformer_h_1_mlp_c_proj_weight_transposed +  +� +� *H +name@ +8 +6"4 +2model_transformer_h_1_mlp_c_proj_weight_transposed*E +val> +  +� +� *& +@model_path/weights/weight.bin��Ë� +linear +x + +var_367@ +weight6 +4 +2model_transformer_h_1_mlp_c_proj_weight_transposed1 +bias) +' +%model_transformer_h_1_mlp_c_proj_bias +x_31 +  +� +�* +name + + +" +x_31b +const +var_369 + + +* +name + + +" +op_369*" +val + + + +  +��v +reshape +x + +x_31 +shape + +var_369& +input_29 +  + +� +�* +name + + " + +input_29s +add +x + + +input_25 +y + + +input_29& +input_31 +  + +� +�* +name + + " + +input_31p +const + x_33_axes_0 + + +*! +name + +" + x_33_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + +input_31 +axes + + x_33_axes_0. +gamma% +# +!model_transformer_h_2_ln_1_weight+ +beta# +! +model_transformer_h_2_ln_1_bias +epsilon + + +var_26" +x_33 +  + +� +�* +name + + +" +x_33i +const +var_389 + + +* +name + + +" +op_389*) +val" + + + + + ����������m +reshape +x + +x_33 +shape + +var_389 +var_390 +  +� +�* +name + + +" +op_390� +constK +3model_transformer_h_2_attn_c_attn_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_2_attn_c_attn_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin��Ó� +linear +x + +var_390A +weight7 +5 +3model_transformer_h_2_attn_c_attn_weight_transposed2 +bias* +( +&model_transformer_h_2_attn_c_attn_bias +x_35 +  +� +�* +name + + +" +x_35b +const +var_392 + + +* +name + + +" +op_392*" +val + + + +  +��s +reshape +x + +x_35 +shape + +var_392% +var_393 +  + +� +�* +name + + +" +op_393b +const +tile_2 + + +* +name + + +" +tile_2*# +val + + + + + +���\ +const +var_394_axis_0 +*# +name + +" + op_394_axis_0* +val + + +� +split +x + +var_393 + split_sizes + + +tile_2 +axis + +var_394_axis_0' + var_394_0 +  + +� +�' + var_394_1 +  + +� +�' + var_394_2 +  + +� +�* +name + + +" +op_394b +const +var_400 + + +* +name + + +" +op_400*" +val + + + +  +�@� +reshape +x + + var_394_0 +shape + +var_400, + tensor_35 +  + +� + +@* +name + + " + tensor_35b +const +var_406 + + +* +name + + +" +op_406*" +val + + + +  +�@� +reshape +x + + var_394_1 +shape + +var_406, + tensor_39 +  + +� + +@* +name + + " + tensor_39b +const +var_412 + + +* +name + + +" +op_412*" +val + + + +  +�@� +reshape +x + + var_394_2 +shape + +var_412, + tensor_43 +  + +� + +@* +name + + " + tensor_43a +const +var_414 + + +* +name + + +" +op_414*! +val + + + + + +{ +const% +attn_weights_25_transpose_x_0 +*3 +name+ +# +!" +attn_weights_25_transpose_x_0* +val + + +{ +const% +attn_weights_25_transpose_y_0 +*3 +name+ +# +!" +attn_weights_25_transpose_y_0* +val + + +z +const# +transpose_76_perm_0 + + +*) +name! + +" +transpose_76_perm_0*! +val + + + + + +z +const# +transpose_77_perm_0 + + +*) +name! + +" +transpose_77_perm_0*! +val + + + + + +� + transpose +x + + tensor_39 +perm + +transpose_77_perm_00 + transpose_205 +  + + +@ +�*# +name + +" + transpose_205� + transpose +x + + tensor_35 +perm + +transpose_76_perm_00 + transpose_206 +  + + +� +@*# +name + +" + transpose_206� +matmul +x + + transpose_206 +y + + transpose_2050 + transpose_x! + +attn_weights_25_transpose_x_00 + transpose_y! + +attn_weights_25_transpose_y_03 +attn_weights_25 +  + + +� +�*% +name + +" +attn_weights_25~ +const% +_inversed_attn_weights_27_y_0 + *3 +name+ +# +!" +_inversed_attn_weights_27_y_0* +val +  + + + +>� +mul +x + +attn_weights_25& +y! + +_inversed_attn_weights_27_y_0= +_inversed_attn_weights_27 +  + + +� +�*/ +name' + +" +_inversed_attn_weights_27� +select) +cond! + +causal_mask_1_after_broadcast" +a + +_inversed_attn_weights_27% +b + +mask_value_1_after_broadcast3 +attn_weights_31 +  + + +� +�*% +name + +" +attn_weights_31� +add +x + +attn_weights_31 +y + +attention_mask_1, +input_33 +  + + +� +�* +name + + " + +input_33� +softmax +x + + +input_33 +axis + +var_63 +attn_weights_33 +  + + +� +�*% +name + +" +attn_weights_33o +const +tensor_45_transpose_x_0 +*- +name% + +" +tensor_45_transpose_x_0* +val + + +o +const +tensor_45_transpose_y_0 +*- +name% + +" +tensor_45_transpose_y_0* +val + + +� + transpose +x + + tensor_43 +perm + +var_4140 + transpose_207 +  + + +� +@*# +name + +" + transpose_207� +matmul +x + +attn_weights_33 +y + + transpose_207* + transpose_x + +tensor_45_transpose_x_0* + transpose_y + +tensor_45_transpose_y_0, + tensor_45 +  + + +� +@* +name + + " + tensor_45a +const +var_445 + + +* +name + + +" +op_445*! +val + + + + + +i +const +var_457 + + +* +name + + +" +op_457*) +val" + + + + + ����������� + transpose +x + + tensor_45 +perm + +var_4450 + transpose_204 +  + +� + +@*# +name + +" + transpose_204v +reshape +x + + transpose_204 +shape + +var_457 +var_458 +  +� +�* +name + + +" +op_458� +constK +3model_transformer_h_2_attn_c_proj_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_2_attn_c_proj_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin��Ù� +linear +x + +var_458A +weight7 +5 +3model_transformer_h_2_attn_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_2_attn_c_proj_bias +x_39 +  +� +�* +name + + +" +x_39b +const +var_460 + + +* +name + + +" +op_460*" +val + + + +  +��v +reshape +x + +x_39 +shape + +var_460& +input_37 +  + +� +�* +name + + " + +input_37s +add +x + + +input_37 +y + + +input_31& +input_39 +  + +� +�* +name + + " + +input_39p +const + x_41_axes_0 + + +*! +name + +" + x_41_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + +input_39 +axes + + x_41_axes_0. +gamma% +# +!model_transformer_h_2_ln_2_weight+ +beta# +! +model_transformer_h_2_ln_2_bias +epsilon + + +var_26" +x_41 +  + +� +�* +name + + +" +x_41i +const +var_475 + + +* +name + + +" +op_475*) +val" + + + + + ����������m +reshape +x + +x_41 +shape + +var_475 +var_476 +  +� +�* +name + + +" +op_476� +constH +0model_transformer_h_2_mlp_c_fc_weight_transposed +  +�  +�*F +name> +6 +4"2 +0model_transformer_h_2_mlp_c_fc_weight_transposed*E +val> +  +�  +�*& +@model_path/weights/weight.bin��Û� +linear +x + +var_476> +weight4 +2 +0model_transformer_h_2_mlp_c_fc_weight_transposed/ +bias' +% +#model_transformer_h_2_mlp_c_fc_bias +x_43 +  +� +� * +name + + +" +x_43b +const +var_478 + + +* +name + + +" +op_478*" +val + + + +  +�� v +reshape +x + +x_43 +shape + +var_478& +input_41 +  + +� +� * +name + + " + +input_41h +const + x_45_mode_0 +*! +name + +" + x_45_mode_0*' +val  + +" +TANH_APPROXIMATIONr +gelu +x + + +input_41 +mode + + x_45_mode_0" +x_45 +  + +� +� * +name + + +" +x_45i +const +var_497 + + +* +name + + +" +op_497*) +val" + + + + + ���������� m +reshape +x + +x_45 +shape + +var_497 +var_498 +  +� +� * +name + + +" +op_498� +constJ +2model_transformer_h_2_mlp_c_proj_weight_transposed +  +� +� *H +name@ +8 +6"4 +2model_transformer_h_2_mlp_c_proj_weight_transposed*E +val> +  +� +� *& +@model_path/weights/weight.bin��ã� +linear +x + +var_498@ +weight6 +4 +2model_transformer_h_2_mlp_c_proj_weight_transposed1 +bias) +' +%model_transformer_h_2_mlp_c_proj_bias +x_47 +  +� +�* +name + + +" +x_47b +const +var_500 + + +* +name + + +" +op_500*" +val + + + +  +��v +reshape +x + +x_47 +shape + +var_500& +input_43 +  + +� +�* +name + + " + +input_43s +add +x + + +input_39 +y + + +input_43& +input_45 +  + +� +�* +name + + " + +input_45p +const + x_49_axes_0 + + +*! +name + +" + x_49_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + +input_45 +axes + + x_49_axes_0. +gamma% +# +!model_transformer_h_3_ln_1_weight+ +beta# +! +model_transformer_h_3_ln_1_bias +epsilon + + +var_26" +x_49 +  + +� +�* +name + + +" +x_49i +const +var_520 + + +* +name + + +" +op_520*) +val" + + + + + ����������m +reshape +x + +x_49 +shape + +var_520 +var_521 +  +� +�* +name + + +" +op_521� +constK +3model_transformer_h_3_attn_c_attn_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_3_attn_c_attn_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin��ë� +linear +x + +var_521A +weight7 +5 +3model_transformer_h_3_attn_c_attn_weight_transposed2 +bias* +( +&model_transformer_h_3_attn_c_attn_bias +x_51 +  +� +�* +name + + +" +x_51b +const +var_523 + + +* +name + + +" +op_523*" +val + + + +  +��s +reshape +x + +x_51 +shape + +var_523% +var_524 +  + +� +�* +name + + +" +op_524b +const +tile_3 + + +* +name + + +" +tile_3*# +val + + + + + +���\ +const +var_525_axis_0 +*# +name + +" + op_525_axis_0* +val + + +� +split +x + +var_524 + split_sizes + + +tile_3 +axis + +var_525_axis_0' + var_525_0 +  + +� +�' + var_525_1 +  + +� +�' + var_525_2 +  + +� +�* +name + + +" +op_525b +const +var_531 + + +* +name + + +" +op_531*" +val + + + +  +�@� +reshape +x + + var_525_0 +shape + +var_531, + tensor_51 +  + +� + +@* +name + + " + tensor_51b +const +var_537 + + +* +name + + +" +op_537*" +val + + + +  +�@� +reshape +x + + var_525_1 +shape + +var_537, + tensor_55 +  + +� + +@* +name + + " + tensor_55b +const +var_543 + + +* +name + + +" +op_543*" +val + + + +  +�@� +reshape +x + + var_525_2 +shape + +var_543, + tensor_59 +  + +� + +@* +name + + " + tensor_59a +const +var_545 + + +* +name + + +" +op_545*! +val + + + + + +{ +const% +attn_weights_37_transpose_x_0 +*3 +name+ +# +!" +attn_weights_37_transpose_x_0* +val + + +{ +const% +attn_weights_37_transpose_y_0 +*3 +name+ +# +!" +attn_weights_37_transpose_y_0* +val + + +z +const# +transpose_78_perm_0 + + +*) +name! + +" +transpose_78_perm_0*! +val + + + + + +z +const# +transpose_79_perm_0 + + +*) +name! + +" +transpose_79_perm_0*! +val + + + + + +� + transpose +x + + tensor_55 +perm + +transpose_79_perm_00 + transpose_201 +  + + +@ +�*# +name + +" + transpose_201� + transpose +x + + tensor_51 +perm + +transpose_78_perm_00 + transpose_202 +  + + +� +@*# +name + +" + transpose_202� +matmul +x + + transpose_202 +y + + transpose_2010 + transpose_x! + +attn_weights_37_transpose_x_00 + transpose_y! + +attn_weights_37_transpose_y_03 +attn_weights_37 +  + + +� +�*% +name + +" +attn_weights_37~ +const% +_inversed_attn_weights_39_y_0 + *3 +name+ +# +!" +_inversed_attn_weights_39_y_0* +val +  + + + +>� +mul +x + +attn_weights_37& +y! + +_inversed_attn_weights_39_y_0= +_inversed_attn_weights_39 +  + + +� +�*/ +name' + +" +_inversed_attn_weights_39� +select) +cond! + +causal_mask_1_after_broadcast" +a + +_inversed_attn_weights_39% +b + +mask_value_1_after_broadcast3 +attn_weights_43 +  + + +� +�*% +name + +" +attn_weights_43� +add +x + +attn_weights_43 +y + +attention_mask_1, +input_47 +  + + +� +�* +name + + " + +input_47� +softmax +x + + +input_47 +axis + +var_63 +attn_weights_45 +  + + +� +�*% +name + +" +attn_weights_45o +const +tensor_61_transpose_x_0 +*- +name% + +" +tensor_61_transpose_x_0* +val + + +o +const +tensor_61_transpose_y_0 +*- +name% + +" +tensor_61_transpose_y_0* +val + + +� + transpose +x + + tensor_59 +perm + +var_5450 + transpose_203 +  + + +� +@*# +name + +" + transpose_203� +matmul +x + +attn_weights_45 +y + + transpose_203* + transpose_x + +tensor_61_transpose_x_0* + transpose_y + +tensor_61_transpose_y_0, + tensor_61 +  + + +� +@* +name + + " + tensor_61a +const +var_576 + + +* +name + + +" +op_576*! +val + + + + + +i +const +var_588 + + +* +name + + +" +op_588*) +val" + + + + + ����������� + transpose +x + + tensor_61 +perm + +var_5760 + transpose_200 +  + +� + +@*# +name + +" + transpose_200v +reshape +x + + transpose_200 +shape + +var_588 +var_589 +  +� +�* +name + + +" +op_589� +constK +3model_transformer_h_3_attn_c_proj_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_3_attn_c_proj_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin��ñ� +linear +x + +var_589A +weight7 +5 +3model_transformer_h_3_attn_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_3_attn_c_proj_bias +x_55 +  +� +�* +name + + +" +x_55b +const +var_591 + + +* +name + + +" +op_591*" +val + + + +  +��v +reshape +x + +x_55 +shape + +var_591& +input_51 +  + +� +�* +name + + " + +input_51s +add +x + + +input_51 +y + + +input_45& +input_53 +  + +� +�* +name + + " + +input_53p +const + x_57_axes_0 + + +*! +name + +" + x_57_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + +input_53 +axes + + x_57_axes_0. +gamma% +# +!model_transformer_h_3_ln_2_weight+ +beta# +! +model_transformer_h_3_ln_2_bias +epsilon + + +var_26" +x_57 +  + +� +�* +name + + +" +x_57i +const +var_606 + + +* +name + + +" +op_606*) +val" + + + + + ����������m +reshape +x + +x_57 +shape + +var_606 +var_607 +  +� +�* +name + + +" +op_607� +constH +0model_transformer_h_3_mlp_c_fc_weight_transposed +  +�  +�*F +name> +6 +4"2 +0model_transformer_h_3_mlp_c_fc_weight_transposed*E +val> +  +�  +�*& +@model_path/weights/weight.bin��ó� +linear +x + +var_607> +weight4 +2 +0model_transformer_h_3_mlp_c_fc_weight_transposed/ +bias' +% +#model_transformer_h_3_mlp_c_fc_bias +x_59 +  +� +� * +name + + +" +x_59b +const +var_609 + + +* +name + + +" +op_609*" +val + + + +  +�� v +reshape +x + +x_59 +shape + +var_609& +input_55 +  + +� +� * +name + + " + +input_55h +const + x_61_mode_0 +*! +name + +" + x_61_mode_0*' +val  + +" +TANH_APPROXIMATIONr +gelu +x + + +input_55 +mode + + x_61_mode_0" +x_61 +  + +� +� * +name + + +" +x_61i +const +var_628 + + +* +name + + +" +op_628*) +val" + + + + + ���������� m +reshape +x + +x_61 +shape + +var_628 +var_629 +  +� +� * +name + + +" +op_629� +constJ +2model_transformer_h_3_mlp_c_proj_weight_transposed +  +� +� *H +name@ +8 +6"4 +2model_transformer_h_3_mlp_c_proj_weight_transposed*E +val> +  +� +� *& +@model_path/weights/weight.bin��û� +linear +x + +var_629@ +weight6 +4 +2model_transformer_h_3_mlp_c_proj_weight_transposed1 +bias) +' +%model_transformer_h_3_mlp_c_proj_bias +x_63 +  +� +�* +name + + +" +x_63b +const +var_631 + + +* +name + + +" +op_631*" +val + + + +  +��v +reshape +x + +x_63 +shape + +var_631& +input_57 +  + +� +�* +name + + " + +input_57s +add +x + + +input_53 +y + + +input_57& +input_59 +  + +� +�* +name + + " + +input_59p +const + x_65_axes_0 + + +*! +name + +" + x_65_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + +input_59 +axes + + x_65_axes_0. +gamma% +# +!model_transformer_h_4_ln_1_weight+ +beta# +! +model_transformer_h_4_ln_1_bias +epsilon + + +var_26" +x_65 +  + +� +�* +name + + +" +x_65i +const +var_651 + + +* +name + + +" +op_651*) +val" + + + + + ����������m +reshape +x + +x_65 +shape + +var_651 +var_652 +  +� +�* +name + + +" +op_652� +constK +3model_transformer_h_4_attn_c_attn_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_4_attn_c_attn_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin����� +linear +x + +var_652A +weight7 +5 +3model_transformer_h_4_attn_c_attn_weight_transposed2 +bias* +( +&model_transformer_h_4_attn_c_attn_bias +x_67 +  +� +�* +name + + +" +x_67b +const +var_654 + + +* +name + + +" +op_654*" +val + + + +  +��s +reshape +x + +x_67 +shape + +var_654% +var_655 +  + +� +�* +name + + +" +op_655b +const +tile_4 + + +* +name + + +" +tile_4*# +val + + + + + +���\ +const +var_656_axis_0 +*# +name + +" + op_656_axis_0* +val + + +� +split +x + +var_655 + split_sizes + + +tile_4 +axis + +var_656_axis_0' + var_656_0 +  + +� +�' + var_656_1 +  + +� +�' + var_656_2 +  + +� +�* +name + + +" +op_656b +const +var_662 + + +* +name + + +" +op_662*" +val + + + +  +�@� +reshape +x + + var_656_0 +shape + +var_662, + tensor_67 +  + +� + +@* +name + + " + tensor_67b +const +var_668 + + +* +name + + +" +op_668*" +val + + + +  +�@� +reshape +x + + var_656_1 +shape + +var_668, + tensor_71 +  + +� + +@* +name + + " + tensor_71b +const +var_674 + + +* +name + + +" +op_674*" +val + + + +  +�@� +reshape +x + + var_656_2 +shape + +var_674, + tensor_75 +  + +� + +@* +name + + " + tensor_75a +const +var_676 + + +* +name + + +" +op_676*! +val + + + + + +{ +const% +attn_weights_49_transpose_x_0 +*3 +name+ +# +!" +attn_weights_49_transpose_x_0* +val + + +{ +const% +attn_weights_49_transpose_y_0 +*3 +name+ +# +!" +attn_weights_49_transpose_y_0* +val + + +z +const# +transpose_80_perm_0 + + +*) +name! + +" +transpose_80_perm_0*! +val + + + + + +z +const# +transpose_81_perm_0 + + +*) +name! + +" +transpose_81_perm_0*! +val + + + + + +� + transpose +x + + tensor_71 +perm + +transpose_81_perm_00 + transpose_197 +  + + +@ +�*# +name + +" + transpose_197� + transpose +x + + tensor_67 +perm + +transpose_80_perm_00 + transpose_198 +  + + +� +@*# +name + +" + transpose_198� +matmul +x + + transpose_198 +y + + transpose_1970 + transpose_x! + +attn_weights_49_transpose_x_00 + transpose_y! + +attn_weights_49_transpose_y_03 +attn_weights_49 +  + + +� +�*% +name + +" +attn_weights_49~ +const% +_inversed_attn_weights_51_y_0 + *3 +name+ +# +!" +_inversed_attn_weights_51_y_0* +val +  + + + +>� +mul +x + +attn_weights_49& +y! + +_inversed_attn_weights_51_y_0= +_inversed_attn_weights_51 +  + + +� +�*/ +name' + +" +_inversed_attn_weights_51� +select) +cond! + +causal_mask_1_after_broadcast" +a + +_inversed_attn_weights_51% +b + +mask_value_1_after_broadcast3 +attn_weights_55 +  + + +� +�*% +name + +" +attn_weights_55� +add +x + +attn_weights_55 +y + +attention_mask_1, +input_61 +  + + +� +�* +name + + " + +input_61� +softmax +x + + +input_61 +axis + +var_63 +attn_weights_57 +  + + +� +�*% +name + +" +attn_weights_57o +const +tensor_77_transpose_x_0 +*- +name% + +" +tensor_77_transpose_x_0* +val + + +o +const +tensor_77_transpose_y_0 +*- +name% + +" +tensor_77_transpose_y_0* +val + + +� + transpose +x + + tensor_75 +perm + +var_6760 + transpose_199 +  + + +� +@*# +name + +" + transpose_199� +matmul +x + +attn_weights_57 +y + + transpose_199* + transpose_x + +tensor_77_transpose_x_0* + transpose_y + +tensor_77_transpose_y_0, + tensor_77 +  + + +� +@* +name + + " + tensor_77a +const +var_707 + + +* +name + + +" +op_707*! +val + + + + + +i +const +var_719 + + +* +name + + +" +op_719*) +val" + + + + + ����������� + transpose +x + + tensor_77 +perm + +var_7070 + transpose_196 +  + +� + +@*# +name + +" + transpose_196v +reshape +x + + transpose_196 +shape + +var_719 +var_720 +  +� +�* +name + + +" +op_720� +constK +3model_transformer_h_4_attn_c_proj_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_4_attn_c_proj_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin����� +linear +x + +var_720A +weight7 +5 +3model_transformer_h_4_attn_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_4_attn_c_proj_bias +x_71 +  +� +�* +name + + +" +x_71b +const +var_722 + + +* +name + + +" +op_722*" +val + + + +  +��v +reshape +x + +x_71 +shape + +var_722& +input_65 +  + +� +�* +name + + " + +input_65s +add +x + + +input_65 +y + + +input_59& +input_67 +  + +� +�* +name + + " + +input_67p +const + x_73_axes_0 + + +*! +name + +" + x_73_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + +input_67 +axes + + x_73_axes_0. +gamma% +# +!model_transformer_h_4_ln_2_weight+ +beta# +! +model_transformer_h_4_ln_2_bias +epsilon + + +var_26" +x_73 +  + +� +�* +name + + +" +x_73i +const +var_737 + + +* +name + + +" +op_737*) +val" + + + + + ����������m +reshape +x + +x_73 +shape + +var_737 +var_738 +  +� +�* +name + + +" +op_738� +constH +0model_transformer_h_4_mlp_c_fc_weight_transposed +  +�  +�*F +name> +6 +4"2 +0model_transformer_h_4_mlp_c_fc_weight_transposed*E +val> +  +�  +�*& +@model_path/weights/weight.bin����� +linear +x + +var_738> +weight4 +2 +0model_transformer_h_4_mlp_c_fc_weight_transposed/ +bias' +% +#model_transformer_h_4_mlp_c_fc_bias +x_75 +  +� +� * +name + + +" +x_75b +const +var_740 + + +* +name + + +" +op_740*" +val + + + +  +�� v +reshape +x + +x_75 +shape + +var_740& +input_69 +  + +� +� * +name + + " + +input_69h +const + x_77_mode_0 +*! +name + +" + x_77_mode_0*' +val  + +" +TANH_APPROXIMATIONr +gelu +x + + +input_69 +mode + + x_77_mode_0" +x_77 +  + +� +� * +name + + +" +x_77i +const +var_759 + + +* +name + + +" +op_759*) +val" + + + + + ���������� m +reshape +x + +x_77 +shape + +var_759 +var_760 +  +� +� * +name + + +" +op_760� +constJ +2model_transformer_h_4_mlp_c_proj_weight_transposed +  +� +� *H +name@ +8 +6"4 +2model_transformer_h_4_mlp_c_proj_weight_transposed*E +val> +  +� +� *& +@model_path/weights/weight.bin����� +linear +x + +var_760@ +weight6 +4 +2model_transformer_h_4_mlp_c_proj_weight_transposed1 +bias) +' +%model_transformer_h_4_mlp_c_proj_bias +x_79 +  +� +�* +name + + +" +x_79b +const +var_762 + + +* +name + + +" +op_762*" +val + + + +  +��v +reshape +x + +x_79 +shape + +var_762& +input_71 +  + +� +�* +name + + " + +input_71s +add +x + + +input_67 +y + + +input_71& +input_73 +  + +� +�* +name + + " + +input_73p +const + x_81_axes_0 + + +*! +name + +" + x_81_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + +input_73 +axes + + x_81_axes_0. +gamma% +# +!model_transformer_h_5_ln_1_weight+ +beta# +! +model_transformer_h_5_ln_1_bias +epsilon + + +var_26" +x_81 +  + +� +�* +name + + +" +x_81i +const +var_782 + + +* +name + + +" +op_782*) +val" + + + + + ����������m +reshape +x + +x_81 +shape + +var_782 +var_783 +  +� +�* +name + + +" +op_783� +constK +3model_transformer_h_5_attn_c_attn_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_5_attn_c_attn_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin����� +linear +x + +var_783A +weight7 +5 +3model_transformer_h_5_attn_c_attn_weight_transposed2 +bias* +( +&model_transformer_h_5_attn_c_attn_bias +x_83 +  +� +�* +name + + +" +x_83b +const +var_785 + + +* +name + + +" +op_785*" +val + + + +  +��s +reshape +x + +x_83 +shape + +var_785% +var_786 +  + +� +�* +name + + +" +op_786b +const +tile_5 + + +* +name + + +" +tile_5*# +val + + + + + +���\ +const +var_787_axis_0 +*# +name + +" + op_787_axis_0* +val + + +� +split +x + +var_786 + split_sizes + + +tile_5 +axis + +var_787_axis_0' + var_787_0 +  + +� +�' + var_787_1 +  + +� +�' + var_787_2 +  + +� +�* +name + + +" +op_787b +const +var_793 + + +* +name + + +" +op_793*" +val + + + +  +�@� +reshape +x + + var_787_0 +shape + +var_793, + tensor_83 +  + +� + +@* +name + + " + tensor_83b +const +var_799 + + +* +name + + +" +op_799*" +val + + + +  +�@� +reshape +x + + var_787_1 +shape + +var_799, + tensor_87 +  + +� + +@* +name + + " + tensor_87b +const +var_805 + + +* +name + + +" +op_805*" +val + + + +  +�@� +reshape +x + + var_787_2 +shape + +var_805, + tensor_91 +  + +� + +@* +name + + " + tensor_91a +const +var_807 + + +* +name + + +" +op_807*! +val + + + + + +{ +const% +attn_weights_61_transpose_x_0 +*3 +name+ +# +!" +attn_weights_61_transpose_x_0* +val + + +{ +const% +attn_weights_61_transpose_y_0 +*3 +name+ +# +!" +attn_weights_61_transpose_y_0* +val + + +z +const# +transpose_82_perm_0 + + +*) +name! + +" +transpose_82_perm_0*! +val + + + + + +z +const# +transpose_83_perm_0 + + +*) +name! + +" +transpose_83_perm_0*! +val + + + + + +� + transpose +x + + tensor_87 +perm + +transpose_83_perm_00 + transpose_193 +  + + +@ +�*# +name + +" + transpose_193� + transpose +x + + tensor_83 +perm + +transpose_82_perm_00 + transpose_194 +  + + +� +@*# +name + +" + transpose_194� +matmul +x + + transpose_194 +y + + transpose_1930 + transpose_x! + +attn_weights_61_transpose_x_00 + transpose_y! + +attn_weights_61_transpose_y_03 +attn_weights_61 +  + + +� +�*% +name + +" +attn_weights_61~ +const% +_inversed_attn_weights_63_y_0 + *3 +name+ +# +!" +_inversed_attn_weights_63_y_0* +val +  + + + +>� +mul +x + +attn_weights_61& +y! + +_inversed_attn_weights_63_y_0= +_inversed_attn_weights_63 +  + + +� +�*/ +name' + +" +_inversed_attn_weights_63� +select) +cond! + +causal_mask_1_after_broadcast" +a + +_inversed_attn_weights_63% +b + +mask_value_1_after_broadcast3 +attn_weights_67 +  + + +� +�*% +name + +" +attn_weights_67� +add +x + +attn_weights_67 +y + +attention_mask_1, +input_75 +  + + +� +�* +name + + " + +input_75� +softmax +x + + +input_75 +axis + +var_63 +attn_weights_69 +  + + +� +�*% +name + +" +attn_weights_69o +const +tensor_93_transpose_x_0 +*- +name% + +" +tensor_93_transpose_x_0* +val + + +o +const +tensor_93_transpose_y_0 +*- +name% + +" +tensor_93_transpose_y_0* +val + + +� + transpose +x + + tensor_91 +perm + +var_8070 + transpose_195 +  + + +� +@*# +name + +" + transpose_195� +matmul +x + +attn_weights_69 +y + + transpose_195* + transpose_x + +tensor_93_transpose_x_0* + transpose_y + +tensor_93_transpose_y_0, + tensor_93 +  + + +� +@* +name + + " + tensor_93a +const +var_838 + + +* +name + + +" +op_838*! +val + + + + + +i +const +var_850 + + +* +name + + +" +op_850*) +val" + + + + + ����������� + transpose +x + + tensor_93 +perm + +var_8380 + transpose_192 +  + +� + +@*# +name + +" + transpose_192v +reshape +x + + transpose_192 +shape + +var_850 +var_851 +  +� +�* +name + + +" +op_851� +constK +3model_transformer_h_5_attn_c_proj_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_5_attn_c_proj_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin����� +linear +x + +var_851A +weight7 +5 +3model_transformer_h_5_attn_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_5_attn_c_proj_bias +x_87 +  +� +�* +name + + +" +x_87b +const +var_853 + + +* +name + + +" +op_853*" +val + + + +  +��v +reshape +x + +x_87 +shape + +var_853& +input_79 +  + +� +�* +name + + " + +input_79s +add +x + + +input_79 +y + + +input_73& +input_81 +  + +� +�* +name + + " + +input_81p +const + x_89_axes_0 + + +*! +name + +" + x_89_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + +input_81 +axes + + x_89_axes_0. +gamma% +# +!model_transformer_h_5_ln_2_weight+ +beta# +! +model_transformer_h_5_ln_2_bias +epsilon + + +var_26" +x_89 +  + +� +�* +name + + +" +x_89i +const +var_868 + + +* +name + + +" +op_868*) +val" + + + + + ����������m +reshape +x + +x_89 +shape + +var_868 +var_869 +  +� +�* +name + + +" +op_869� +constH +0model_transformer_h_5_mlp_c_fc_weight_transposed +  +�  +�*F +name> +6 +4"2 +0model_transformer_h_5_mlp_c_fc_weight_transposed*E +val> +  +�  +�*& +@model_path/weights/weight.bin����� +linear +x + +var_869> +weight4 +2 +0model_transformer_h_5_mlp_c_fc_weight_transposed/ +bias' +% +#model_transformer_h_5_mlp_c_fc_bias +x_91 +  +� +� * +name + + +" +x_91b +const +var_871 + + +* +name + + +" +op_871*" +val + + + +  +�� v +reshape +x + +x_91 +shape + +var_871& +input_83 +  + +� +� * +name + + " + +input_83h +const + x_93_mode_0 +*! +name + +" + x_93_mode_0*' +val  + +" +TANH_APPROXIMATIONr +gelu +x + + +input_83 +mode + + x_93_mode_0" +x_93 +  + +� +� * +name + + +" +x_93i +const +var_890 + + +* +name + + +" +op_890*) +val" + + + + + ���������� m +reshape +x + +x_93 +shape + +var_890 +var_891 +  +� +� * +name + + +" +op_891� +constJ +2model_transformer_h_5_mlp_c_proj_weight_transposed +  +� +� *H +name@ +8 +6"4 +2model_transformer_h_5_mlp_c_proj_weight_transposed*E +val> +  +� +� *& +@model_path/weights/weight.bin����� +linear +x + +var_891@ +weight6 +4 +2model_transformer_h_5_mlp_c_proj_weight_transposed1 +bias) +' +%model_transformer_h_5_mlp_c_proj_bias +x_95 +  +� +�* +name + + +" +x_95b +const +var_893 + + +* +name + + +" +op_893*" +val + + + +  +��v +reshape +x + +x_95 +shape + +var_893& +input_85 +  + +� +�* +name + + " + +input_85s +add +x + + +input_81 +y + + +input_85& +input_87 +  + +� +�* +name + + " + +input_87p +const + x_97_axes_0 + + +*! +name + +" + x_97_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + +input_87 +axes + + x_97_axes_0. +gamma% +# +!model_transformer_h_6_ln_1_weight+ +beta# +! +model_transformer_h_6_ln_1_bias +epsilon + + +var_26" +x_97 +  + +� +�* +name + + +" +x_97i +const +var_913 + + +* +name + + +" +op_913*) +val" + + + + + ����������m +reshape +x + +x_97 +shape + +var_913 +var_914 +  +� +�* +name + + +" +op_914� +constK +3model_transformer_h_6_attn_c_attn_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_6_attn_c_attn_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin����� +linear +x + +var_914A +weight7 +5 +3model_transformer_h_6_attn_c_attn_weight_transposed2 +bias* +( +&model_transformer_h_6_attn_c_attn_bias +x_99 +  +� +�* +name + + +" +x_99b +const +var_916 + + +* +name + + +" +op_916*" +val + + + +  +��s +reshape +x + +x_99 +shape + +var_916% +var_917 +  + +� +�* +name + + +" +op_917b +const +tile_6 + + +* +name + + +" +tile_6*# +val + + + + + +���\ +const +var_918_axis_0 +*# +name + +" + op_918_axis_0* +val + + +� +split +x + +var_917 + split_sizes + + +tile_6 +axis + +var_918_axis_0' + var_918_0 +  + +� +�' + var_918_1 +  + +� +�' + var_918_2 +  + +� +�* +name + + +" +op_918b +const +var_924 + + +* +name + + +" +op_924*" +val + + + +  +�@� +reshape +x + + var_918_0 +shape + +var_924, + tensor_99 +  + +� + +@* +name + + " + tensor_99b +const +var_930 + + +* +name + + +" +op_930*" +val + + + +  +�@� +reshape +x + + var_918_1 +shape + +var_930- + +tensor_103 +  + +� + +@* +name + +" + +tensor_103b +const +var_936 + + +* +name + + +" +op_936*" +val + + + +  +�@� +reshape +x + + var_918_2 +shape + +var_936- + +tensor_107 +  + +� + +@* +name + +" + +tensor_107a +const +var_938 + + +* +name + + +" +op_938*! +val + + + + + +{ +const% +attn_weights_73_transpose_x_0 +*3 +name+ +# +!" +attn_weights_73_transpose_x_0* +val + + +{ +const% +attn_weights_73_transpose_y_0 +*3 +name+ +# +!" +attn_weights_73_transpose_y_0* +val + + +z +const# +transpose_84_perm_0 + + +*) +name! + +" +transpose_84_perm_0*! +val + + + + + +z +const# +transpose_85_perm_0 + + +*) +name! + +" +transpose_85_perm_0*! +val + + + + + +� + transpose +x + + +tensor_103 +perm + +transpose_85_perm_00 + transpose_189 +  + + +@ +�*# +name + +" + transpose_189� + transpose +x + + tensor_99 +perm + +transpose_84_perm_00 + transpose_190 +  + + +� +@*# +name + +" + transpose_190� +matmul +x + + transpose_190 +y + + transpose_1890 + transpose_x! + +attn_weights_73_transpose_x_00 + transpose_y! + +attn_weights_73_transpose_y_03 +attn_weights_73 +  + + +� +�*% +name + +" +attn_weights_73~ +const% +_inversed_attn_weights_75_y_0 + *3 +name+ +# +!" +_inversed_attn_weights_75_y_0* +val +  + + + +>� +mul +x + +attn_weights_73& +y! + +_inversed_attn_weights_75_y_0= +_inversed_attn_weights_75 +  + + +� +�*/ +name' + +" +_inversed_attn_weights_75� +select) +cond! + +causal_mask_1_after_broadcast" +a + +_inversed_attn_weights_75% +b + +mask_value_1_after_broadcast3 +attn_weights_79 +  + + +� +�*% +name + +" +attn_weights_79� +add +x + +attn_weights_79 +y + +attention_mask_1, +input_89 +  + + +� +�* +name + + " + +input_89� +softmax +x + + +input_89 +axis + +var_63 +attn_weights_81 +  + + +� +�*% +name + +" +attn_weights_81q +const +tensor_109_transpose_x_0 +*. +name& + +" +tensor_109_transpose_x_0* +val + + +q +const +tensor_109_transpose_y_0 +*. +name& + +" +tensor_109_transpose_y_0* +val + + +� + transpose +x + + +tensor_107 +perm + +var_9380 + transpose_191 +  + + +� +@*# +name + +" + transpose_191� +matmul +x + +attn_weights_81 +y + + transpose_191+ + transpose_x + +tensor_109_transpose_x_0+ + transpose_y + +tensor_109_transpose_y_0- + +tensor_109 +  + + +� +@* +name + +" + +tensor_109a +const +var_969 + + +* +name + + +" +op_969*! +val + + + + + +i +const +var_981 + + +* +name + + +" +op_981*) +val" + + + + + ����������� + transpose +x + + +tensor_109 +perm + +var_9690 + transpose_188 +  + +� + +@*# +name + +" + transpose_188v +reshape +x + + transpose_188 +shape + +var_981 +var_982 +  +� +�* +name + + +" +op_982� +constK +3model_transformer_h_6_attn_c_proj_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_6_attn_c_proj_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin����� +linear +x + +var_982A +weight7 +5 +3model_transformer_h_6_attn_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_6_attn_c_proj_bias +x_103 +  +� +�* +name + + " +x_103b +const +var_984 + + +* +name + + +" +op_984*" +val + + + +  +��w +reshape +x + +x_103 +shape + +var_984& +input_93 +  + +� +�* +name + + " + +input_93s +add +x + + +input_93 +y + + +input_87& +input_95 +  + +� +�* +name + + " + +input_95r +const + x_105_axes_0 + + +*" +name + +" + x_105_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + +input_95 +axes + + x_105_axes_0. +gamma% +# +!model_transformer_h_6_ln_2_weight+ +beta# +! +model_transformer_h_6_ln_2_bias +epsilon + + +var_26# +x_105 +  + +� +�* +name + + " +x_105i +const +var_999 + + +* +name + + +" +op_999*) +val" + + + + + ����������p +reshape +x + +x_105 +shape + +var_999 +var_1000 +  +� +�* +name + + " +op_1000� +constH +0model_transformer_h_6_mlp_c_fc_weight_transposed +  +�  +�*F +name> +6 +4"2 +0model_transformer_h_6_mlp_c_fc_weight_transposed*E +val> +  +�  +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_1000> +weight4 +2 +0model_transformer_h_6_mlp_c_fc_weight_transposed/ +bias' +% +#model_transformer_h_6_mlp_c_fc_bias +x_107 +  +� +� * +name + + " +x_107d +const +var_1002 + + +* +name + + " +op_1002*" +val + + + +  +�� x +reshape +x + +x_107 +shape + + +var_1002& +input_97 +  + +� +� * +name + + " + +input_97j +const + x_109_mode_0 +*" +name + +" + x_109_mode_0*' +val  + +" +TANH_APPROXIMATIONu +gelu +x + + +input_97 +mode + + x_109_mode_0# +x_109 +  + +� +� * +name + + " +x_109k +const +var_1021 + + +* +name + + " +op_1021*) +val" + + + + + ���������� q +reshape +x + +x_109 +shape + + +var_1021 +var_1022 +  +� +� * +name + + " +op_1022� +constJ +2model_transformer_h_6_mlp_c_proj_weight_transposed +  +� +� *H +name@ +8 +6"4 +2model_transformer_h_6_mlp_c_proj_weight_transposed*E +val> +  +� +� *& +@model_path/weights/weight.bin��Ã� +linear +x + + +var_1022@ +weight6 +4 +2model_transformer_h_6_mlp_c_proj_weight_transposed1 +bias) +' +%model_transformer_h_6_mlp_c_proj_bias +x_111 +  +� +�* +name + + " +x_111d +const +var_1024 + + +* +name + + " +op_1024*" +val + + + +  +��x +reshape +x + +x_111 +shape + + +var_1024& +input_99 +  + +� +�* +name + + " + +input_99u +add +x + + +input_95 +y + + +input_99' + input_101 +  + +� +�* +name + + " + input_101r +const + x_113_axes_0 + + +*" +name + +" + x_113_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_101 +axes + + x_113_axes_0. +gamma% +# +!model_transformer_h_7_ln_1_weight+ +beta# +! +model_transformer_h_7_ln_1_bias +epsilon + + +var_26# +x_113 +  + +� +�* +name + + " +x_113k +const +var_1044 + + +* +name + + " +op_1044*) +val" + + + + + ����������q +reshape +x + +x_113 +shape + + +var_1044 +var_1045 +  +� +�* +name + + " +op_1045� +constK +3model_transformer_h_7_attn_c_attn_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_7_attn_c_attn_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin��Ë� +linear +x + + +var_1045A +weight7 +5 +3model_transformer_h_7_attn_c_attn_weight_transposed2 +bias* +( +&model_transformer_h_7_attn_c_attn_bias +x_115 +  +� +�* +name + + " +x_115d +const +var_1047 + + +* +name + + " +op_1047*" +val + + + +  +��w +reshape +x + +x_115 +shape + + +var_1047& +var_1048 +  + +� +�* +name + + " +op_1048b +const +tile_7 + + +* +name + + +" +tile_7*# +val + + + + + +���^ +const +var_1049_axis_0 +*$ +name + +" +op_1049_axis_0* +val + + +� +split +x + + +var_1048 + split_sizes + + +tile_7 +axis + +var_1049_axis_0( + +var_1049_0 +  + +� +�( + +var_1049_1 +  + +� +�( + +var_1049_2 +  + +� +�* +name + + " +op_1049d +const +var_1055 + + +* +name + + " +op_1055*" +val + + + +  +�@� +reshape +x + + +var_1049_0 +shape + + +var_1055- + +tensor_115 +  + +� + +@* +name + +" + +tensor_115d +const +var_1061 + + +* +name + + " +op_1061*" +val + + + +  +�@� +reshape +x + + +var_1049_1 +shape + + +var_1061- + +tensor_119 +  + +� + +@* +name + +" + +tensor_119d +const +var_1067 + + +* +name + + " +op_1067*" +val + + + +  +�@� +reshape +x + + +var_1049_2 +shape + + +var_1067- + +tensor_123 +  + +� + +@* +name + +" + +tensor_123c +const +var_1069 + + +* +name + + " +op_1069*! +val + + + + + +{ +const% +attn_weights_85_transpose_x_0 +*3 +name+ +# +!" +attn_weights_85_transpose_x_0* +val + + +{ +const% +attn_weights_85_transpose_y_0 +*3 +name+ +# +!" +attn_weights_85_transpose_y_0* +val + + +z +const# +transpose_86_perm_0 + + +*) +name! + +" +transpose_86_perm_0*! +val + + + + + +z +const# +transpose_87_perm_0 + + +*) +name! + +" +transpose_87_perm_0*! +val + + + + + +� + transpose +x + + +tensor_119 +perm + +transpose_87_perm_00 + transpose_185 +  + + +@ +�*# +name + +" + transpose_185� + transpose +x + + +tensor_115 +perm + +transpose_86_perm_00 + transpose_186 +  + + +� +@*# +name + +" + transpose_186� +matmul +x + + transpose_186 +y + + transpose_1850 + transpose_x! + +attn_weights_85_transpose_x_00 + transpose_y! + +attn_weights_85_transpose_y_03 +attn_weights_85 +  + + +� +�*% +name + +" +attn_weights_85~ +const% +_inversed_attn_weights_87_y_0 + *3 +name+ +# +!" +_inversed_attn_weights_87_y_0* +val +  + + + +>� +mul +x + +attn_weights_85& +y! + +_inversed_attn_weights_87_y_0= +_inversed_attn_weights_87 +  + + +� +�*/ +name' + +" +_inversed_attn_weights_87� +select) +cond! + +causal_mask_1_after_broadcast" +a + +_inversed_attn_weights_87% +b + +mask_value_1_after_broadcast3 +attn_weights_91 +  + + +� +�*% +name + +" +attn_weights_91� +add +x + +attn_weights_91 +y + +attention_mask_1- + input_103 +  + + +� +�* +name + + " + input_103� +softmax +x + + input_103 +axis + +var_63 +attn_weights_93 +  + + +� +�*% +name + +" +attn_weights_93q +const +tensor_125_transpose_x_0 +*. +name& + +" +tensor_125_transpose_x_0* +val + + +q +const +tensor_125_transpose_y_0 +*. +name& + +" +tensor_125_transpose_y_0* +val + + +� + transpose +x + + +tensor_123 +perm + + +var_10690 + transpose_187 +  + + +� +@*# +name + +" + transpose_187� +matmul +x + +attn_weights_93 +y + + transpose_187+ + transpose_x + +tensor_125_transpose_x_0+ + transpose_y + +tensor_125_transpose_y_0- + +tensor_125 +  + + +� +@* +name + +" + +tensor_125c +const +var_1100 + + +* +name + + " +op_1100*! +val + + + + + +k +const +var_1112 + + +* +name + + " +op_1112*) +val" + + + + + ����������� + transpose +x + + +tensor_125 +perm + + +var_11000 + transpose_184 +  + +� + +@*# +name + +" + transpose_184y +reshape +x + + transpose_184 +shape + + +var_1112 +var_1113 +  +� +�* +name + + " +op_1113� +constK +3model_transformer_h_7_attn_c_proj_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_7_attn_c_proj_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin��Ñ� +linear +x + + +var_1113A +weight7 +5 +3model_transformer_h_7_attn_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_7_attn_c_proj_bias +x_119 +  +� +�* +name + + " +x_119d +const +var_1115 + + +* +name + + " +op_1115*" +val + + + +  +��z +reshape +x + +x_119 +shape + + +var_1115' + input_107 +  + +� +�* +name + + " + input_107w +add +x + + input_107 +y + + input_101' + input_109 +  + +� +�* +name + + " + input_109r +const + x_121_axes_0 + + +*" +name + +" + x_121_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_109 +axes + + x_121_axes_0. +gamma% +# +!model_transformer_h_7_ln_2_weight+ +beta# +! +model_transformer_h_7_ln_2_bias +epsilon + + +var_26# +x_121 +  + +� +�* +name + + " +x_121k +const +var_1130 + + +* +name + + " +op_1130*) +val" + + + + + ����������q +reshape +x + +x_121 +shape + + +var_1130 +var_1131 +  +� +�* +name + + " +op_1131� +constH +0model_transformer_h_7_mlp_c_fc_weight_transposed +  +�  +�*F +name> +6 +4"2 +0model_transformer_h_7_mlp_c_fc_weight_transposed*E +val> +  +�  +�*& +@model_path/weights/weight.bin��Ó� +linear +x + + +var_1131> +weight4 +2 +0model_transformer_h_7_mlp_c_fc_weight_transposed/ +bias' +% +#model_transformer_h_7_mlp_c_fc_bias +x_123 +  +� +� * +name + + " +x_123d +const +var_1133 + + +* +name + + " +op_1133*" +val + + + +  +�� z +reshape +x + +x_123 +shape + + +var_1133' + input_111 +  + +� +� * +name + + " + input_111j +const + x_125_mode_0 +*" +name + +" + x_125_mode_0*' +val  + +" +TANH_APPROXIMATIONv +gelu +x + + input_111 +mode + + x_125_mode_0# +x_125 +  + +� +� * +name + + " +x_125k +const +var_1152 + + +* +name + + " +op_1152*) +val" + + + + + ���������� q +reshape +x + +x_125 +shape + + +var_1152 +var_1153 +  +� +� * +name + + " +op_1153� +constJ +2model_transformer_h_7_mlp_c_proj_weight_transposed +  +� +� *H +name@ +8 +6"4 +2model_transformer_h_7_mlp_c_proj_weight_transposed*E +val> +  +� +� *& +@model_path/weights/weight.bin��Û� +linear +x + + +var_1153@ +weight6 +4 +2model_transformer_h_7_mlp_c_proj_weight_transposed1 +bias) +' +%model_transformer_h_7_mlp_c_proj_bias +x_127 +  +� +�* +name + + " +x_127d +const +var_1155 + + +* +name + + " +op_1155*" +val + + + +  +��z +reshape +x + +x_127 +shape + + +var_1155' + input_113 +  + +� +�* +name + + " + input_113w +add +x + + input_109 +y + + input_113' + input_115 +  + +� +�* +name + + " + input_115r +const + x_129_axes_0 + + +*" +name + +" + x_129_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_115 +axes + + x_129_axes_0. +gamma% +# +!model_transformer_h_8_ln_1_weight+ +beta# +! +model_transformer_h_8_ln_1_bias +epsilon + + +var_26# +x_129 +  + +� +�* +name + + " +x_129k +const +var_1175 + + +* +name + + " +op_1175*) +val" + + + + + ����������q +reshape +x + +x_129 +shape + + +var_1175 +var_1176 +  +� +�* +name + + " +op_1176� +constK +3model_transformer_h_8_attn_c_attn_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_8_attn_c_attn_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin��ã� +linear +x + + +var_1176A +weight7 +5 +3model_transformer_h_8_attn_c_attn_weight_transposed2 +bias* +( +&model_transformer_h_8_attn_c_attn_bias +x_131 +  +� +�* +name + + " +x_131d +const +var_1178 + + +* +name + + " +op_1178*" +val + + + +  +��w +reshape +x + +x_131 +shape + + +var_1178& +var_1179 +  + +� +�* +name + + " +op_1179b +const +tile_8 + + +* +name + + +" +tile_8*# +val + + + + + +���^ +const +var_1180_axis_0 +*$ +name + +" +op_1180_axis_0* +val + + +� +split +x + + +var_1179 + split_sizes + + +tile_8 +axis + +var_1180_axis_0( + +var_1180_0 +  + +� +�( + +var_1180_1 +  + +� +�( + +var_1180_2 +  + +� +�* +name + + " +op_1180d +const +var_1186 + + +* +name + + " +op_1186*" +val + + + +  +�@� +reshape +x + + +var_1180_0 +shape + + +var_1186- + +tensor_131 +  + +� + +@* +name + +" + +tensor_131d +const +var_1192 + + +* +name + + " +op_1192*" +val + + + +  +�@� +reshape +x + + +var_1180_1 +shape + + +var_1192- + +tensor_135 +  + +� + +@* +name + +" + +tensor_135d +const +var_1198 + + +* +name + + " +op_1198*" +val + + + +  +�@� +reshape +x + + +var_1180_2 +shape + + +var_1198- + +tensor_139 +  + +� + +@* +name + +" + +tensor_139c +const +var_1200 + + +* +name + + " +op_1200*! +val + + + + + +{ +const% +attn_weights_97_transpose_x_0 +*3 +name+ +# +!" +attn_weights_97_transpose_x_0* +val + + +{ +const% +attn_weights_97_transpose_y_0 +*3 +name+ +# +!" +attn_weights_97_transpose_y_0* +val + + +z +const# +transpose_88_perm_0 + + +*) +name! + +" +transpose_88_perm_0*! +val + + + + + +z +const# +transpose_89_perm_0 + + +*) +name! + +" +transpose_89_perm_0*! +val + + + + + +� + transpose +x + + +tensor_135 +perm + +transpose_89_perm_00 + transpose_181 +  + + +@ +�*# +name + +" + transpose_181� + transpose +x + + +tensor_131 +perm + +transpose_88_perm_00 + transpose_182 +  + + +� +@*# +name + +" + transpose_182� +matmul +x + + transpose_182 +y + + transpose_1810 + transpose_x! + +attn_weights_97_transpose_x_00 + transpose_y! + +attn_weights_97_transpose_y_03 +attn_weights_97 +  + + +� +�*% +name + +" +attn_weights_97~ +const% +_inversed_attn_weights_99_y_0 + *3 +name+ +# +!" +_inversed_attn_weights_99_y_0* +val +  + + + +>� +mul +x + +attn_weights_97& +y! + +_inversed_attn_weights_99_y_0= +_inversed_attn_weights_99 +  + + +� +�*/ +name' + +" +_inversed_attn_weights_99� +select) +cond! + +causal_mask_1_after_broadcast" +a + +_inversed_attn_weights_99% +b + +mask_value_1_after_broadcast4 +attn_weights_103 +  + + +� +�*& +name + +" +attn_weights_103� +add +x + +attn_weights_103 +y + +attention_mask_1- + input_117 +  + + +� +�* +name + + " + input_117� +softmax +x + + input_117 +axis + +var_64 +attn_weights_105 +  + + +� +�*& +name + +" +attn_weights_105q +const +tensor_141_transpose_x_0 +*. +name& + +" +tensor_141_transpose_x_0* +val + + +q +const +tensor_141_transpose_y_0 +*. +name& + +" +tensor_141_transpose_y_0* +val + + +� + transpose +x + + +tensor_139 +perm + + +var_12000 + transpose_183 +  + + +� +@*# +name + +" + transpose_183� +matmul +x + +attn_weights_105 +y + + transpose_183+ + transpose_x + +tensor_141_transpose_x_0+ + transpose_y + +tensor_141_transpose_y_0- + +tensor_141 +  + + +� +@* +name + +" + +tensor_141c +const +var_1231 + + +* +name + + " +op_1231*! +val + + + + + +k +const +var_1243 + + +* +name + + " +op_1243*) +val" + + + + + ����������� + transpose +x + + +tensor_141 +perm + + +var_12310 + transpose_180 +  + +� + +@*# +name + +" + transpose_180y +reshape +x + + transpose_180 +shape + + +var_1243 +var_1244 +  +� +�* +name + + " +op_1244� +constK +3model_transformer_h_8_attn_c_proj_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_8_attn_c_proj_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin��é� +linear +x + + +var_1244A +weight7 +5 +3model_transformer_h_8_attn_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_8_attn_c_proj_bias +x_135 +  +� +�* +name + + " +x_135d +const +var_1246 + + +* +name + + " +op_1246*" +val + + + +  +��z +reshape +x + +x_135 +shape + + +var_1246' + input_121 +  + +� +�* +name + + " + input_121w +add +x + + input_121 +y + + input_115' + input_123 +  + +� +�* +name + + " + input_123r +const + x_137_axes_0 + + +*" +name + +" + x_137_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_123 +axes + + x_137_axes_0. +gamma% +# +!model_transformer_h_8_ln_2_weight+ +beta# +! +model_transformer_h_8_ln_2_bias +epsilon + + +var_26# +x_137 +  + +� +�* +name + + " +x_137k +const +var_1261 + + +* +name + + " +op_1261*) +val" + + + + + ����������q +reshape +x + +x_137 +shape + + +var_1261 +var_1262 +  +� +�* +name + + " +op_1262� +constH +0model_transformer_h_8_mlp_c_fc_weight_transposed +  +�  +�*F +name> +6 +4"2 +0model_transformer_h_8_mlp_c_fc_weight_transposed*E +val> +  +�  +�*& +@model_path/weights/weight.bin��ë� +linear +x + + +var_1262> +weight4 +2 +0model_transformer_h_8_mlp_c_fc_weight_transposed/ +bias' +% +#model_transformer_h_8_mlp_c_fc_bias +x_139 +  +� +� * +name + + " +x_139d +const +var_1264 + + +* +name + + " +op_1264*" +val + + + +  +�� z +reshape +x + +x_139 +shape + + +var_1264' + input_125 +  + +� +� * +name + + " + input_125j +const + x_141_mode_0 +*" +name + +" + x_141_mode_0*' +val  + +" +TANH_APPROXIMATIONv +gelu +x + + input_125 +mode + + x_141_mode_0# +x_141 +  + +� +� * +name + + " +x_141k +const +var_1283 + + +* +name + + " +op_1283*) +val" + + + + + ���������� q +reshape +x + +x_141 +shape + + +var_1283 +var_1284 +  +� +� * +name + + " +op_1284� +constJ +2model_transformer_h_8_mlp_c_proj_weight_transposed +  +� +� *H +name@ +8 +6"4 +2model_transformer_h_8_mlp_c_proj_weight_transposed*E +val> +  +� +� *& +@model_path/weights/weight.bin��ó� +linear +x + + +var_1284@ +weight6 +4 +2model_transformer_h_8_mlp_c_proj_weight_transposed1 +bias) +' +%model_transformer_h_8_mlp_c_proj_bias +x_143 +  +� +�* +name + + " +x_143d +const +var_1286 + + +* +name + + " +op_1286*" +val + + + +  +��z +reshape +x + +x_143 +shape + + +var_1286' + input_127 +  + +� +�* +name + + " + input_127w +add +x + + input_123 +y + + input_127' + input_129 +  + +� +�* +name + + " + input_129r +const + x_145_axes_0 + + +*" +name + +" + x_145_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_129 +axes + + x_145_axes_0. +gamma% +# +!model_transformer_h_9_ln_1_weight+ +beta# +! +model_transformer_h_9_ln_1_bias +epsilon + + +var_26# +x_145 +  + +� +�* +name + + " +x_145k +const +var_1306 + + +* +name + + " +op_1306*) +val" + + + + + ����������q +reshape +x + +x_145 +shape + + +var_1306 +var_1307 +  +� +�* +name + + " +op_1307� +constK +3model_transformer_h_9_attn_c_attn_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_9_attn_c_attn_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin��û� +linear +x + + +var_1307A +weight7 +5 +3model_transformer_h_9_attn_c_attn_weight_transposed2 +bias* +( +&model_transformer_h_9_attn_c_attn_bias +x_147 +  +� +�* +name + + " +x_147d +const +var_1309 + + +* +name + + " +op_1309*" +val + + + +  +��w +reshape +x + +x_147 +shape + + +var_1309& +var_1310 +  + +� +�* +name + + " +op_1310b +const +tile_9 + + +* +name + + +" +tile_9*# +val + + + + + +���^ +const +var_1311_axis_0 +*$ +name + +" +op_1311_axis_0* +val + + +� +split +x + + +var_1310 + split_sizes + + +tile_9 +axis + +var_1311_axis_0( + +var_1311_0 +  + +� +�( + +var_1311_1 +  + +� +�( + +var_1311_2 +  + +� +�* +name + + " +op_1311d +const +var_1317 + + +* +name + + " +op_1317*" +val + + + +  +�@� +reshape +x + + +var_1311_0 +shape + + +var_1317- + +tensor_147 +  + +� + +@* +name + +" + +tensor_147d +const +var_1323 + + +* +name + + " +op_1323*" +val + + + +  +�@� +reshape +x + + +var_1311_1 +shape + + +var_1323- + +tensor_151 +  + +� + +@* +name + +" + +tensor_151d +const +var_1329 + + +* +name + + " +op_1329*" +val + + + +  +�@� +reshape +x + + +var_1311_2 +shape + + +var_1329- + +tensor_155 +  + +� + +@* +name + +" + +tensor_155c +const +var_1331 + + +* +name + + " +op_1331*! +val + + + + + +} +const& +attn_weights_109_transpose_x_0 +*4 +name, +$ +"" +attn_weights_109_transpose_x_0* +val + + +} +const& +attn_weights_109_transpose_y_0 +*4 +name, +$ +"" +attn_weights_109_transpose_y_0* +val + + +z +const# +transpose_90_perm_0 + + +*) +name! + +" +transpose_90_perm_0*! +val + + + + + +z +const# +transpose_91_perm_0 + + +*) +name! + +" +transpose_91_perm_0*! +val + + + + + +� + transpose +x + + +tensor_151 +perm + +transpose_91_perm_00 + transpose_177 +  + + +@ +�*# +name + +" + transpose_177� + transpose +x + + +tensor_147 +perm + +transpose_90_perm_00 + transpose_178 +  + + +� +@*# +name + +" + transpose_178� +matmul +x + + transpose_178 +y + + transpose_1771 + transpose_x" + +attn_weights_109_transpose_x_01 + transpose_y" + +attn_weights_109_transpose_y_04 +attn_weights_109 +  + + +� +�*& +name + +" +attn_weights_109� +const& +_inversed_attn_weights_111_y_0 + *4 +name, +$ +"" +_inversed_attn_weights_111_y_0* +val +  + + + +>� +mul +x + +attn_weights_109' +y" + +_inversed_attn_weights_111_y_0> +_inversed_attn_weights_111 +  + + +� +�*0 +name( + +" +_inversed_attn_weights_111� +select) +cond! + +causal_mask_1_after_broadcast# +a + +_inversed_attn_weights_111% +b + +mask_value_1_after_broadcast4 +attn_weights_115 +  + + +� +�*& +name + +" +attn_weights_115� +add +x + +attn_weights_115 +y + +attention_mask_1- + input_131 +  + + +� +�* +name + + " + input_131� +softmax +x + + input_131 +axis + +var_64 +attn_weights_117 +  + + +� +�*& +name + +" +attn_weights_117q +const +tensor_157_transpose_x_0 +*. +name& + +" +tensor_157_transpose_x_0* +val + + +q +const +tensor_157_transpose_y_0 +*. +name& + +" +tensor_157_transpose_y_0* +val + + +� + transpose +x + + +tensor_155 +perm + + +var_13310 + transpose_179 +  + + +� +@*# +name + +" + transpose_179� +matmul +x + +attn_weights_117 +y + + transpose_179+ + transpose_x + +tensor_157_transpose_x_0+ + transpose_y + +tensor_157_transpose_y_0- + +tensor_157 +  + + +� +@* +name + +" + +tensor_157c +const +var_1362 + + +* +name + + " +op_1362*! +val + + + + + +k +const +var_1374 + + +* +name + + " +op_1374*) +val" + + + + + ����������� + transpose +x + + +tensor_157 +perm + + +var_13620 + transpose_176 +  + +� + +@*# +name + +" + transpose_176y +reshape +x + + transpose_176 +shape + + +var_1374 +var_1375 +  +� +�* +name + + " +op_1375� +constK +3model_transformer_h_9_attn_c_proj_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_9_attn_c_proj_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_1375A +weight7 +5 +3model_transformer_h_9_attn_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_9_attn_c_proj_bias +x_151 +  +� +�* +name + + " +x_151d +const +var_1377 + + +* +name + + " +op_1377*" +val + + + +  +��z +reshape +x + +x_151 +shape + + +var_1377' + input_135 +  + +� +�* +name + + " + input_135w +add +x + + input_135 +y + + input_129' + input_137 +  + +� +�* +name + + " + input_137r +const + x_153_axes_0 + + +*" +name + +" + x_153_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_137 +axes + + x_153_axes_0. +gamma% +# +!model_transformer_h_9_ln_2_weight+ +beta# +! +model_transformer_h_9_ln_2_bias +epsilon + + +var_26# +x_153 +  + +� +�* +name + + " +x_153k +const +var_1392 + + +* +name + + " +op_1392*) +val" + + + + + ����������q +reshape +x + +x_153 +shape + + +var_1392 +var_1393 +  +� +�* +name + + " +op_1393� +constH +0model_transformer_h_9_mlp_c_fc_weight_transposed +  +�  +�*F +name> +6 +4"2 +0model_transformer_h_9_mlp_c_fc_weight_transposed*E +val> +  +�  +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_1393> +weight4 +2 +0model_transformer_h_9_mlp_c_fc_weight_transposed/ +bias' +% +#model_transformer_h_9_mlp_c_fc_bias +x_155 +  +� +� * +name + + " +x_155d +const +var_1395 + + +* +name + + " +op_1395*" +val + + + +  +�� z +reshape +x + +x_155 +shape + + +var_1395' + input_139 +  + +� +� * +name + + " + input_139j +const + x_157_mode_0 +*" +name + +" + x_157_mode_0*' +val  + +" +TANH_APPROXIMATIONv +gelu +x + + input_139 +mode + + x_157_mode_0# +x_157 +  + +� +� * +name + + " +x_157k +const +var_1414 + + +* +name + + " +op_1414*) +val" + + + + + ���������� q +reshape +x + +x_157 +shape + + +var_1414 +var_1415 +  +� +� * +name + + " +op_1415� +constJ +2model_transformer_h_9_mlp_c_proj_weight_transposed +  +� +� *H +name@ +8 +6"4 +2model_transformer_h_9_mlp_c_proj_weight_transposed*E +val> +  +� +� *& +@model_path/weights/weight.bin����� +linear +x + + +var_1415@ +weight6 +4 +2model_transformer_h_9_mlp_c_proj_weight_transposed1 +bias) +' +%model_transformer_h_9_mlp_c_proj_bias +x_159 +  +� +�* +name + + " +x_159d +const +var_1417 + + +* +name + + " +op_1417*" +val + + + +  +��z +reshape +x + +x_159 +shape + + +var_1417' + input_141 +  + +� +�* +name + + " + input_141w +add +x + + input_137 +y + + input_141' + input_143 +  + +� +�* +name + + " + input_143r +const + x_161_axes_0 + + +*" +name + +" + x_161_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_143 +axes + + x_161_axes_0/ +gamma& +$ +"model_transformer_h_10_ln_1_weight, +beta$ +" + model_transformer_h_10_ln_1_bias +epsilon + + +var_26# +x_161 +  + +� +�* +name + + " +x_161k +const +var_1437 + + +* +name + + " +op_1437*) +val" + + + + + ����������q +reshape +x + +x_161 +shape + + +var_1437 +var_1438 +  +� +�* +name + + " +op_1438� +constL +4model_transformer_h_10_attn_c_attn_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_10_attn_c_attn_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_1438B +weight8 +6 +4model_transformer_h_10_attn_c_attn_weight_transposed3 +bias+ +) +'model_transformer_h_10_attn_c_attn_bias +x_163 +  +� +�* +name + + " +x_163d +const +var_1440 + + +* +name + + " +op_1440*" +val + + + +  +��w +reshape +x + +x_163 +shape + + +var_1440& +var_1441 +  + +� +�* +name + + " +op_1441d +const +tile_10 + + +* +name + + " +tile_10*# +val + + + + + +���^ +const +var_1442_axis_0 +*$ +name + +" +op_1442_axis_0* +val + + +� +split +x + + +var_1441 + split_sizes + +tile_10 +axis + +var_1442_axis_0( + +var_1442_0 +  + +� +�( + +var_1442_1 +  + +� +�( + +var_1442_2 +  + +� +�* +name + + " +op_1442d +const +var_1448 + + +* +name + + " +op_1448*" +val + + + +  +�@� +reshape +x + + +var_1442_0 +shape + + +var_1448- + +tensor_163 +  + +� + +@* +name + +" + +tensor_163d +const +var_1454 + + +* +name + + " +op_1454*" +val + + + +  +�@� +reshape +x + + +var_1442_1 +shape + + +var_1454- + +tensor_167 +  + +� + +@* +name + +" + +tensor_167d +const +var_1460 + + +* +name + + " +op_1460*" +val + + + +  +�@� +reshape +x + + +var_1442_2 +shape + + +var_1460- + +tensor_171 +  + +� + +@* +name + +" + +tensor_171c +const +var_1462 + + +* +name + + " +op_1462*! +val + + + + + +} +const& +attn_weights_121_transpose_x_0 +*4 +name, +$ +"" +attn_weights_121_transpose_x_0* +val + + +} +const& +attn_weights_121_transpose_y_0 +*4 +name, +$ +"" +attn_weights_121_transpose_y_0* +val + + +z +const# +transpose_92_perm_0 + + +*) +name! + +" +transpose_92_perm_0*! +val + + + + + +z +const# +transpose_93_perm_0 + + +*) +name! + +" +transpose_93_perm_0*! +val + + + + + +� + transpose +x + + +tensor_167 +perm + +transpose_93_perm_00 + transpose_173 +  + + +@ +�*# +name + +" + transpose_173� + transpose +x + + +tensor_163 +perm + +transpose_92_perm_00 + transpose_174 +  + + +� +@*# +name + +" + transpose_174� +matmul +x + + transpose_174 +y + + transpose_1731 + transpose_x" + +attn_weights_121_transpose_x_01 + transpose_y" + +attn_weights_121_transpose_y_04 +attn_weights_121 +  + + +� +�*& +name + +" +attn_weights_121� +const& +_inversed_attn_weights_123_y_0 + *4 +name, +$ +"" +_inversed_attn_weights_123_y_0* +val +  + + + +>� +mul +x + +attn_weights_121' +y" + +_inversed_attn_weights_123_y_0> +_inversed_attn_weights_123 +  + + +� +�*0 +name( + +" +_inversed_attn_weights_123� +select) +cond! + +causal_mask_1_after_broadcast# +a + +_inversed_attn_weights_123% +b + +mask_value_1_after_broadcast4 +attn_weights_127 +  + + +� +�*& +name + +" +attn_weights_127� +add +x + +attn_weights_127 +y + +attention_mask_1- + input_145 +  + + +� +�* +name + + " + input_145� +softmax +x + + input_145 +axis + +var_64 +attn_weights_129 +  + + +� +�*& +name + +" +attn_weights_129q +const +tensor_173_transpose_x_0 +*. +name& + +" +tensor_173_transpose_x_0* +val + + +q +const +tensor_173_transpose_y_0 +*. +name& + +" +tensor_173_transpose_y_0* +val + + +� + transpose +x + + +tensor_171 +perm + + +var_14620 + transpose_175 +  + + +� +@*# +name + +" + transpose_175� +matmul +x + +attn_weights_129 +y + + transpose_175+ + transpose_x + +tensor_173_transpose_x_0+ + transpose_y + +tensor_173_transpose_y_0- + +tensor_173 +  + + +� +@* +name + +" + +tensor_173c +const +var_1493 + + +* +name + + " +op_1493*! +val + + + + + +k +const +var_1505 + + +* +name + + " +op_1505*) +val" + + + + + ����������� + transpose +x + + +tensor_173 +perm + + +var_14930 + transpose_172 +  + +� + +@*# +name + +" + transpose_172y +reshape +x + + transpose_172 +shape + + +var_1505 +var_1506 +  +� +�* +name + + " +op_1506� +constL +4model_transformer_h_10_attn_c_proj_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_10_attn_c_proj_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_1506B +weight8 +6 +4model_transformer_h_10_attn_c_proj_weight_transposed3 +bias+ +) +'model_transformer_h_10_attn_c_proj_bias +x_167 +  +� +�* +name + + " +x_167d +const +var_1508 + + +* +name + + " +op_1508*" +val + + + +  +��z +reshape +x + +x_167 +shape + + +var_1508' + input_149 +  + +� +�* +name + + " + input_149w +add +x + + input_149 +y + + input_143' + input_151 +  + +� +�* +name + + " + input_151r +const + x_169_axes_0 + + +*" +name + +" + x_169_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_151 +axes + + x_169_axes_0/ +gamma& +$ +"model_transformer_h_10_ln_2_weight, +beta$ +" + model_transformer_h_10_ln_2_bias +epsilon + + +var_26# +x_169 +  + +� +�* +name + + " +x_169k +const +var_1523 + + +* +name + + " +op_1523*) +val" + + + + + ����������q +reshape +x + +x_169 +shape + + +var_1523 +var_1524 +  +� +�* +name + + " +op_1524� +constI +1model_transformer_h_10_mlp_c_fc_weight_transposed +  +�  +�*G +name? +7 +5"3 +1model_transformer_h_10_mlp_c_fc_weight_transposed*E +val> +  +�  +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_1524? +weight5 +3 +1model_transformer_h_10_mlp_c_fc_weight_transposed0 +bias( +& +$model_transformer_h_10_mlp_c_fc_bias +x_171 +  +� +� * +name + + " +x_171d +const +var_1526 + + +* +name + + " +op_1526*" +val + + + +  +�� z +reshape +x + +x_171 +shape + + +var_1526' + input_153 +  + +� +� * +name + + " + input_153j +const + x_173_mode_0 +*" +name + +" + x_173_mode_0*' +val  + +" +TANH_APPROXIMATIONv +gelu +x + + input_153 +mode + + x_173_mode_0# +x_173 +  + +� +� * +name + + " +x_173k +const +var_1545 + + +* +name + + " +op_1545*) +val" + + + + + ���������� q +reshape +x + +x_173 +shape + + +var_1545 +var_1546 +  +� +� * +name + + " +op_1546� +constK +3model_transformer_h_10_mlp_c_proj_weight_transposed +  +� +� *I +nameA +9 +7"5 +3model_transformer_h_10_mlp_c_proj_weight_transposed*E +val> +  +� +� *& +@model_path/weights/weight.bin����� +linear +x + + +var_1546A +weight7 +5 +3model_transformer_h_10_mlp_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_10_mlp_c_proj_bias +x_175 +  +� +�* +name + + " +x_175d +const +var_1548 + + +* +name + + " +op_1548*" +val + + + +  +��z +reshape +x + +x_175 +shape + + +var_1548' + input_155 +  + +� +�* +name + + " + input_155w +add +x + + input_151 +y + + input_155' + input_157 +  + +� +�* +name + + " + input_157r +const + x_177_axes_0 + + +*" +name + +" + x_177_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_157 +axes + + x_177_axes_0/ +gamma& +$ +"model_transformer_h_11_ln_1_weight, +beta$ +" + model_transformer_h_11_ln_1_bias +epsilon + + +var_26# +x_177 +  + +� +�* +name + + " +x_177k +const +var_1568 + + +* +name + + " +op_1568*) +val" + + + + + ����������q +reshape +x + +x_177 +shape + + +var_1568 +var_1569 +  +� +�* +name + + " +op_1569� +constL +4model_transformer_h_11_attn_c_attn_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_11_attn_c_attn_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_1569B +weight8 +6 +4model_transformer_h_11_attn_c_attn_weight_transposed3 +bias+ +) +'model_transformer_h_11_attn_c_attn_bias +x_179 +  +� +�* +name + + " +x_179d +const +var_1571 + + +* +name + + " +op_1571*" +val + + + +  +��w +reshape +x + +x_179 +shape + + +var_1571& +var_1572 +  + +� +�* +name + + " +op_1572d +const +tile_11 + + +* +name + + " +tile_11*# +val + + + + + +���^ +const +var_1573_axis_0 +*$ +name + +" +op_1573_axis_0* +val + + +� +split +x + + +var_1572 + split_sizes + +tile_11 +axis + +var_1573_axis_0( + +var_1573_0 +  + +� +�( + +var_1573_1 +  + +� +�( + +var_1573_2 +  + +� +�* +name + + " +op_1573d +const +var_1579 + + +* +name + + " +op_1579*" +val + + + +  +�@� +reshape +x + + +var_1573_0 +shape + + +var_1579- + +tensor_179 +  + +� + +@* +name + +" + +tensor_179d +const +var_1585 + + +* +name + + " +op_1585*" +val + + + +  +�@� +reshape +x + + +var_1573_1 +shape + + +var_1585- + +tensor_183 +  + +� + +@* +name + +" + +tensor_183d +const +var_1591 + + +* +name + + " +op_1591*" +val + + + +  +�@� +reshape +x + + +var_1573_2 +shape + + +var_1591- + +tensor_187 +  + +� + +@* +name + +" + +tensor_187c +const +var_1593 + + +* +name + + " +op_1593*! +val + + + + + +} +const& +attn_weights_133_transpose_x_0 +*4 +name, +$ +"" +attn_weights_133_transpose_x_0* +val + + +} +const& +attn_weights_133_transpose_y_0 +*4 +name, +$ +"" +attn_weights_133_transpose_y_0* +val + + +z +const# +transpose_94_perm_0 + + +*) +name! + +" +transpose_94_perm_0*! +val + + + + + +z +const# +transpose_95_perm_0 + + +*) +name! + +" +transpose_95_perm_0*! +val + + + + + +� + transpose +x + + +tensor_183 +perm + +transpose_95_perm_00 + transpose_169 +  + + +@ +�*# +name + +" + transpose_169� + transpose +x + + +tensor_179 +perm + +transpose_94_perm_00 + transpose_170 +  + + +� +@*# +name + +" + transpose_170� +matmul +x + + transpose_170 +y + + transpose_1691 + transpose_x" + +attn_weights_133_transpose_x_01 + transpose_y" + +attn_weights_133_transpose_y_04 +attn_weights_133 +  + + +� +�*& +name + +" +attn_weights_133� +const& +_inversed_attn_weights_135_y_0 + *4 +name, +$ +"" +_inversed_attn_weights_135_y_0* +val +  + + + +>� +mul +x + +attn_weights_133' +y" + +_inversed_attn_weights_135_y_0> +_inversed_attn_weights_135 +  + + +� +�*0 +name( + +" +_inversed_attn_weights_135� +select) +cond! + +causal_mask_1_after_broadcast# +a + +_inversed_attn_weights_135% +b + +mask_value_1_after_broadcast4 +attn_weights_139 +  + + +� +�*& +name + +" +attn_weights_139� +add +x + +attn_weights_139 +y + +attention_mask_1- + input_159 +  + + +� +�* +name + + " + input_159� +softmax +x + + input_159 +axis + +var_64 +attn_weights_141 +  + + +� +�*& +name + +" +attn_weights_141q +const +tensor_189_transpose_x_0 +*. +name& + +" +tensor_189_transpose_x_0* +val + + +q +const +tensor_189_transpose_y_0 +*. +name& + +" +tensor_189_transpose_y_0* +val + + +� + transpose +x + + +tensor_187 +perm + + +var_15930 + transpose_171 +  + + +� +@*# +name + +" + transpose_171� +matmul +x + +attn_weights_141 +y + + transpose_171+ + transpose_x + +tensor_189_transpose_x_0+ + transpose_y + +tensor_189_transpose_y_0- + +tensor_189 +  + + +� +@* +name + +" + +tensor_189c +const +var_1624 + + +* +name + + " +op_1624*! +val + + + + + +k +const +var_1636 + + +* +name + + " +op_1636*) +val" + + + + + ����������� + transpose +x + + +tensor_189 +perm + + +var_16240 + transpose_168 +  + +� + +@*# +name + +" + transpose_168y +reshape +x + + transpose_168 +shape + + +var_1636 +var_1637 +  +� +�* +name + + " +op_1637� +constL +4model_transformer_h_11_attn_c_proj_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_11_attn_c_proj_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_1637B +weight8 +6 +4model_transformer_h_11_attn_c_proj_weight_transposed3 +bias+ +) +'model_transformer_h_11_attn_c_proj_bias +x_183 +  +� +�* +name + + " +x_183d +const +var_1639 + + +* +name + + " +op_1639*" +val + + + +  +��z +reshape +x + +x_183 +shape + + +var_1639' + input_163 +  + +� +�* +name + + " + input_163w +add +x + + input_163 +y + + input_157' + input_165 +  + +� +�* +name + + " + input_165r +const + x_185_axes_0 + + +*" +name + +" + x_185_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_165 +axes + + x_185_axes_0/ +gamma& +$ +"model_transformer_h_11_ln_2_weight, +beta$ +" + model_transformer_h_11_ln_2_bias +epsilon + + +var_26# +x_185 +  + +� +�* +name + + " +x_185k +const +var_1654 + + +* +name + + " +op_1654*) +val" + + + + + ����������q +reshape +x + +x_185 +shape + + +var_1654 +var_1655 +  +� +�* +name + + " +op_1655� +constI +1model_transformer_h_11_mlp_c_fc_weight_transposed +  +�  +�*G +name? +7 +5"3 +1model_transformer_h_11_mlp_c_fc_weight_transposed*E +val> +  +�  +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_1655? +weight5 +3 +1model_transformer_h_11_mlp_c_fc_weight_transposed0 +bias( +& +$model_transformer_h_11_mlp_c_fc_bias +x_187 +  +� +� * +name + + " +x_187d +const +var_1657 + + +* +name + + " +op_1657*" +val + + + +  +�� z +reshape +x + +x_187 +shape + + +var_1657' + input_167 +  + +� +� * +name + + " + input_167j +const + x_189_mode_0 +*" +name + +" + x_189_mode_0*' +val  + +" +TANH_APPROXIMATIONv +gelu +x + + input_167 +mode + + x_189_mode_0# +x_189 +  + +� +� * +name + + " +x_189k +const +var_1676 + + +* +name + + " +op_1676*) +val" + + + + + ���������� q +reshape +x + +x_189 +shape + + +var_1676 +var_1677 +  +� +� * +name + + " +op_1677� +constK +3model_transformer_h_11_mlp_c_proj_weight_transposed +  +� +� *I +nameA +9 +7"5 +3model_transformer_h_11_mlp_c_proj_weight_transposed*E +val> +  +� +� *& +@model_path/weights/weight.bin����� +linear +x + + +var_1677A +weight7 +5 +3model_transformer_h_11_mlp_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_11_mlp_c_proj_bias +x_191 +  +� +�* +name + + " +x_191d +const +var_1679 + + +* +name + + " +op_1679*" +val + + + +  +��z +reshape +x + +x_191 +shape + + +var_1679' + input_169 +  + +� +�* +name + + " + input_169w +add +x + + input_165 +y + + input_169' + input_171 +  + +� +�* +name + + " + input_171r +const + x_193_axes_0 + + +*" +name + +" + x_193_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_171 +axes + + x_193_axes_0/ +gamma& +$ +"model_transformer_h_12_ln_1_weight, +beta$ +" + model_transformer_h_12_ln_1_bias +epsilon + + +var_26# +x_193 +  + +� +�* +name + + " +x_193k +const +var_1699 + + +* +name + + " +op_1699*) +val" + + + + + ����������q +reshape +x + +x_193 +shape + + +var_1699 +var_1700 +  +� +�* +name + + " +op_1700� +constL +4model_transformer_h_12_attn_c_attn_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_12_attn_c_attn_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin��Ã� +linear +x + + +var_1700B +weight8 +6 +4model_transformer_h_12_attn_c_attn_weight_transposed3 +bias+ +) +'model_transformer_h_12_attn_c_attn_bias +x_195 +  +� +�* +name + + " +x_195d +const +var_1702 + + +* +name + + " +op_1702*" +val + + + +  +��w +reshape +x + +x_195 +shape + + +var_1702& +var_1703 +  + +� +�* +name + + " +op_1703d +const +tile_12 + + +* +name + + " +tile_12*# +val + + + + + +���^ +const +var_1704_axis_0 +*$ +name + +" +op_1704_axis_0* +val + + +� +split +x + + +var_1703 + split_sizes + +tile_12 +axis + +var_1704_axis_0( + +var_1704_0 +  + +� +�( + +var_1704_1 +  + +� +�( + +var_1704_2 +  + +� +�* +name + + " +op_1704d +const +var_1710 + + +* +name + + " +op_1710*" +val + + + +  +�@� +reshape +x + + +var_1704_0 +shape + + +var_1710- + +tensor_195 +  + +� + +@* +name + +" + +tensor_195d +const +var_1716 + + +* +name + + " +op_1716*" +val + + + +  +�@� +reshape +x + + +var_1704_1 +shape + + +var_1716- + +tensor_199 +  + +� + +@* +name + +" + +tensor_199d +const +var_1722 + + +* +name + + " +op_1722*" +val + + + +  +�@� +reshape +x + + +var_1704_2 +shape + + +var_1722- + +tensor_203 +  + +� + +@* +name + +" + +tensor_203c +const +var_1724 + + +* +name + + " +op_1724*! +val + + + + + +} +const& +attn_weights_145_transpose_x_0 +*4 +name, +$ +"" +attn_weights_145_transpose_x_0* +val + + +} +const& +attn_weights_145_transpose_y_0 +*4 +name, +$ +"" +attn_weights_145_transpose_y_0* +val + + +z +const# +transpose_96_perm_0 + + +*) +name! + +" +transpose_96_perm_0*! +val + + + + + +z +const# +transpose_97_perm_0 + + +*) +name! + +" +transpose_97_perm_0*! +val + + + + + +� + transpose +x + + +tensor_199 +perm + +transpose_97_perm_00 + transpose_165 +  + + +@ +�*# +name + +" + transpose_165� + transpose +x + + +tensor_195 +perm + +transpose_96_perm_00 + transpose_166 +  + + +� +@*# +name + +" + transpose_166� +matmul +x + + transpose_166 +y + + transpose_1651 + transpose_x" + +attn_weights_145_transpose_x_01 + transpose_y" + +attn_weights_145_transpose_y_04 +attn_weights_145 +  + + +� +�*& +name + +" +attn_weights_145� +const& +_inversed_attn_weights_147_y_0 + *4 +name, +$ +"" +_inversed_attn_weights_147_y_0* +val +  + + + +>� +mul +x + +attn_weights_145' +y" + +_inversed_attn_weights_147_y_0> +_inversed_attn_weights_147 +  + + +� +�*0 +name( + +" +_inversed_attn_weights_147� +select) +cond! + +causal_mask_1_after_broadcast# +a + +_inversed_attn_weights_147% +b + +mask_value_1_after_broadcast4 +attn_weights_151 +  + + +� +�*& +name + +" +attn_weights_151� +add +x + +attn_weights_151 +y + +attention_mask_1- + input_173 +  + + +� +�* +name + + " + input_173� +softmax +x + + input_173 +axis + +var_64 +attn_weights_153 +  + + +� +�*& +name + +" +attn_weights_153q +const +tensor_205_transpose_x_0 +*. +name& + +" +tensor_205_transpose_x_0* +val + + +q +const +tensor_205_transpose_y_0 +*. +name& + +" +tensor_205_transpose_y_0* +val + + +� + transpose +x + + +tensor_203 +perm + + +var_17240 + transpose_167 +  + + +� +@*# +name + +" + transpose_167� +matmul +x + +attn_weights_153 +y + + transpose_167+ + transpose_x + +tensor_205_transpose_x_0+ + transpose_y + +tensor_205_transpose_y_0- + +tensor_205 +  + + +� +@* +name + +" + +tensor_205c +const +var_1755 + + +* +name + + " +op_1755*! +val + + + + + +k +const +var_1767 + + +* +name + + " +op_1767*) +val" + + + + + ����������� + transpose +x + + +tensor_205 +perm + + +var_17550 + transpose_164 +  + +� + +@*# +name + +" + transpose_164y +reshape +x + + transpose_164 +shape + + +var_1767 +var_1768 +  +� +�* +name + + " +op_1768� +constL +4model_transformer_h_12_attn_c_proj_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_12_attn_c_proj_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin��É� +linear +x + + +var_1768B +weight8 +6 +4model_transformer_h_12_attn_c_proj_weight_transposed3 +bias+ +) +'model_transformer_h_12_attn_c_proj_bias +x_199 +  +� +�* +name + + " +x_199d +const +var_1770 + + +* +name + + " +op_1770*" +val + + + +  +��z +reshape +x + +x_199 +shape + + +var_1770' + input_177 +  + +� +�* +name + + " + input_177w +add +x + + input_177 +y + + input_171' + input_179 +  + +� +�* +name + + " + input_179r +const + x_201_axes_0 + + +*" +name + +" + x_201_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_179 +axes + + x_201_axes_0/ +gamma& +$ +"model_transformer_h_12_ln_2_weight, +beta$ +" + model_transformer_h_12_ln_2_bias +epsilon + + +var_26# +x_201 +  + +� +�* +name + + " +x_201k +const +var_1785 + + +* +name + + " +op_1785*) +val" + + + + + ����������q +reshape +x + +x_201 +shape + + +var_1785 +var_1786 +  +� +�* +name + + " +op_1786� +constI +1model_transformer_h_12_mlp_c_fc_weight_transposed +  +�  +�*G +name? +7 +5"3 +1model_transformer_h_12_mlp_c_fc_weight_transposed*E +val> +  +�  +�*& +@model_path/weights/weight.bin��Ë� +linear +x + + +var_1786? +weight5 +3 +1model_transformer_h_12_mlp_c_fc_weight_transposed0 +bias( +& +$model_transformer_h_12_mlp_c_fc_bias +x_203 +  +� +� * +name + + " +x_203d +const +var_1788 + + +* +name + + " +op_1788*" +val + + + +  +�� z +reshape +x + +x_203 +shape + + +var_1788' + input_181 +  + +� +� * +name + + " + input_181j +const + x_205_mode_0 +*" +name + +" + x_205_mode_0*' +val  + +" +TANH_APPROXIMATIONv +gelu +x + + input_181 +mode + + x_205_mode_0# +x_205 +  + +� +� * +name + + " +x_205k +const +var_1807 + + +* +name + + " +op_1807*) +val" + + + + + ���������� q +reshape +x + +x_205 +shape + + +var_1807 +var_1808 +  +� +� * +name + + " +op_1808� +constK +3model_transformer_h_12_mlp_c_proj_weight_transposed +  +� +� *I +nameA +9 +7"5 +3model_transformer_h_12_mlp_c_proj_weight_transposed*E +val> +  +� +� *& +@model_path/weights/weight.bin��Ó� +linear +x + + +var_1808A +weight7 +5 +3model_transformer_h_12_mlp_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_12_mlp_c_proj_bias +x_207 +  +� +�* +name + + " +x_207d +const +var_1810 + + +* +name + + " +op_1810*" +val + + + +  +��z +reshape +x + +x_207 +shape + + +var_1810' + input_183 +  + +� +�* +name + + " + input_183w +add +x + + input_179 +y + + input_183' + input_185 +  + +� +�* +name + + " + input_185r +const + x_209_axes_0 + + +*" +name + +" + x_209_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_185 +axes + + x_209_axes_0/ +gamma& +$ +"model_transformer_h_13_ln_1_weight, +beta$ +" + model_transformer_h_13_ln_1_bias +epsilon + + +var_26# +x_209 +  + +� +�* +name + + " +x_209k +const +var_1830 + + +* +name + + " +op_1830*) +val" + + + + + ����������q +reshape +x + +x_209 +shape + + +var_1830 +var_1831 +  +� +�* +name + + " +op_1831� +constL +4model_transformer_h_13_attn_c_attn_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_13_attn_c_attn_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin��Û� +linear +x + + +var_1831B +weight8 +6 +4model_transformer_h_13_attn_c_attn_weight_transposed3 +bias+ +) +'model_transformer_h_13_attn_c_attn_bias +x_211 +  +� +�* +name + + " +x_211d +const +var_1833 + + +* +name + + " +op_1833*" +val + + + +  +��w +reshape +x + +x_211 +shape + + +var_1833& +var_1834 +  + +� +�* +name + + " +op_1834d +const +tile_13 + + +* +name + + " +tile_13*# +val + + + + + +���^ +const +var_1835_axis_0 +*$ +name + +" +op_1835_axis_0* +val + + +� +split +x + + +var_1834 + split_sizes + +tile_13 +axis + +var_1835_axis_0( + +var_1835_0 +  + +� +�( + +var_1835_1 +  + +� +�( + +var_1835_2 +  + +� +�* +name + + " +op_1835d +const +var_1841 + + +* +name + + " +op_1841*" +val + + + +  +�@� +reshape +x + + +var_1835_0 +shape + + +var_1841- + +tensor_211 +  + +� + +@* +name + +" + +tensor_211d +const +var_1847 + + +* +name + + " +op_1847*" +val + + + +  +�@� +reshape +x + + +var_1835_1 +shape + + +var_1847- + +tensor_215 +  + +� + +@* +name + +" + +tensor_215d +const +var_1853 + + +* +name + + " +op_1853*" +val + + + +  +�@� +reshape +x + + +var_1835_2 +shape + + +var_1853- + +tensor_219 +  + +� + +@* +name + +" + +tensor_219c +const +var_1855 + + +* +name + + " +op_1855*! +val + + + + + +} +const& +attn_weights_157_transpose_x_0 +*4 +name, +$ +"" +attn_weights_157_transpose_x_0* +val + + +} +const& +attn_weights_157_transpose_y_0 +*4 +name, +$ +"" +attn_weights_157_transpose_y_0* +val + + +z +const# +transpose_98_perm_0 + + +*) +name! + +" +transpose_98_perm_0*! +val + + + + + +z +const# +transpose_99_perm_0 + + +*) +name! + +" +transpose_99_perm_0*! +val + + + + + +� + transpose +x + + +tensor_215 +perm + +transpose_99_perm_00 + transpose_161 +  + + +@ +�*# +name + +" + transpose_161� + transpose +x + + +tensor_211 +perm + +transpose_98_perm_00 + transpose_162 +  + + +� +@*# +name + +" + transpose_162� +matmul +x + + transpose_162 +y + + transpose_1611 + transpose_x" + +attn_weights_157_transpose_x_01 + transpose_y" + +attn_weights_157_transpose_y_04 +attn_weights_157 +  + + +� +�*& +name + +" +attn_weights_157� +const& +_inversed_attn_weights_159_y_0 + *4 +name, +$ +"" +_inversed_attn_weights_159_y_0* +val +  + + + +>� +mul +x + +attn_weights_157' +y" + +_inversed_attn_weights_159_y_0> +_inversed_attn_weights_159 +  + + +� +�*0 +name( + +" +_inversed_attn_weights_159� +select) +cond! + +causal_mask_1_after_broadcast# +a + +_inversed_attn_weights_159% +b + +mask_value_1_after_broadcast4 +attn_weights_163 +  + + +� +�*& +name + +" +attn_weights_163� +add +x + +attn_weights_163 +y + +attention_mask_1- + input_187 +  + + +� +�* +name + + " + input_187� +softmax +x + + input_187 +axis + +var_64 +attn_weights_165 +  + + +� +�*& +name + +" +attn_weights_165q +const +tensor_221_transpose_x_0 +*. +name& + +" +tensor_221_transpose_x_0* +val + + +q +const +tensor_221_transpose_y_0 +*. +name& + +" +tensor_221_transpose_y_0* +val + + +� + transpose +x + + +tensor_219 +perm + + +var_18550 + transpose_163 +  + + +� +@*# +name + +" + transpose_163� +matmul +x + +attn_weights_165 +y + + transpose_163+ + transpose_x + +tensor_221_transpose_x_0+ + transpose_y + +tensor_221_transpose_y_0- + +tensor_221 +  + + +� +@* +name + +" + +tensor_221c +const +var_1886 + + +* +name + + " +op_1886*! +val + + + + + +k +const +var_1898 + + +* +name + + " +op_1898*) +val" + + + + + ����������� + transpose +x + + +tensor_221 +perm + + +var_18860 + transpose_160 +  + +� + +@*# +name + +" + transpose_160y +reshape +x + + transpose_160 +shape + + +var_1898 +var_1899 +  +� +�* +name + + " +op_1899� +constL +4model_transformer_h_13_attn_c_proj_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_13_attn_c_proj_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin��á� +linear +x + + +var_1899B +weight8 +6 +4model_transformer_h_13_attn_c_proj_weight_transposed3 +bias+ +) +'model_transformer_h_13_attn_c_proj_bias +x_215 +  +� +�* +name + + " +x_215d +const +var_1901 + + +* +name + + " +op_1901*" +val + + + +  +��z +reshape +x + +x_215 +shape + + +var_1901' + input_191 +  + +� +�* +name + + " + input_191w +add +x + + input_191 +y + + input_185' + input_193 +  + +� +�* +name + + " + input_193r +const + x_217_axes_0 + + +*" +name + +" + x_217_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_193 +axes + + x_217_axes_0/ +gamma& +$ +"model_transformer_h_13_ln_2_weight, +beta$ +" + model_transformer_h_13_ln_2_bias +epsilon + + +var_26# +x_217 +  + +� +�* +name + + " +x_217k +const +var_1916 + + +* +name + + " +op_1916*) +val" + + + + + ����������q +reshape +x + +x_217 +shape + + +var_1916 +var_1917 +  +� +�* +name + + " +op_1917� +constI +1model_transformer_h_13_mlp_c_fc_weight_transposed +  +�  +�*G +name? +7 +5"3 +1model_transformer_h_13_mlp_c_fc_weight_transposed*E +val> +  +�  +�*& +@model_path/weights/weight.bin��ã� +linear +x + + +var_1917? +weight5 +3 +1model_transformer_h_13_mlp_c_fc_weight_transposed0 +bias( +& +$model_transformer_h_13_mlp_c_fc_bias +x_219 +  +� +� * +name + + " +x_219d +const +var_1919 + + +* +name + + " +op_1919*" +val + + + +  +�� z +reshape +x + +x_219 +shape + + +var_1919' + input_195 +  + +� +� * +name + + " + input_195j +const + x_221_mode_0 +*" +name + +" + x_221_mode_0*' +val  + +" +TANH_APPROXIMATIONv +gelu +x + + input_195 +mode + + x_221_mode_0# +x_221 +  + +� +� * +name + + " +x_221k +const +var_1938 + + +* +name + + " +op_1938*) +val" + + + + + ���������� q +reshape +x + +x_221 +shape + + +var_1938 +var_1939 +  +� +� * +name + + " +op_1939� +constK +3model_transformer_h_13_mlp_c_proj_weight_transposed +  +� +� *I +nameA +9 +7"5 +3model_transformer_h_13_mlp_c_proj_weight_transposed*E +val> +  +� +� *& +@model_path/weights/weight.bin��ë� +linear +x + + +var_1939A +weight7 +5 +3model_transformer_h_13_mlp_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_13_mlp_c_proj_bias +x_223 +  +� +�* +name + + " +x_223d +const +var_1941 + + +* +name + + " +op_1941*" +val + + + +  +��z +reshape +x + +x_223 +shape + + +var_1941' + input_197 +  + +� +�* +name + + " + input_197w +add +x + + input_193 +y + + input_197' + input_199 +  + +� +�* +name + + " + input_199r +const + x_225_axes_0 + + +*" +name + +" + x_225_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_199 +axes + + x_225_axes_0/ +gamma& +$ +"model_transformer_h_14_ln_1_weight, +beta$ +" + model_transformer_h_14_ln_1_bias +epsilon + + +var_26# +x_225 +  + +� +�* +name + + " +x_225k +const +var_1961 + + +* +name + + " +op_1961*) +val" + + + + + ����������q +reshape +x + +x_225 +shape + + +var_1961 +var_1962 +  +� +�* +name + + " +op_1962� +constL +4model_transformer_h_14_attn_c_attn_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_14_attn_c_attn_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin��ó� +linear +x + + +var_1962B +weight8 +6 +4model_transformer_h_14_attn_c_attn_weight_transposed3 +bias+ +) +'model_transformer_h_14_attn_c_attn_bias +x_227 +  +� +�* +name + + " +x_227d +const +var_1964 + + +* +name + + " +op_1964*" +val + + + +  +��w +reshape +x + +x_227 +shape + + +var_1964& +var_1965 +  + +� +�* +name + + " +op_1965d +const +tile_14 + + +* +name + + " +tile_14*# +val + + + + + +���^ +const +var_1966_axis_0 +*$ +name + +" +op_1966_axis_0* +val + + +� +split +x + + +var_1965 + split_sizes + +tile_14 +axis + +var_1966_axis_0( + +var_1966_0 +  + +� +�( + +var_1966_1 +  + +� +�( + +var_1966_2 +  + +� +�* +name + + " +op_1966d +const +var_1972 + + +* +name + + " +op_1972*" +val + + + +  +�@� +reshape +x + + +var_1966_0 +shape + + +var_1972- + +tensor_227 +  + +� + +@* +name + +" + +tensor_227d +const +var_1978 + + +* +name + + " +op_1978*" +val + + + +  +�@� +reshape +x + + +var_1966_1 +shape + + +var_1978- + +tensor_231 +  + +� + +@* +name + +" + +tensor_231d +const +var_1984 + + +* +name + + " +op_1984*" +val + + + +  +�@� +reshape +x + + +var_1966_2 +shape + + +var_1984- + +tensor_235 +  + +� + +@* +name + +" + +tensor_235c +const +var_1986 + + +* +name + + " +op_1986*! +val + + + + + +} +const& +attn_weights_169_transpose_x_0 +*4 +name, +$ +"" +attn_weights_169_transpose_x_0* +val + + +} +const& +attn_weights_169_transpose_y_0 +*4 +name, +$ +"" +attn_weights_169_transpose_y_0* +val + + +| +const$ +transpose_100_perm_0 + + +** +name" + +" +transpose_100_perm_0*! +val + + + + + +| +const$ +transpose_101_perm_0 + + +** +name" + +" +transpose_101_perm_0*! +val + + + + + +� + transpose +x + + +tensor_231 +perm + +transpose_101_perm_00 + transpose_157 +  + + +@ +�*# +name + +" + transpose_157� + transpose +x + + +tensor_227 +perm + +transpose_100_perm_00 + transpose_158 +  + + +� +@*# +name + +" + transpose_158� +matmul +x + + transpose_158 +y + + transpose_1571 + transpose_x" + +attn_weights_169_transpose_x_01 + transpose_y" + +attn_weights_169_transpose_y_04 +attn_weights_169 +  + + +� +�*& +name + +" +attn_weights_169� +const& +_inversed_attn_weights_171_y_0 + *4 +name, +$ +"" +_inversed_attn_weights_171_y_0* +val +  + + + +>� +mul +x + +attn_weights_169' +y" + +_inversed_attn_weights_171_y_0> +_inversed_attn_weights_171 +  + + +� +�*0 +name( + +" +_inversed_attn_weights_171� +select) +cond! + +causal_mask_1_after_broadcast# +a + +_inversed_attn_weights_171% +b + +mask_value_1_after_broadcast4 +attn_weights_175 +  + + +� +�*& +name + +" +attn_weights_175� +add +x + +attn_weights_175 +y + +attention_mask_1- + input_201 +  + + +� +�* +name + + " + input_201� +softmax +x + + input_201 +axis + +var_64 +attn_weights_177 +  + + +� +�*& +name + +" +attn_weights_177q +const +tensor_237_transpose_x_0 +*. +name& + +" +tensor_237_transpose_x_0* +val + + +q +const +tensor_237_transpose_y_0 +*. +name& + +" +tensor_237_transpose_y_0* +val + + +� + transpose +x + + +tensor_235 +perm + + +var_19860 + transpose_159 +  + + +� +@*# +name + +" + transpose_159� +matmul +x + +attn_weights_177 +y + + transpose_159+ + transpose_x + +tensor_237_transpose_x_0+ + transpose_y + +tensor_237_transpose_y_0- + +tensor_237 +  + + +� +@* +name + +" + +tensor_237c +const +var_2017 + + +* +name + + " +op_2017*! +val + + + + + +k +const +var_2029 + + +* +name + + " +op_2029*) +val" + + + + + ����������� + transpose +x + + +tensor_237 +perm + + +var_20170 + transpose_156 +  + +� + +@*# +name + +" + transpose_156y +reshape +x + + transpose_156 +shape + + +var_2029 +var_2030 +  +� +�* +name + + " +op_2030� +constL +4model_transformer_h_14_attn_c_proj_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_14_attn_c_proj_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin��ù� +linear +x + + +var_2030B +weight8 +6 +4model_transformer_h_14_attn_c_proj_weight_transposed3 +bias+ +) +'model_transformer_h_14_attn_c_proj_bias +x_231 +  +� +�* +name + + " +x_231d +const +var_2032 + + +* +name + + " +op_2032*" +val + + + +  +��z +reshape +x + +x_231 +shape + + +var_2032' + input_205 +  + +� +�* +name + + " + input_205w +add +x + + input_205 +y + + input_199' + input_207 +  + +� +�* +name + + " + input_207r +const + x_233_axes_0 + + +*" +name + +" + x_233_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_207 +axes + + x_233_axes_0/ +gamma& +$ +"model_transformer_h_14_ln_2_weight, +beta$ +" + model_transformer_h_14_ln_2_bias +epsilon + + +var_26# +x_233 +  + +� +�* +name + + " +x_233k +const +var_2047 + + +* +name + + " +op_2047*) +val" + + + + + ����������q +reshape +x + +x_233 +shape + + +var_2047 +var_2048 +  +� +�* +name + + " +op_2048� +constI +1model_transformer_h_14_mlp_c_fc_weight_transposed +  +�  +�*G +name? +7 +5"3 +1model_transformer_h_14_mlp_c_fc_weight_transposed*E +val> +  +�  +�*& +@model_path/weights/weight.bin��û� +linear +x + + +var_2048? +weight5 +3 +1model_transformer_h_14_mlp_c_fc_weight_transposed0 +bias( +& +$model_transformer_h_14_mlp_c_fc_bias +x_235 +  +� +� * +name + + " +x_235d +const +var_2050 + + +* +name + + " +op_2050*" +val + + + +  +�� z +reshape +x + +x_235 +shape + + +var_2050' + input_209 +  + +� +� * +name + + " + input_209j +const + x_237_mode_0 +*" +name + +" + x_237_mode_0*' +val  + +" +TANH_APPROXIMATIONv +gelu +x + + input_209 +mode + + x_237_mode_0# +x_237 +  + +� +� * +name + + " +x_237k +const +var_2069 + + +* +name + + " +op_2069*) +val" + + + + + ���������� q +reshape +x + +x_237 +shape + + +var_2069 +var_2070 +  +� +� * +name + + " +op_2070� +constK +3model_transformer_h_14_mlp_c_proj_weight_transposed +  +� +� *I +nameA +9 +7"5 +3model_transformer_h_14_mlp_c_proj_weight_transposed*E +val> +  +� +� *& +@model_path/weights/weight.bin����� +linear +x + + +var_2070A +weight7 +5 +3model_transformer_h_14_mlp_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_14_mlp_c_proj_bias +x_239 +  +� +�* +name + + " +x_239d +const +var_2072 + + +* +name + + " +op_2072*" +val + + + +  +��z +reshape +x + +x_239 +shape + + +var_2072' + input_211 +  + +� +�* +name + + " + input_211w +add +x + + input_207 +y + + input_211' + input_213 +  + +� +�* +name + + " + input_213r +const + x_241_axes_0 + + +*" +name + +" + x_241_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_213 +axes + + x_241_axes_0/ +gamma& +$ +"model_transformer_h_15_ln_1_weight, +beta$ +" + model_transformer_h_15_ln_1_bias +epsilon + + +var_26# +x_241 +  + +� +�* +name + + " +x_241k +const +var_2092 + + +* +name + + " +op_2092*) +val" + + + + + ����������q +reshape +x + +x_241 +shape + + +var_2092 +var_2093 +  +� +�* +name + + " +op_2093� +constL +4model_transformer_h_15_attn_c_attn_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_15_attn_c_attn_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_2093B +weight8 +6 +4model_transformer_h_15_attn_c_attn_weight_transposed3 +bias+ +) +'model_transformer_h_15_attn_c_attn_bias +x_243 +  +� +�* +name + + " +x_243d +const +var_2095 + + +* +name + + " +op_2095*" +val + + + +  +��w +reshape +x + +x_243 +shape + + +var_2095& +var_2096 +  + +� +�* +name + + " +op_2096d +const +tile_15 + + +* +name + + " +tile_15*# +val + + + + + +���^ +const +var_2097_axis_0 +*$ +name + +" +op_2097_axis_0* +val + + +� +split +x + + +var_2096 + split_sizes + +tile_15 +axis + +var_2097_axis_0( + +var_2097_0 +  + +� +�( + +var_2097_1 +  + +� +�( + +var_2097_2 +  + +� +�* +name + + " +op_2097d +const +var_2103 + + +* +name + + " +op_2103*" +val + + + +  +�@� +reshape +x + + +var_2097_0 +shape + + +var_2103- + +tensor_243 +  + +� + +@* +name + +" + +tensor_243d +const +var_2109 + + +* +name + + " +op_2109*" +val + + + +  +�@� +reshape +x + + +var_2097_1 +shape + + +var_2109- + +tensor_247 +  + +� + +@* +name + +" + +tensor_247d +const +var_2115 + + +* +name + + " +op_2115*" +val + + + +  +�@� +reshape +x + + +var_2097_2 +shape + + +var_2115- + +tensor_251 +  + +� + +@* +name + +" + +tensor_251c +const +var_2117 + + +* +name + + " +op_2117*! +val + + + + + +} +const& +attn_weights_181_transpose_x_0 +*4 +name, +$ +"" +attn_weights_181_transpose_x_0* +val + + +} +const& +attn_weights_181_transpose_y_0 +*4 +name, +$ +"" +attn_weights_181_transpose_y_0* +val + + +| +const$ +transpose_102_perm_0 + + +** +name" + +" +transpose_102_perm_0*! +val + + + + + +| +const$ +transpose_103_perm_0 + + +** +name" + +" +transpose_103_perm_0*! +val + + + + + +� + transpose +x + + +tensor_247 +perm + +transpose_103_perm_00 + transpose_153 +  + + +@ +�*# +name + +" + transpose_153� + transpose +x + + +tensor_243 +perm + +transpose_102_perm_00 + transpose_154 +  + + +� +@*# +name + +" + transpose_154� +matmul +x + + transpose_154 +y + + transpose_1531 + transpose_x" + +attn_weights_181_transpose_x_01 + transpose_y" + +attn_weights_181_transpose_y_04 +attn_weights_181 +  + + +� +�*& +name + +" +attn_weights_181� +const& +_inversed_attn_weights_183_y_0 + *4 +name, +$ +"" +_inversed_attn_weights_183_y_0* +val +  + + + +>� +mul +x + +attn_weights_181' +y" + +_inversed_attn_weights_183_y_0> +_inversed_attn_weights_183 +  + + +� +�*0 +name( + +" +_inversed_attn_weights_183� +select) +cond! + +causal_mask_1_after_broadcast# +a + +_inversed_attn_weights_183% +b + +mask_value_1_after_broadcast4 +attn_weights_187 +  + + +� +�*& +name + +" +attn_weights_187� +add +x + +attn_weights_187 +y + +attention_mask_1- + input_215 +  + + +� +�* +name + + " + input_215� +softmax +x + + input_215 +axis + +var_64 +attn_weights_189 +  + + +� +�*& +name + +" +attn_weights_189q +const +tensor_253_transpose_x_0 +*. +name& + +" +tensor_253_transpose_x_0* +val + + +q +const +tensor_253_transpose_y_0 +*. +name& + +" +tensor_253_transpose_y_0* +val + + +� + transpose +x + + +tensor_251 +perm + + +var_21170 + transpose_155 +  + + +� +@*# +name + +" + transpose_155� +matmul +x + +attn_weights_189 +y + + transpose_155+ + transpose_x + +tensor_253_transpose_x_0+ + transpose_y + +tensor_253_transpose_y_0- + +tensor_253 +  + + +� +@* +name + +" + +tensor_253c +const +var_2148 + + +* +name + + " +op_2148*! +val + + + + + +k +const +var_2160 + + +* +name + + " +op_2160*) +val" + + + + + ����������� + transpose +x + + +tensor_253 +perm + + +var_21480 + transpose_152 +  + +� + +@*# +name + +" + transpose_152y +reshape +x + + transpose_152 +shape + + +var_2160 +var_2161 +  +� +�* +name + + " +op_2161� +constL +4model_transformer_h_15_attn_c_proj_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_15_attn_c_proj_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_2161B +weight8 +6 +4model_transformer_h_15_attn_c_proj_weight_transposed3 +bias+ +) +'model_transformer_h_15_attn_c_proj_bias +x_247 +  +� +�* +name + + " +x_247d +const +var_2163 + + +* +name + + " +op_2163*" +val + + + +  +��z +reshape +x + +x_247 +shape + + +var_2163' + input_219 +  + +� +�* +name + + " + input_219w +add +x + + input_219 +y + + input_213' + input_221 +  + +� +�* +name + + " + input_221r +const + x_249_axes_0 + + +*" +name + +" + x_249_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_221 +axes + + x_249_axes_0/ +gamma& +$ +"model_transformer_h_15_ln_2_weight, +beta$ +" + model_transformer_h_15_ln_2_bias +epsilon + + +var_26# +x_249 +  + +� +�* +name + + " +x_249k +const +var_2178 + + +* +name + + " +op_2178*) +val" + + + + + ����������q +reshape +x + +x_249 +shape + + +var_2178 +var_2179 +  +� +�* +name + + " +op_2179� +constI +1model_transformer_h_15_mlp_c_fc_weight_transposed +  +�  +�*G +name? +7 +5"3 +1model_transformer_h_15_mlp_c_fc_weight_transposed*E +val> +  +�  +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_2179? +weight5 +3 +1model_transformer_h_15_mlp_c_fc_weight_transposed0 +bias( +& +$model_transformer_h_15_mlp_c_fc_bias +x_251 +  +� +� * +name + + " +x_251d +const +var_2181 + + +* +name + + " +op_2181*" +val + + + +  +�� z +reshape +x + +x_251 +shape + + +var_2181' + input_223 +  + +� +� * +name + + " + input_223j +const + x_253_mode_0 +*" +name + +" + x_253_mode_0*' +val  + +" +TANH_APPROXIMATIONv +gelu +x + + input_223 +mode + + x_253_mode_0# +x_253 +  + +� +� * +name + + " +x_253k +const +var_2200 + + +* +name + + " +op_2200*) +val" + + + + + ���������� q +reshape +x + +x_253 +shape + + +var_2200 +var_2201 +  +� +� * +name + + " +op_2201� +constK +3model_transformer_h_15_mlp_c_proj_weight_transposed +  +� +� *I +nameA +9 +7"5 +3model_transformer_h_15_mlp_c_proj_weight_transposed*E +val> +  +� +� *& +@model_path/weights/weight.bin����� +linear +x + + +var_2201A +weight7 +5 +3model_transformer_h_15_mlp_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_15_mlp_c_proj_bias +x_255 +  +� +�* +name + + " +x_255d +const +var_2203 + + +* +name + + " +op_2203*" +val + + + +  +��z +reshape +x + +x_255 +shape + + +var_2203' + input_225 +  + +� +�* +name + + " + input_225w +add +x + + input_221 +y + + input_225' + input_227 +  + +� +�* +name + + " + input_227r +const + x_257_axes_0 + + +*" +name + +" + x_257_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_227 +axes + + x_257_axes_0/ +gamma& +$ +"model_transformer_h_16_ln_1_weight, +beta$ +" + model_transformer_h_16_ln_1_bias +epsilon + + +var_26# +x_257 +  + +� +�* +name + + " +x_257k +const +var_2223 + + +* +name + + " +op_2223*) +val" + + + + + ����������q +reshape +x + +x_257 +shape + + +var_2223 +var_2224 +  +� +�* +name + + " +op_2224� +constL +4model_transformer_h_16_attn_c_attn_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_16_attn_c_attn_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_2224B +weight8 +6 +4model_transformer_h_16_attn_c_attn_weight_transposed3 +bias+ +) +'model_transformer_h_16_attn_c_attn_bias +x_259 +  +� +�* +name + + " +x_259d +const +var_2226 + + +* +name + + " +op_2226*" +val + + + +  +��w +reshape +x + +x_259 +shape + + +var_2226& +var_2227 +  + +� +�* +name + + " +op_2227d +const +tile_16 + + +* +name + + " +tile_16*# +val + + + + + +���^ +const +var_2228_axis_0 +*$ +name + +" +op_2228_axis_0* +val + + +� +split +x + + +var_2227 + split_sizes + +tile_16 +axis + +var_2228_axis_0( + +var_2228_0 +  + +� +�( + +var_2228_1 +  + +� +�( + +var_2228_2 +  + +� +�* +name + + " +op_2228d +const +var_2234 + + +* +name + + " +op_2234*" +val + + + +  +�@� +reshape +x + + +var_2228_0 +shape + + +var_2234- + +tensor_259 +  + +� + +@* +name + +" + +tensor_259d +const +var_2240 + + +* +name + + " +op_2240*" +val + + + +  +�@� +reshape +x + + +var_2228_1 +shape + + +var_2240- + +tensor_263 +  + +� + +@* +name + +" + +tensor_263d +const +var_2246 + + +* +name + + " +op_2246*" +val + + + +  +�@� +reshape +x + + +var_2228_2 +shape + + +var_2246- + +tensor_267 +  + +� + +@* +name + +" + +tensor_267c +const +var_2248 + + +* +name + + " +op_2248*! +val + + + + + +} +const& +attn_weights_193_transpose_x_0 +*4 +name, +$ +"" +attn_weights_193_transpose_x_0* +val + + +} +const& +attn_weights_193_transpose_y_0 +*4 +name, +$ +"" +attn_weights_193_transpose_y_0* +val + + +| +const$ +transpose_104_perm_0 + + +** +name" + +" +transpose_104_perm_0*! +val + + + + + +| +const$ +transpose_105_perm_0 + + +** +name" + +" +transpose_105_perm_0*! +val + + + + + +� + transpose +x + + +tensor_263 +perm + +transpose_105_perm_00 + transpose_149 +  + + +@ +�*# +name + +" + transpose_149� + transpose +x + + +tensor_259 +perm + +transpose_104_perm_00 + transpose_150 +  + + +� +@*# +name + +" + transpose_150� +matmul +x + + transpose_150 +y + + transpose_1491 + transpose_x" + +attn_weights_193_transpose_x_01 + transpose_y" + +attn_weights_193_transpose_y_04 +attn_weights_193 +  + + +� +�*& +name + +" +attn_weights_193� +const& +_inversed_attn_weights_195_y_0 + *4 +name, +$ +"" +_inversed_attn_weights_195_y_0* +val +  + + + +>� +mul +x + +attn_weights_193' +y" + +_inversed_attn_weights_195_y_0> +_inversed_attn_weights_195 +  + + +� +�*0 +name( + +" +_inversed_attn_weights_195� +select) +cond! + +causal_mask_1_after_broadcast# +a + +_inversed_attn_weights_195% +b + +mask_value_1_after_broadcast4 +attn_weights_199 +  + + +� +�*& +name + +" +attn_weights_199� +add +x + +attn_weights_199 +y + +attention_mask_1- + input_229 +  + + +� +�* +name + + " + input_229� +softmax +x + + input_229 +axis + +var_64 +attn_weights_201 +  + + +� +�*& +name + +" +attn_weights_201q +const +tensor_269_transpose_x_0 +*. +name& + +" +tensor_269_transpose_x_0* +val + + +q +const +tensor_269_transpose_y_0 +*. +name& + +" +tensor_269_transpose_y_0* +val + + +� + transpose +x + + +tensor_267 +perm + + +var_22480 + transpose_151 +  + + +� +@*# +name + +" + transpose_151� +matmul +x + +attn_weights_201 +y + + transpose_151+ + transpose_x + +tensor_269_transpose_x_0+ + transpose_y + +tensor_269_transpose_y_0- + +tensor_269 +  + + +� +@* +name + +" + +tensor_269c +const +var_2279 + + +* +name + + " +op_2279*! +val + + + + + +k +const +var_2291 + + +* +name + + " +op_2291*) +val" + + + + + ����������� + transpose +x + + +tensor_269 +perm + + +var_22790 + transpose_148 +  + +� + +@*# +name + +" + transpose_148y +reshape +x + + transpose_148 +shape + + +var_2291 +var_2292 +  +� +�* +name + + " +op_2292� +constL +4model_transformer_h_16_attn_c_proj_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_16_attn_c_proj_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_2292B +weight8 +6 +4model_transformer_h_16_attn_c_proj_weight_transposed3 +bias+ +) +'model_transformer_h_16_attn_c_proj_bias +x_263 +  +� +�* +name + + " +x_263d +const +var_2294 + + +* +name + + " +op_2294*" +val + + + +  +��z +reshape +x + +x_263 +shape + + +var_2294' + input_233 +  + +� +�* +name + + " + input_233w +add +x + + input_233 +y + + input_227' + input_235 +  + +� +�* +name + + " + input_235r +const + x_265_axes_0 + + +*" +name + +" + x_265_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_235 +axes + + x_265_axes_0/ +gamma& +$ +"model_transformer_h_16_ln_2_weight, +beta$ +" + model_transformer_h_16_ln_2_bias +epsilon + + +var_26# +x_265 +  + +� +�* +name + + " +x_265k +const +var_2309 + + +* +name + + " +op_2309*) +val" + + + + + ����������q +reshape +x + +x_265 +shape + + +var_2309 +var_2310 +  +� +�* +name + + " +op_2310� +constI +1model_transformer_h_16_mlp_c_fc_weight_transposed +  +�  +�*G +name? +7 +5"3 +1model_transformer_h_16_mlp_c_fc_weight_transposed*E +val> +  +�  +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_2310? +weight5 +3 +1model_transformer_h_16_mlp_c_fc_weight_transposed0 +bias( +& +$model_transformer_h_16_mlp_c_fc_bias +x_267 +  +� +� * +name + + " +x_267d +const +var_2312 + + +* +name + + " +op_2312*" +val + + + +  +�� z +reshape +x + +x_267 +shape + + +var_2312' + input_237 +  + +� +� * +name + + " + input_237j +const + x_269_mode_0 +*" +name + +" + x_269_mode_0*' +val  + +" +TANH_APPROXIMATIONv +gelu +x + + input_237 +mode + + x_269_mode_0# +x_269 +  + +� +� * +name + + " +x_269k +const +var_2331 + + +* +name + + " +op_2331*) +val" + + + + + ���������� q +reshape +x + +x_269 +shape + + +var_2331 +var_2332 +  +� +� * +name + + " +op_2332� +constK +3model_transformer_h_16_mlp_c_proj_weight_transposed +  +� +� *I +nameA +9 +7"5 +3model_transformer_h_16_mlp_c_proj_weight_transposed*E +val> +  +� +� *& +@model_path/weights/weight.bin����� +linear +x + + +var_2332A +weight7 +5 +3model_transformer_h_16_mlp_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_16_mlp_c_proj_bias +x_271 +  +� +�* +name + + " +x_271d +const +var_2334 + + +* +name + + " +op_2334*" +val + + + +  +��z +reshape +x + +x_271 +shape + + +var_2334' + input_239 +  + +� +�* +name + + " + input_239w +add +x + + input_235 +y + + input_239' + input_241 +  + +� +�* +name + + " + input_241r +const + x_273_axes_0 + + +*" +name + +" + x_273_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_241 +axes + + x_273_axes_0/ +gamma& +$ +"model_transformer_h_17_ln_1_weight, +beta$ +" + model_transformer_h_17_ln_1_bias +epsilon + + +var_26# +x_273 +  + +� +�* +name + + " +x_273k +const +var_2354 + + +* +name + + " +op_2354*) +val" + + + + + ����������q +reshape +x + +x_273 +shape + + +var_2354 +var_2355 +  +� +�* +name + + " +op_2355� +constL +4model_transformer_h_17_attn_c_attn_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_17_attn_c_attn_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_2355B +weight8 +6 +4model_transformer_h_17_attn_c_attn_weight_transposed3 +bias+ +) +'model_transformer_h_17_attn_c_attn_bias +x_275 +  +� +�* +name + + " +x_275d +const +var_2357 + + +* +name + + " +op_2357*" +val + + + +  +��w +reshape +x + +x_275 +shape + + +var_2357& +var_2358 +  + +� +�* +name + + " +op_2358d +const +tile_17 + + +* +name + + " +tile_17*# +val + + + + + +���^ +const +var_2359_axis_0 +*$ +name + +" +op_2359_axis_0* +val + + +� +split +x + + +var_2358 + split_sizes + +tile_17 +axis + +var_2359_axis_0( + +var_2359_0 +  + +� +�( + +var_2359_1 +  + +� +�( + +var_2359_2 +  + +� +�* +name + + " +op_2359d +const +var_2365 + + +* +name + + " +op_2365*" +val + + + +  +�@� +reshape +x + + +var_2359_0 +shape + + +var_2365- + +tensor_275 +  + +� + +@* +name + +" + +tensor_275d +const +var_2371 + + +* +name + + " +op_2371*" +val + + + +  +�@� +reshape +x + + +var_2359_1 +shape + + +var_2371- + +tensor_279 +  + +� + +@* +name + +" + +tensor_279d +const +var_2377 + + +* +name + + " +op_2377*" +val + + + +  +�@� +reshape +x + + +var_2359_2 +shape + + +var_2377- + +tensor_283 +  + +� + +@* +name + +" + +tensor_283c +const +var_2379 + + +* +name + + " +op_2379*! +val + + + + + +} +const& +attn_weights_205_transpose_x_0 +*4 +name, +$ +"" +attn_weights_205_transpose_x_0* +val + + +} +const& +attn_weights_205_transpose_y_0 +*4 +name, +$ +"" +attn_weights_205_transpose_y_0* +val + + +| +const$ +transpose_106_perm_0 + + +** +name" + +" +transpose_106_perm_0*! +val + + + + + +| +const$ +transpose_107_perm_0 + + +** +name" + +" +transpose_107_perm_0*! +val + + + + + +� + transpose +x + + +tensor_279 +perm + +transpose_107_perm_00 + transpose_145 +  + + +@ +�*# +name + +" + transpose_145� + transpose +x + + +tensor_275 +perm + +transpose_106_perm_00 + transpose_146 +  + + +� +@*# +name + +" + transpose_146� +matmul +x + + transpose_146 +y + + transpose_1451 + transpose_x" + +attn_weights_205_transpose_x_01 + transpose_y" + +attn_weights_205_transpose_y_04 +attn_weights_205 +  + + +� +�*& +name + +" +attn_weights_205� +const& +_inversed_attn_weights_207_y_0 + *4 +name, +$ +"" +_inversed_attn_weights_207_y_0* +val +  + + + +>� +mul +x + +attn_weights_205' +y" + +_inversed_attn_weights_207_y_0> +_inversed_attn_weights_207 +  + + +� +�*0 +name( + +" +_inversed_attn_weights_207� +select) +cond! + +causal_mask_1_after_broadcast# +a + +_inversed_attn_weights_207% +b + +mask_value_1_after_broadcast4 +attn_weights_211 +  + + +� +�*& +name + +" +attn_weights_211� +add +x + +attn_weights_211 +y + +attention_mask_1- + input_243 +  + + +� +�* +name + + " + input_243� +softmax +x + + input_243 +axis + +var_64 +attn_weights_213 +  + + +� +�*& +name + +" +attn_weights_213q +const +tensor_285_transpose_x_0 +*. +name& + +" +tensor_285_transpose_x_0* +val + + +q +const +tensor_285_transpose_y_0 +*. +name& + +" +tensor_285_transpose_y_0* +val + + +� + transpose +x + + +tensor_283 +perm + + +var_23790 + transpose_147 +  + + +� +@*# +name + +" + transpose_147� +matmul +x + +attn_weights_213 +y + + transpose_147+ + transpose_x + +tensor_285_transpose_x_0+ + transpose_y + +tensor_285_transpose_y_0- + +tensor_285 +  + + +� +@* +name + +" + +tensor_285c +const +var_2410 + + +* +name + + " +op_2410*! +val + + + + + +k +const +var_2422 + + +* +name + + " +op_2422*) +val" + + + + + ����������� + transpose +x + + +tensor_285 +perm + + +var_24100 + transpose_144 +  + +� + +@*# +name + +" + transpose_144y +reshape +x + + transpose_144 +shape + + +var_2422 +var_2423 +  +� +�* +name + + " +op_2423� +constL +4model_transformer_h_17_attn_c_proj_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_17_attn_c_proj_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin��Á� +linear +x + + +var_2423B +weight8 +6 +4model_transformer_h_17_attn_c_proj_weight_transposed3 +bias+ +) +'model_transformer_h_17_attn_c_proj_bias +x_279 +  +� +�* +name + + " +x_279d +const +var_2425 + + +* +name + + " +op_2425*" +val + + + +  +��z +reshape +x + +x_279 +shape + + +var_2425' + input_247 +  + +� +�* +name + + " + input_247w +add +x + + input_247 +y + + input_241' + input_249 +  + +� +�* +name + + " + input_249r +const + x_281_axes_0 + + +*" +name + +" + x_281_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_249 +axes + + x_281_axes_0/ +gamma& +$ +"model_transformer_h_17_ln_2_weight, +beta$ +" + model_transformer_h_17_ln_2_bias +epsilon + + +var_26# +x_281 +  + +� +�* +name + + " +x_281k +const +var_2440 + + +* +name + + " +op_2440*) +val" + + + + + ����������q +reshape +x + +x_281 +shape + + +var_2440 +var_2441 +  +� +�* +name + + " +op_2441� +constI +1model_transformer_h_17_mlp_c_fc_weight_transposed +  +�  +�*G +name? +7 +5"3 +1model_transformer_h_17_mlp_c_fc_weight_transposed*E +val> +  +�  +�*& +@model_path/weights/weight.bin��Ã� +linear +x + + +var_2441? +weight5 +3 +1model_transformer_h_17_mlp_c_fc_weight_transposed0 +bias( +& +$model_transformer_h_17_mlp_c_fc_bias +x_283 +  +� +� * +name + + " +x_283d +const +var_2443 + + +* +name + + " +op_2443*" +val + + + +  +�� z +reshape +x + +x_283 +shape + + +var_2443' + input_251 +  + +� +� * +name + + " + input_251j +const + x_285_mode_0 +*" +name + +" + x_285_mode_0*' +val  + +" +TANH_APPROXIMATIONv +gelu +x + + input_251 +mode + + x_285_mode_0# +x_285 +  + +� +� * +name + + " +x_285k +const +var_2462 + + +* +name + + " +op_2462*) +val" + + + + + ���������� q +reshape +x + +x_285 +shape + + +var_2462 +var_2463 +  +� +� * +name + + " +op_2463� +constK +3model_transformer_h_17_mlp_c_proj_weight_transposed +  +� +� *I +nameA +9 +7"5 +3model_transformer_h_17_mlp_c_proj_weight_transposed*E +val> +  +� +� *& +@model_path/weights/weight.bin��Ë� +linear +x + + +var_2463A +weight7 +5 +3model_transformer_h_17_mlp_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_17_mlp_c_proj_bias +x_287 +  +� +�* +name + + " +x_287d +const +var_2465 + + +* +name + + " +op_2465*" +val + + + +  +��z +reshape +x + +x_287 +shape + + +var_2465' + input_253 +  + +� +�* +name + + " + input_253w +add +x + + input_249 +y + + input_253' + input_255 +  + +� +�* +name + + " + input_255r +const + x_289_axes_0 + + +*" +name + +" + x_289_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_255 +axes + + x_289_axes_0/ +gamma& +$ +"model_transformer_h_18_ln_1_weight, +beta$ +" + model_transformer_h_18_ln_1_bias +epsilon + + +var_26# +x_289 +  + +� +�* +name + + " +x_289k +const +var_2485 + + +* +name + + " +op_2485*) +val" + + + + + ����������q +reshape +x + +x_289 +shape + + +var_2485 +var_2486 +  +� +�* +name + + " +op_2486� +constL +4model_transformer_h_18_attn_c_attn_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_18_attn_c_attn_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin��Ó� +linear +x + + +var_2486B +weight8 +6 +4model_transformer_h_18_attn_c_attn_weight_transposed3 +bias+ +) +'model_transformer_h_18_attn_c_attn_bias +x_291 +  +� +�* +name + + " +x_291d +const +var_2488 + + +* +name + + " +op_2488*" +val + + + +  +��w +reshape +x + +x_291 +shape + + +var_2488& +var_2489 +  + +� +�* +name + + " +op_2489d +const +tile_18 + + +* +name + + " +tile_18*# +val + + + + + +���^ +const +var_2490_axis_0 +*$ +name + +" +op_2490_axis_0* +val + + +� +split +x + + +var_2489 + split_sizes + +tile_18 +axis + +var_2490_axis_0( + +var_2490_0 +  + +� +�( + +var_2490_1 +  + +� +�( + +var_2490_2 +  + +� +�* +name + + " +op_2490d +const +var_2496 + + +* +name + + " +op_2496*" +val + + + +  +�@� +reshape +x + + +var_2490_0 +shape + + +var_2496- + +tensor_291 +  + +� + +@* +name + +" + +tensor_291d +const +var_2502 + + +* +name + + " +op_2502*" +val + + + +  +�@� +reshape +x + + +var_2490_1 +shape + + +var_2502- + +tensor_295 +  + +� + +@* +name + +" + +tensor_295d +const +var_2508 + + +* +name + + " +op_2508*" +val + + + +  +�@� +reshape +x + + +var_2490_2 +shape + + +var_2508- + +tensor_299 +  + +� + +@* +name + +" + +tensor_299c +const +var_2510 + + +* +name + + " +op_2510*! +val + + + + + +} +const& +attn_weights_217_transpose_x_0 +*4 +name, +$ +"" +attn_weights_217_transpose_x_0* +val + + +} +const& +attn_weights_217_transpose_y_0 +*4 +name, +$ +"" +attn_weights_217_transpose_y_0* +val + + +| +const$ +transpose_108_perm_0 + + +** +name" + +" +transpose_108_perm_0*! +val + + + + + +| +const$ +transpose_109_perm_0 + + +** +name" + +" +transpose_109_perm_0*! +val + + + + + +� + transpose +x + + +tensor_295 +perm + +transpose_109_perm_00 + transpose_141 +  + + +@ +�*# +name + +" + transpose_141� + transpose +x + + +tensor_291 +perm + +transpose_108_perm_00 + transpose_142 +  + + +� +@*# +name + +" + transpose_142� +matmul +x + + transpose_142 +y + + transpose_1411 + transpose_x" + +attn_weights_217_transpose_x_01 + transpose_y" + +attn_weights_217_transpose_y_04 +attn_weights_217 +  + + +� +�*& +name + +" +attn_weights_217� +const& +_inversed_attn_weights_219_y_0 + *4 +name, +$ +"" +_inversed_attn_weights_219_y_0* +val +  + + + +>� +mul +x + +attn_weights_217' +y" + +_inversed_attn_weights_219_y_0> +_inversed_attn_weights_219 +  + + +� +�*0 +name( + +" +_inversed_attn_weights_219� +select) +cond! + +causal_mask_1_after_broadcast# +a + +_inversed_attn_weights_219% +b + +mask_value_1_after_broadcast4 +attn_weights_223 +  + + +� +�*& +name + +" +attn_weights_223� +add +x + +attn_weights_223 +y + +attention_mask_1- + input_257 +  + + +� +�* +name + + " + input_257� +softmax +x + + input_257 +axis + +var_64 +attn_weights_225 +  + + +� +�*& +name + +" +attn_weights_225q +const +tensor_301_transpose_x_0 +*. +name& + +" +tensor_301_transpose_x_0* +val + + +q +const +tensor_301_transpose_y_0 +*. +name& + +" +tensor_301_transpose_y_0* +val + + +� + transpose +x + + +tensor_299 +perm + + +var_25100 + transpose_143 +  + + +� +@*# +name + +" + transpose_143� +matmul +x + +attn_weights_225 +y + + transpose_143+ + transpose_x + +tensor_301_transpose_x_0+ + transpose_y + +tensor_301_transpose_y_0- + +tensor_301 +  + + +� +@* +name + +" + +tensor_301c +const +var_2541 + + +* +name + + " +op_2541*! +val + + + + + +k +const +var_2553 + + +* +name + + " +op_2553*) +val" + + + + + ����������� + transpose +x + + +tensor_301 +perm + + +var_25410 + transpose_140 +  + +� + +@*# +name + +" + transpose_140y +reshape +x + + transpose_140 +shape + + +var_2553 +var_2554 +  +� +�* +name + + " +op_2554� +constL +4model_transformer_h_18_attn_c_proj_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_18_attn_c_proj_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin��Ù� +linear +x + + +var_2554B +weight8 +6 +4model_transformer_h_18_attn_c_proj_weight_transposed3 +bias+ +) +'model_transformer_h_18_attn_c_proj_bias +x_295 +  +� +�* +name + + " +x_295d +const +var_2556 + + +* +name + + " +op_2556*" +val + + + +  +��z +reshape +x + +x_295 +shape + + +var_2556' + input_261 +  + +� +�* +name + + " + input_261w +add +x + + input_261 +y + + input_255' + input_263 +  + +� +�* +name + + " + input_263r +const + x_297_axes_0 + + +*" +name + +" + x_297_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_263 +axes + + x_297_axes_0/ +gamma& +$ +"model_transformer_h_18_ln_2_weight, +beta$ +" + model_transformer_h_18_ln_2_bias +epsilon + + +var_26# +x_297 +  + +� +�* +name + + " +x_297k +const +var_2571 + + +* +name + + " +op_2571*) +val" + + + + + ����������q +reshape +x + +x_297 +shape + + +var_2571 +var_2572 +  +� +�* +name + + " +op_2572� +constI +1model_transformer_h_18_mlp_c_fc_weight_transposed +  +�  +�*G +name? +7 +5"3 +1model_transformer_h_18_mlp_c_fc_weight_transposed*E +val> +  +�  +�*& +@model_path/weights/weight.bin��Û� +linear +x + + +var_2572? +weight5 +3 +1model_transformer_h_18_mlp_c_fc_weight_transposed0 +bias( +& +$model_transformer_h_18_mlp_c_fc_bias +x_299 +  +� +� * +name + + " +x_299d +const +var_2574 + + +* +name + + " +op_2574*" +val + + + +  +�� z +reshape +x + +x_299 +shape + + +var_2574' + input_265 +  + +� +� * +name + + " + input_265j +const + x_301_mode_0 +*" +name + +" + x_301_mode_0*' +val  + +" +TANH_APPROXIMATIONv +gelu +x + + input_265 +mode + + x_301_mode_0# +x_301 +  + +� +� * +name + + " +x_301k +const +var_2593 + + +* +name + + " +op_2593*) +val" + + + + + ���������� q +reshape +x + +x_301 +shape + + +var_2593 +var_2594 +  +� +� * +name + + " +op_2594� +constK +3model_transformer_h_18_mlp_c_proj_weight_transposed +  +� +� *I +nameA +9 +7"5 +3model_transformer_h_18_mlp_c_proj_weight_transposed*E +val> +  +� +� *& +@model_path/weights/weight.bin��ã� +linear +x + + +var_2594A +weight7 +5 +3model_transformer_h_18_mlp_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_18_mlp_c_proj_bias +x_303 +  +� +�* +name + + " +x_303d +const +var_2596 + + +* +name + + " +op_2596*" +val + + + +  +��z +reshape +x + +x_303 +shape + + +var_2596' + input_267 +  + +� +�* +name + + " + input_267w +add +x + + input_263 +y + + input_267' + input_269 +  + +� +�* +name + + " + input_269r +const + x_305_axes_0 + + +*" +name + +" + x_305_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_269 +axes + + x_305_axes_0/ +gamma& +$ +"model_transformer_h_19_ln_1_weight, +beta$ +" + model_transformer_h_19_ln_1_bias +epsilon + + +var_26# +x_305 +  + +� +�* +name + + " +x_305k +const +var_2616 + + +* +name + + " +op_2616*) +val" + + + + + ����������q +reshape +x + +x_305 +shape + + +var_2616 +var_2617 +  +� +�* +name + + " +op_2617� +constL +4model_transformer_h_19_attn_c_attn_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_19_attn_c_attn_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin��ë� +linear +x + + +var_2617B +weight8 +6 +4model_transformer_h_19_attn_c_attn_weight_transposed3 +bias+ +) +'model_transformer_h_19_attn_c_attn_bias +x_307 +  +� +�* +name + + " +x_307d +const +var_2619 + + +* +name + + " +op_2619*" +val + + + +  +��w +reshape +x + +x_307 +shape + + +var_2619& +var_2620 +  + +� +�* +name + + " +op_2620d +const +tile_19 + + +* +name + + " +tile_19*# +val + + + + + +���^ +const +var_2621_axis_0 +*$ +name + +" +op_2621_axis_0* +val + + +� +split +x + + +var_2620 + split_sizes + +tile_19 +axis + +var_2621_axis_0( + +var_2621_0 +  + +� +�( + +var_2621_1 +  + +� +�( + +var_2621_2 +  + +� +�* +name + + " +op_2621d +const +var_2627 + + +* +name + + " +op_2627*" +val + + + +  +�@� +reshape +x + + +var_2621_0 +shape + + +var_2627- + +tensor_307 +  + +� + +@* +name + +" + +tensor_307d +const +var_2633 + + +* +name + + " +op_2633*" +val + + + +  +�@� +reshape +x + + +var_2621_1 +shape + + +var_2633- + +tensor_311 +  + +� + +@* +name + +" + +tensor_311d +const +var_2639 + + +* +name + + " +op_2639*" +val + + + +  +�@� +reshape +x + + +var_2621_2 +shape + + +var_2639- + +tensor_315 +  + +� + +@* +name + +" + +tensor_315c +const +var_2641 + + +* +name + + " +op_2641*! +val + + + + + +} +const& +attn_weights_229_transpose_x_0 +*4 +name, +$ +"" +attn_weights_229_transpose_x_0* +val + + +} +const& +attn_weights_229_transpose_y_0 +*4 +name, +$ +"" +attn_weights_229_transpose_y_0* +val + + +| +const$ +transpose_110_perm_0 + + +** +name" + +" +transpose_110_perm_0*! +val + + + + + +| +const$ +transpose_111_perm_0 + + +** +name" + +" +transpose_111_perm_0*! +val + + + + + +� + transpose +x + + +tensor_311 +perm + +transpose_111_perm_00 + transpose_137 +  + + +@ +�*# +name + +" + transpose_137� + transpose +x + + +tensor_307 +perm + +transpose_110_perm_00 + transpose_138 +  + + +� +@*# +name + +" + transpose_138� +matmul +x + + transpose_138 +y + + transpose_1371 + transpose_x" + +attn_weights_229_transpose_x_01 + transpose_y" + +attn_weights_229_transpose_y_04 +attn_weights_229 +  + + +� +�*& +name + +" +attn_weights_229� +const& +_inversed_attn_weights_231_y_0 + *4 +name, +$ +"" +_inversed_attn_weights_231_y_0* +val +  + + + +>� +mul +x + +attn_weights_229' +y" + +_inversed_attn_weights_231_y_0> +_inversed_attn_weights_231 +  + + +� +�*0 +name( + +" +_inversed_attn_weights_231� +select) +cond! + +causal_mask_1_after_broadcast# +a + +_inversed_attn_weights_231% +b + +mask_value_1_after_broadcast4 +attn_weights_235 +  + + +� +�*& +name + +" +attn_weights_235� +add +x + +attn_weights_235 +y + +attention_mask_1- + input_271 +  + + +� +�* +name + + " + input_271� +softmax +x + + input_271 +axis + +var_64 +attn_weights_237 +  + + +� +�*& +name + +" +attn_weights_237q +const +tensor_317_transpose_x_0 +*. +name& + +" +tensor_317_transpose_x_0* +val + + +q +const +tensor_317_transpose_y_0 +*. +name& + +" +tensor_317_transpose_y_0* +val + + +� + transpose +x + + +tensor_315 +perm + + +var_26410 + transpose_139 +  + + +� +@*# +name + +" + transpose_139� +matmul +x + +attn_weights_237 +y + + transpose_139+ + transpose_x + +tensor_317_transpose_x_0+ + transpose_y + +tensor_317_transpose_y_0- + +tensor_317 +  + + +� +@* +name + +" + +tensor_317c +const +var_2672 + + +* +name + + " +op_2672*! +val + + + + + +k +const +var_2684 + + +* +name + + " +op_2684*) +val" + + + + + ����������� + transpose +x + + +tensor_317 +perm + + +var_26720 + transpose_136 +  + +� + +@*# +name + +" + transpose_136y +reshape +x + + transpose_136 +shape + + +var_2684 +var_2685 +  +� +�* +name + + " +op_2685� +constL +4model_transformer_h_19_attn_c_proj_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_19_attn_c_proj_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin��ñ� +linear +x + + +var_2685B +weight8 +6 +4model_transformer_h_19_attn_c_proj_weight_transposed3 +bias+ +) +'model_transformer_h_19_attn_c_proj_bias +x_311 +  +� +�* +name + + " +x_311d +const +var_2687 + + +* +name + + " +op_2687*" +val + + + +  +��z +reshape +x + +x_311 +shape + + +var_2687' + input_275 +  + +� +�* +name + + " + input_275w +add +x + + input_275 +y + + input_269' + input_277 +  + +� +�* +name + + " + input_277r +const + x_313_axes_0 + + +*" +name + +" + x_313_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_277 +axes + + x_313_axes_0/ +gamma& +$ +"model_transformer_h_19_ln_2_weight, +beta$ +" + model_transformer_h_19_ln_2_bias +epsilon + + +var_26# +x_313 +  + +� +�* +name + + " +x_313k +const +var_2702 + + +* +name + + " +op_2702*) +val" + + + + + ����������q +reshape +x + +x_313 +shape + + +var_2702 +var_2703 +  +� +�* +name + + " +op_2703� +constI +1model_transformer_h_19_mlp_c_fc_weight_transposed +  +�  +�*G +name? +7 +5"3 +1model_transformer_h_19_mlp_c_fc_weight_transposed*E +val> +  +�  +�*& +@model_path/weights/weight.bin��ó� +linear +x + + +var_2703? +weight5 +3 +1model_transformer_h_19_mlp_c_fc_weight_transposed0 +bias( +& +$model_transformer_h_19_mlp_c_fc_bias +x_315 +  +� +� * +name + + " +x_315d +const +var_2705 + + +* +name + + " +op_2705*" +val + + + +  +�� z +reshape +x + +x_315 +shape + + +var_2705' + input_279 +  + +� +� * +name + + " + input_279j +const + x_317_mode_0 +*" +name + +" + x_317_mode_0*' +val  + +" +TANH_APPROXIMATIONv +gelu +x + + input_279 +mode + + x_317_mode_0# +x_317 +  + +� +� * +name + + " +x_317k +const +var_2724 + + +* +name + + " +op_2724*) +val" + + + + + ���������� q +reshape +x + +x_317 +shape + + +var_2724 +var_2725 +  +� +� * +name + + " +op_2725� +constK +3model_transformer_h_19_mlp_c_proj_weight_transposed +  +� +� *I +nameA +9 +7"5 +3model_transformer_h_19_mlp_c_proj_weight_transposed*E +val> +  +� +� *& +@model_path/weights/weight.bin��û� +linear +x + + +var_2725A +weight7 +5 +3model_transformer_h_19_mlp_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_19_mlp_c_proj_bias +x_319 +  +� +�* +name + + " +x_319d +const +var_2727 + + +* +name + + " +op_2727*" +val + + + +  +��z +reshape +x + +x_319 +shape + + +var_2727' + input_281 +  + +� +�* +name + + " + input_281w +add +x + + input_277 +y + + input_281' + input_283 +  + +� +�* +name + + " + input_283r +const + x_321_axes_0 + + +*" +name + +" + x_321_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_283 +axes + + x_321_axes_0/ +gamma& +$ +"model_transformer_h_20_ln_1_weight, +beta$ +" + model_transformer_h_20_ln_1_bias +epsilon + + +var_26# +x_321 +  + +� +�* +name + + " +x_321k +const +var_2747 + + +* +name + + " +op_2747*) +val" + + + + + ����������q +reshape +x + +x_321 +shape + + +var_2747 +var_2748 +  +� +�* +name + + " +op_2748� +constL +4model_transformer_h_20_attn_c_attn_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_20_attn_c_attn_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_2748B +weight8 +6 +4model_transformer_h_20_attn_c_attn_weight_transposed3 +bias+ +) +'model_transformer_h_20_attn_c_attn_bias +x_323 +  +� +�* +name + + " +x_323d +const +var_2750 + + +* +name + + " +op_2750*" +val + + + +  +��w +reshape +x + +x_323 +shape + + +var_2750& +var_2751 +  + +� +�* +name + + " +op_2751d +const +tile_20 + + +* +name + + " +tile_20*# +val + + + + + +���^ +const +var_2752_axis_0 +*$ +name + +" +op_2752_axis_0* +val + + +� +split +x + + +var_2751 + split_sizes + +tile_20 +axis + +var_2752_axis_0( + +var_2752_0 +  + +� +�( + +var_2752_1 +  + +� +�( + +var_2752_2 +  + +� +�* +name + + " +op_2752d +const +var_2758 + + +* +name + + " +op_2758*" +val + + + +  +�@� +reshape +x + + +var_2752_0 +shape + + +var_2758- + +tensor_323 +  + +� + +@* +name + +" + +tensor_323d +const +var_2764 + + +* +name + + " +op_2764*" +val + + + +  +�@� +reshape +x + + +var_2752_1 +shape + + +var_2764- + +tensor_327 +  + +� + +@* +name + +" + +tensor_327d +const +var_2770 + + +* +name + + " +op_2770*" +val + + + +  +�@� +reshape +x + + +var_2752_2 +shape + + +var_2770- + +tensor_331 +  + +� + +@* +name + +" + +tensor_331c +const +var_2772 + + +* +name + + " +op_2772*! +val + + + + + +} +const& +attn_weights_241_transpose_x_0 +*4 +name, +$ +"" +attn_weights_241_transpose_x_0* +val + + +} +const& +attn_weights_241_transpose_y_0 +*4 +name, +$ +"" +attn_weights_241_transpose_y_0* +val + + +| +const$ +transpose_112_perm_0 + + +** +name" + +" +transpose_112_perm_0*! +val + + + + + +| +const$ +transpose_113_perm_0 + + +** +name" + +" +transpose_113_perm_0*! +val + + + + + +� + transpose +x + + +tensor_327 +perm + +transpose_113_perm_00 + transpose_133 +  + + +@ +�*# +name + +" + transpose_133� + transpose +x + + +tensor_323 +perm + +transpose_112_perm_00 + transpose_134 +  + + +� +@*# +name + +" + transpose_134� +matmul +x + + transpose_134 +y + + transpose_1331 + transpose_x" + +attn_weights_241_transpose_x_01 + transpose_y" + +attn_weights_241_transpose_y_04 +attn_weights_241 +  + + +� +�*& +name + +" +attn_weights_241� +const& +_inversed_attn_weights_243_y_0 + *4 +name, +$ +"" +_inversed_attn_weights_243_y_0* +val +  + + + +>� +mul +x + +attn_weights_241' +y" + +_inversed_attn_weights_243_y_0> +_inversed_attn_weights_243 +  + + +� +�*0 +name( + +" +_inversed_attn_weights_243� +select) +cond! + +causal_mask_1_after_broadcast# +a + +_inversed_attn_weights_243% +b + +mask_value_1_after_broadcast4 +attn_weights_247 +  + + +� +�*& +name + +" +attn_weights_247� +add +x + +attn_weights_247 +y + +attention_mask_1- + input_285 +  + + +� +�* +name + + " + input_285� +softmax +x + + input_285 +axis + +var_64 +attn_weights_249 +  + + +� +�*& +name + +" +attn_weights_249q +const +tensor_333_transpose_x_0 +*. +name& + +" +tensor_333_transpose_x_0* +val + + +q +const +tensor_333_transpose_y_0 +*. +name& + +" +tensor_333_transpose_y_0* +val + + +� + transpose +x + + +tensor_331 +perm + + +var_27720 + transpose_135 +  + + +� +@*# +name + +" + transpose_135� +matmul +x + +attn_weights_249 +y + + transpose_135+ + transpose_x + +tensor_333_transpose_x_0+ + transpose_y + +tensor_333_transpose_y_0- + +tensor_333 +  + + +� +@* +name + +" + +tensor_333c +const +var_2803 + + +* +name + + " +op_2803*! +val + + + + + +k +const +var_2815 + + +* +name + + " +op_2815*) +val" + + + + + ����������� + transpose +x + + +tensor_333 +perm + + +var_28030 + transpose_132 +  + +� + +@*# +name + +" + transpose_132y +reshape +x + + transpose_132 +shape + + +var_2815 +var_2816 +  +� +�* +name + + " +op_2816� +constL +4model_transformer_h_20_attn_c_proj_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_20_attn_c_proj_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_2816B +weight8 +6 +4model_transformer_h_20_attn_c_proj_weight_transposed3 +bias+ +) +'model_transformer_h_20_attn_c_proj_bias +x_327 +  +� +�* +name + + " +x_327d +const +var_2818 + + +* +name + + " +op_2818*" +val + + + +  +��z +reshape +x + +x_327 +shape + + +var_2818' + input_289 +  + +� +�* +name + + " + input_289w +add +x + + input_289 +y + + input_283' + input_291 +  + +� +�* +name + + " + input_291r +const + x_329_axes_0 + + +*" +name + +" + x_329_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_291 +axes + + x_329_axes_0/ +gamma& +$ +"model_transformer_h_20_ln_2_weight, +beta$ +" + model_transformer_h_20_ln_2_bias +epsilon + + +var_26# +x_329 +  + +� +�* +name + + " +x_329k +const +var_2833 + + +* +name + + " +op_2833*) +val" + + + + + ����������q +reshape +x + +x_329 +shape + + +var_2833 +var_2834 +  +� +�* +name + + " +op_2834� +constI +1model_transformer_h_20_mlp_c_fc_weight_transposed +  +�  +�*G +name? +7 +5"3 +1model_transformer_h_20_mlp_c_fc_weight_transposed*E +val> +  +�  +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_2834? +weight5 +3 +1model_transformer_h_20_mlp_c_fc_weight_transposed0 +bias( +& +$model_transformer_h_20_mlp_c_fc_bias +x_331 +  +� +� * +name + + " +x_331d +const +var_2836 + + +* +name + + " +op_2836*" +val + + + +  +�� z +reshape +x + +x_331 +shape + + +var_2836' + input_293 +  + +� +� * +name + + " + input_293j +const + x_333_mode_0 +*" +name + +" + x_333_mode_0*' +val  + +" +TANH_APPROXIMATIONv +gelu +x + + input_293 +mode + + x_333_mode_0# +x_333 +  + +� +� * +name + + " +x_333k +const +var_2855 + + +* +name + + " +op_2855*) +val" + + + + + ���������� q +reshape +x + +x_333 +shape + + +var_2855 +var_2856 +  +� +� * +name + + " +op_2856� +constK +3model_transformer_h_20_mlp_c_proj_weight_transposed +  +� +� *I +nameA +9 +7"5 +3model_transformer_h_20_mlp_c_proj_weight_transposed*E +val> +  +� +� *& +@model_path/weights/weight.bin����� +linear +x + + +var_2856A +weight7 +5 +3model_transformer_h_20_mlp_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_20_mlp_c_proj_bias +x_335 +  +� +�* +name + + " +x_335d +const +var_2858 + + +* +name + + " +op_2858*" +val + + + +  +��z +reshape +x + +x_335 +shape + + +var_2858' + input_295 +  + +� +�* +name + + " + input_295w +add +x + + input_291 +y + + input_295' + input_297 +  + +� +�* +name + + " + input_297r +const + x_337_axes_0 + + +*" +name + +" + x_337_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_297 +axes + + x_337_axes_0/ +gamma& +$ +"model_transformer_h_21_ln_1_weight, +beta$ +" + model_transformer_h_21_ln_1_bias +epsilon + + +var_26# +x_337 +  + +� +�* +name + + " +x_337k +const +var_2878 + + +* +name + + " +op_2878*) +val" + + + + + ����������q +reshape +x + +x_337 +shape + + +var_2878 +var_2879 +  +� +�* +name + + " +op_2879� +constL +4model_transformer_h_21_attn_c_attn_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_21_attn_c_attn_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_2879B +weight8 +6 +4model_transformer_h_21_attn_c_attn_weight_transposed3 +bias+ +) +'model_transformer_h_21_attn_c_attn_bias +x_339 +  +� +�* +name + + " +x_339d +const +var_2881 + + +* +name + + " +op_2881*" +val + + + +  +��w +reshape +x + +x_339 +shape + + +var_2881& +var_2882 +  + +� +�* +name + + " +op_2882d +const +tile_21 + + +* +name + + " +tile_21*# +val + + + + + +���^ +const +var_2883_axis_0 +*$ +name + +" +op_2883_axis_0* +val + + +� +split +x + + +var_2882 + split_sizes + +tile_21 +axis + +var_2883_axis_0( + +var_2883_0 +  + +� +�( + +var_2883_1 +  + +� +�( + +var_2883_2 +  + +� +�* +name + + " +op_2883d +const +var_2889 + + +* +name + + " +op_2889*" +val + + + +  +�@� +reshape +x + + +var_2883_0 +shape + + +var_2889- + +tensor_339 +  + +� + +@* +name + +" + +tensor_339d +const +var_2895 + + +* +name + + " +op_2895*" +val + + + +  +�@� +reshape +x + + +var_2883_1 +shape + + +var_2895- + +tensor_343 +  + +� + +@* +name + +" + +tensor_343d +const +var_2901 + + +* +name + + " +op_2901*" +val + + + +  +�@� +reshape +x + + +var_2883_2 +shape + + +var_2901- + +tensor_347 +  + +� + +@* +name + +" + +tensor_347c +const +var_2903 + + +* +name + + " +op_2903*! +val + + + + + +} +const& +attn_weights_253_transpose_x_0 +*4 +name, +$ +"" +attn_weights_253_transpose_x_0* +val + + +} +const& +attn_weights_253_transpose_y_0 +*4 +name, +$ +"" +attn_weights_253_transpose_y_0* +val + + +| +const$ +transpose_114_perm_0 + + +** +name" + +" +transpose_114_perm_0*! +val + + + + + +| +const$ +transpose_115_perm_0 + + +** +name" + +" +transpose_115_perm_0*! +val + + + + + +� + transpose +x + + +tensor_343 +perm + +transpose_115_perm_00 + transpose_129 +  + + +@ +�*# +name + +" + transpose_129� + transpose +x + + +tensor_339 +perm + +transpose_114_perm_00 + transpose_130 +  + + +� +@*# +name + +" + transpose_130� +matmul +x + + transpose_130 +y + + transpose_1291 + transpose_x" + +attn_weights_253_transpose_x_01 + transpose_y" + +attn_weights_253_transpose_y_04 +attn_weights_253 +  + + +� +�*& +name + +" +attn_weights_253� +const& +_inversed_attn_weights_255_y_0 + *4 +name, +$ +"" +_inversed_attn_weights_255_y_0* +val +  + + + +>� +mul +x + +attn_weights_253' +y" + +_inversed_attn_weights_255_y_0> +_inversed_attn_weights_255 +  + + +� +�*0 +name( + +" +_inversed_attn_weights_255� +select) +cond! + +causal_mask_1_after_broadcast# +a + +_inversed_attn_weights_255% +b + +mask_value_1_after_broadcast4 +attn_weights_259 +  + + +� +�*& +name + +" +attn_weights_259� +add +x + +attn_weights_259 +y + +attention_mask_1- + input_299 +  + + +� +�* +name + + " + input_299� +softmax +x + + input_299 +axis + +var_64 +attn_weights_261 +  + + +� +�*& +name + +" +attn_weights_261q +const +tensor_349_transpose_x_0 +*. +name& + +" +tensor_349_transpose_x_0* +val + + +q +const +tensor_349_transpose_y_0 +*. +name& + +" +tensor_349_transpose_y_0* +val + + +� + transpose +x + + +tensor_347 +perm + + +var_29030 + transpose_131 +  + + +� +@*# +name + +" + transpose_131� +matmul +x + +attn_weights_261 +y + + transpose_131+ + transpose_x + +tensor_349_transpose_x_0+ + transpose_y + +tensor_349_transpose_y_0- + +tensor_349 +  + + +� +@* +name + +" + +tensor_349c +const +var_2934 + + +* +name + + " +op_2934*! +val + + + + + +k +const +var_2946 + + +* +name + + " +op_2946*) +val" + + + + + ����������� + transpose +x + + +tensor_349 +perm + + +var_29340 + transpose_128 +  + +� + +@*# +name + +" + transpose_128y +reshape +x + + transpose_128 +shape + + +var_2946 +var_2947 +  +� +�* +name + + " +op_2947� +constL +4model_transformer_h_21_attn_c_proj_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_21_attn_c_proj_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_2947B +weight8 +6 +4model_transformer_h_21_attn_c_proj_weight_transposed3 +bias+ +) +'model_transformer_h_21_attn_c_proj_bias +x_343 +  +� +�* +name + + " +x_343d +const +var_2949 + + +* +name + + " +op_2949*" +val + + + +  +��z +reshape +x + +x_343 +shape + + +var_2949' + input_303 +  + +� +�* +name + + " + input_303w +add +x + + input_303 +y + + input_297' + input_305 +  + +� +�* +name + + " + input_305r +const + x_345_axes_0 + + +*" +name + +" + x_345_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_305 +axes + + x_345_axes_0/ +gamma& +$ +"model_transformer_h_21_ln_2_weight, +beta$ +" + model_transformer_h_21_ln_2_bias +epsilon + + +var_26# +x_345 +  + +� +�* +name + + " +x_345k +const +var_2964 + + +* +name + + " +op_2964*) +val" + + + + + ����������q +reshape +x + +x_345 +shape + + +var_2964 +var_2965 +  +� +�* +name + + " +op_2965� +constI +1model_transformer_h_21_mlp_c_fc_weight_transposed +  +�  +�*G +name? +7 +5"3 +1model_transformer_h_21_mlp_c_fc_weight_transposed*E +val> +  +�  +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_2965? +weight5 +3 +1model_transformer_h_21_mlp_c_fc_weight_transposed0 +bias( +& +$model_transformer_h_21_mlp_c_fc_bias +x_347 +  +� +� * +name + + " +x_347d +const +var_2967 + + +* +name + + " +op_2967*" +val + + + +  +�� z +reshape +x + +x_347 +shape + + +var_2967' + input_307 +  + +� +� * +name + + " + input_307j +const + x_349_mode_0 +*" +name + +" + x_349_mode_0*' +val  + +" +TANH_APPROXIMATIONv +gelu +x + + input_307 +mode + + x_349_mode_0# +x_349 +  + +� +� * +name + + " +x_349k +const +var_2986 + + +* +name + + " +op_2986*) +val" + + + + + ���������� q +reshape +x + +x_349 +shape + + +var_2986 +var_2987 +  +� +� * +name + + " +op_2987� +constK +3model_transformer_h_21_mlp_c_proj_weight_transposed +  +� +� *I +nameA +9 +7"5 +3model_transformer_h_21_mlp_c_proj_weight_transposed*E +val> +  +� +� *& +@model_path/weights/weight.bin����� +linear +x + + +var_2987A +weight7 +5 +3model_transformer_h_21_mlp_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_21_mlp_c_proj_bias +x_351 +  +� +�* +name + + " +x_351d +const +var_2989 + + +* +name + + " +op_2989*" +val + + + +  +��z +reshape +x + +x_351 +shape + + +var_2989' + input_309 +  + +� +�* +name + + " + input_309w +add +x + + input_305 +y + + input_309' + input_311 +  + +� +�* +name + + " + input_311r +const + x_353_axes_0 + + +*" +name + +" + x_353_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_311 +axes + + x_353_axes_0/ +gamma& +$ +"model_transformer_h_22_ln_1_weight, +beta$ +" + model_transformer_h_22_ln_1_bias +epsilon + + +var_26# +x_353 +  + +� +�* +name + + " +x_353k +const +var_3009 + + +* +name + + " +op_3009*) +val" + + + + + ����������q +reshape +x + +x_353 +shape + + +var_3009 +var_3010 +  +� +�* +name + + " +op_3010� +constL +4model_transformer_h_22_attn_c_attn_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_22_attn_c_attn_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_3010B +weight8 +6 +4model_transformer_h_22_attn_c_attn_weight_transposed3 +bias+ +) +'model_transformer_h_22_attn_c_attn_bias +x_355 +  +� +�* +name + + " +x_355d +const +var_3012 + + +* +name + + " +op_3012*" +val + + + +  +��w +reshape +x + +x_355 +shape + + +var_3012& +var_3013 +  + +� +�* +name + + " +op_3013d +const +tile_22 + + +* +name + + " +tile_22*# +val + + + + + +���^ +const +var_3014_axis_0 +*$ +name + +" +op_3014_axis_0* +val + + +� +split +x + + +var_3013 + split_sizes + +tile_22 +axis + +var_3014_axis_0( + +var_3014_0 +  + +� +�( + +var_3014_1 +  + +� +�( + +var_3014_2 +  + +� +�* +name + + " +op_3014d +const +var_3020 + + +* +name + + " +op_3020*" +val + + + +  +�@� +reshape +x + + +var_3014_0 +shape + + +var_3020- + +tensor_355 +  + +� + +@* +name + +" + +tensor_355d +const +var_3026 + + +* +name + + " +op_3026*" +val + + + +  +�@� +reshape +x + + +var_3014_1 +shape + + +var_3026- + +tensor_359 +  + +� + +@* +name + +" + +tensor_359d +const +var_3032 + + +* +name + + " +op_3032*" +val + + + +  +�@� +reshape +x + + +var_3014_2 +shape + + +var_3032- + +tensor_363 +  + +� + +@* +name + +" + +tensor_363c +const +var_3034 + + +* +name + + " +op_3034*! +val + + + + + +} +const& +attn_weights_265_transpose_x_0 +*4 +name, +$ +"" +attn_weights_265_transpose_x_0* +val + + +} +const& +attn_weights_265_transpose_y_0 +*4 +name, +$ +"" +attn_weights_265_transpose_y_0* +val + + +| +const$ +transpose_116_perm_0 + + +** +name" + +" +transpose_116_perm_0*! +val + + + + + +| +const$ +transpose_117_perm_0 + + +** +name" + +" +transpose_117_perm_0*! +val + + + + + +� + transpose +x + + +tensor_359 +perm + +transpose_117_perm_00 + transpose_125 +  + + +@ +�*# +name + +" + transpose_125� + transpose +x + + +tensor_355 +perm + +transpose_116_perm_00 + transpose_126 +  + + +� +@*# +name + +" + transpose_126� +matmul +x + + transpose_126 +y + + transpose_1251 + transpose_x" + +attn_weights_265_transpose_x_01 + transpose_y" + +attn_weights_265_transpose_y_04 +attn_weights_265 +  + + +� +�*& +name + +" +attn_weights_265� +const& +_inversed_attn_weights_267_y_0 + *4 +name, +$ +"" +_inversed_attn_weights_267_y_0* +val +  + + + +>� +mul +x + +attn_weights_265' +y" + +_inversed_attn_weights_267_y_0> +_inversed_attn_weights_267 +  + + +� +�*0 +name( + +" +_inversed_attn_weights_267� +select) +cond! + +causal_mask_1_after_broadcast# +a + +_inversed_attn_weights_267% +b + +mask_value_1_after_broadcast4 +attn_weights_271 +  + + +� +�*& +name + +" +attn_weights_271� +add +x + +attn_weights_271 +y + +attention_mask_1- + input_313 +  + + +� +�* +name + + " + input_313� +softmax +x + + input_313 +axis + +var_64 +attn_weights_273 +  + + +� +�*& +name + +" +attn_weights_273q +const +tensor_365_transpose_x_0 +*. +name& + +" +tensor_365_transpose_x_0* +val + + +q +const +tensor_365_transpose_y_0 +*. +name& + +" +tensor_365_transpose_y_0* +val + + +� + transpose +x + + +tensor_363 +perm + + +var_30340 + transpose_127 +  + + +� +@*# +name + +" + transpose_127� +matmul +x + +attn_weights_273 +y + + transpose_127+ + transpose_x + +tensor_365_transpose_x_0+ + transpose_y + +tensor_365_transpose_y_0- + +tensor_365 +  + + +� +@* +name + +" + +tensor_365c +const +var_3065 + + +* +name + + " +op_3065*! +val + + + + + +k +const +var_3077 + + +* +name + + " +op_3077*) +val" + + + + + ����������� + transpose +x + + +tensor_365 +perm + + +var_30650 + transpose_124 +  + +� + +@*# +name + +" + transpose_124y +reshape +x + + transpose_124 +shape + + +var_3077 +var_3078 +  +� +�* +name + + " +op_3078� +constL +4model_transformer_h_22_attn_c_proj_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_22_attn_c_proj_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_3078B +weight8 +6 +4model_transformer_h_22_attn_c_proj_weight_transposed3 +bias+ +) +'model_transformer_h_22_attn_c_proj_bias +x_359 +  +� +�* +name + + " +x_359d +const +var_3080 + + +* +name + + " +op_3080*" +val + + + +  +��z +reshape +x + +x_359 +shape + + +var_3080' + input_317 +  + +� +�* +name + + " + input_317w +add +x + + input_317 +y + + input_311' + input_319 +  + +� +�* +name + + " + input_319r +const + x_361_axes_0 + + +*" +name + +" + x_361_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_319 +axes + + x_361_axes_0/ +gamma& +$ +"model_transformer_h_22_ln_2_weight, +beta$ +" + model_transformer_h_22_ln_2_bias +epsilon + + +var_26# +x_361 +  + +� +�* +name + + " +x_361k +const +var_3095 + + +* +name + + " +op_3095*) +val" + + + + + ����������q +reshape +x + +x_361 +shape + + +var_3095 +var_3096 +  +� +�* +name + + " +op_3096� +constI +1model_transformer_h_22_mlp_c_fc_weight_transposed +  +�  +�*G +name? +7 +5"3 +1model_transformer_h_22_mlp_c_fc_weight_transposed*E +val> +  +�  +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_3096? +weight5 +3 +1model_transformer_h_22_mlp_c_fc_weight_transposed0 +bias( +& +$model_transformer_h_22_mlp_c_fc_bias +x_363 +  +� +� * +name + + " +x_363d +const +var_3098 + + +* +name + + " +op_3098*" +val + + + +  +�� z +reshape +x + +x_363 +shape + + +var_3098' + input_321 +  + +� +� * +name + + " + input_321j +const + x_365_mode_0 +*" +name + +" + x_365_mode_0*' +val  + +" +TANH_APPROXIMATIONv +gelu +x + + input_321 +mode + + x_365_mode_0# +x_365 +  + +� +� * +name + + " +x_365k +const +var_3117 + + +* +name + + " +op_3117*) +val" + + + + + ���������� q +reshape +x + +x_365 +shape + + +var_3117 +var_3118 +  +� +� * +name + + " +op_3118� +constK +3model_transformer_h_22_mlp_c_proj_weight_transposed +  +� +� *I +nameA +9 +7"5 +3model_transformer_h_22_mlp_c_proj_weight_transposed*E +val> +  +� +� *& +@model_path/weights/weight.bin��Ã� +linear +x + + +var_3118A +weight7 +5 +3model_transformer_h_22_mlp_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_22_mlp_c_proj_bias +x_367 +  +� +�* +name + + " +x_367d +const +var_3120 + + +* +name + + " +op_3120*" +val + + + +  +��z +reshape +x + +x_367 +shape + + +var_3120' + input_323 +  + +� +�* +name + + " + input_323w +add +x + + input_319 +y + + input_323' + input_325 +  + +� +�* +name + + " + input_325r +const + x_369_axes_0 + + +*" +name + +" + x_369_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_325 +axes + + x_369_axes_0/ +gamma& +$ +"model_transformer_h_23_ln_1_weight, +beta$ +" + model_transformer_h_23_ln_1_bias +epsilon + + +var_26# +x_369 +  + +� +�* +name + + " +x_369k +const +var_3140 + + +* +name + + " +op_3140*) +val" + + + + + ����������q +reshape +x + +x_369 +shape + + +var_3140 +var_3141 +  +� +�* +name + + " +op_3141� +constL +4model_transformer_h_23_attn_c_attn_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_23_attn_c_attn_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin��Ë� +linear +x + + +var_3141B +weight8 +6 +4model_transformer_h_23_attn_c_attn_weight_transposed3 +bias+ +) +'model_transformer_h_23_attn_c_attn_bias +x_371 +  +� +�* +name + + " +x_371d +const +var_3143 + + +* +name + + " +op_3143*" +val + + + +  +��w +reshape +x + +x_371 +shape + + +var_3143& +var_3144 +  + +� +�* +name + + " +op_3144d +const +tile_23 + + +* +name + + " +tile_23*# +val + + + + + +���^ +const +var_3145_axis_0 +*$ +name + +" +op_3145_axis_0* +val + + +� +split +x + + +var_3144 + split_sizes + +tile_23 +axis + +var_3145_axis_0( + +var_3145_0 +  + +� +�( + +var_3145_1 +  + +� +�( + +var_3145_2 +  + +� +�* +name + + " +op_3145d +const +var_3151 + + +* +name + + " +op_3151*" +val + + + +  +�@� +reshape +x + + +var_3145_0 +shape + + +var_3151- + +tensor_371 +  + +� + +@* +name + +" + +tensor_371d +const +var_3157 + + +* +name + + " +op_3157*" +val + + + +  +�@� +reshape +x + + +var_3145_1 +shape + + +var_3157- + +tensor_375 +  + +� + +@* +name + +" + +tensor_375d +const +var_3163 + + +* +name + + " +op_3163*" +val + + + +  +�@� +reshape +x + + +var_3145_2 +shape + + +var_3163- + +tensor_379 +  + +� + +@* +name + +" + +tensor_379c +const +var_3165 + + +* +name + + " +op_3165*! +val + + + + + +} +const& +attn_weights_277_transpose_x_0 +*4 +name, +$ +"" +attn_weights_277_transpose_x_0* +val + + +} +const& +attn_weights_277_transpose_y_0 +*4 +name, +$ +"" +attn_weights_277_transpose_y_0* +val + + +| +const$ +transpose_118_perm_0 + + +** +name" + +" +transpose_118_perm_0*! +val + + + + + +| +const$ +transpose_119_perm_0 + + +** +name" + +" +transpose_119_perm_0*! +val + + + + + +� + transpose +x + + +tensor_375 +perm + +transpose_119_perm_00 + transpose_121 +  + + +@ +�*# +name + +" + transpose_121� + transpose +x + + +tensor_371 +perm + +transpose_118_perm_00 + transpose_122 +  + + +� +@*# +name + +" + transpose_122� +matmul +x + + transpose_122 +y + + transpose_1211 + transpose_x" + +attn_weights_277_transpose_x_01 + transpose_y" + +attn_weights_277_transpose_y_04 +attn_weights_277 +  + + +� +�*& +name + +" +attn_weights_277� +const& +_inversed_attn_weights_279_y_0 + *4 +name, +$ +"" +_inversed_attn_weights_279_y_0* +val +  + + + +>� +mul +x + +attn_weights_277' +y" + +_inversed_attn_weights_279_y_0> +_inversed_attn_weights_279 +  + + +� +�*0 +name( + +" +_inversed_attn_weights_279� +select) +cond! + +causal_mask_1_after_broadcast# +a + +_inversed_attn_weights_279% +b + +mask_value_1_after_broadcast4 +attn_weights_283 +  + + +� +�*& +name + +" +attn_weights_283� +add +x + +attn_weights_283 +y + +attention_mask_1- + input_327 +  + + +� +�* +name + + " + input_327� +softmax +x + + input_327 +axis + +var_64 +attn_weights_285 +  + + +� +�*& +name + +" +attn_weights_285q +const +tensor_381_transpose_x_0 +*. +name& + +" +tensor_381_transpose_x_0* +val + + +q +const +tensor_381_transpose_y_0 +*. +name& + +" +tensor_381_transpose_y_0* +val + + +� + transpose +x + + +tensor_379 +perm + + +var_31650 + transpose_123 +  + + +� +@*# +name + +" + transpose_123� +matmul +x + +attn_weights_285 +y + + transpose_123+ + transpose_x + +tensor_381_transpose_x_0+ + transpose_y + +tensor_381_transpose_y_0- + +tensor_381 +  + + +� +@* +name + +" + +tensor_381c +const +var_3196 + + +* +name + + " +op_3196*! +val + + + + + +k +const +var_3208 + + +* +name + + " +op_3208*) +val" + + + + + ����������� + transpose +x + + +tensor_381 +perm + + +var_31960 + transpose_120 +  + +� + +@*# +name + +" + transpose_120y +reshape +x + + transpose_120 +shape + + +var_3208 +var_3209 +  +� +�* +name + + " +op_3209� +constL +4model_transformer_h_23_attn_c_proj_weight_transposed +  +� +�*J +nameB +: +8"6 +4model_transformer_h_23_attn_c_proj_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin��Ñ� +linear +x + + +var_3209B +weight8 +6 +4model_transformer_h_23_attn_c_proj_weight_transposed3 +bias+ +) +'model_transformer_h_23_attn_c_proj_bias +x_375 +  +� +�* +name + + " +x_375d +const +var_3211 + + +* +name + + " +op_3211*" +val + + + +  +��z +reshape +x + +x_375 +shape + + +var_3211' + input_331 +  + +� +�* +name + + " + input_331w +add +x + + input_331 +y + + input_325' + input_333 +  + +� +�* +name + + " + input_333r +const + x_377_axes_0 + + +*" +name + +" + x_377_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_333 +axes + + x_377_axes_0/ +gamma& +$ +"model_transformer_h_23_ln_2_weight, +beta$ +" + model_transformer_h_23_ln_2_bias +epsilon + + +var_26# +x_377 +  + +� +�* +name + + " +x_377k +const +var_3226 + + +* +name + + " +op_3226*) +val" + + + + + ����������q +reshape +x + +x_377 +shape + + +var_3226 +var_3227 +  +� +�* +name + + " +op_3227� +constI +1model_transformer_h_23_mlp_c_fc_weight_transposed +  +�  +�*G +name? +7 +5"3 +1model_transformer_h_23_mlp_c_fc_weight_transposed*E +val> +  +�  +�*& +@model_path/weights/weight.bin��Ó� +linear +x + + +var_3227? +weight5 +3 +1model_transformer_h_23_mlp_c_fc_weight_transposed0 +bias( +& +$model_transformer_h_23_mlp_c_fc_bias +x_379 +  +� +� * +name + + " +x_379d +const +var_3229 + + +* +name + + " +op_3229*" +val + + + +  +�� z +reshape +x + +x_379 +shape + + +var_3229' + input_335 +  + +� +� * +name + + " + input_335j +const + x_381_mode_0 +*" +name + +" + x_381_mode_0*' +val  + +" +TANH_APPROXIMATIONv +gelu +x + + input_335 +mode + + x_381_mode_0# +x_381 +  + +� +� * +name + + " +x_381k +const +var_3248 + + +* +name + + " +op_3248*) +val" + + + + + ���������� q +reshape +x + +x_381 +shape + + +var_3248 +var_3249 +  +� +� * +name + + " +op_3249� +constK +3model_transformer_h_23_mlp_c_proj_weight_transposed +  +� +� *I +nameA +9 +7"5 +3model_transformer_h_23_mlp_c_proj_weight_transposed*E +val> +  +� +� *& +@model_path/weights/weight.bin��Û� +linear +x + + +var_3249A +weight7 +5 +3model_transformer_h_23_mlp_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_23_mlp_c_proj_bias +x +  +� +�* +name + +" +xd +const +var_3251 + + +* +name + + " +op_3251*" +val + + + +  +��v +reshape + +x + +x +shape + + +var_3251' + input_337 +  + +� +�* +name + + " + input_337w +add +x + + input_333 +y + + input_337' + input_339 +  + +� +�* +name + + " + input_339� +const$ +hidden_states_axes_0 + + +** +name" + +" +hidden_states_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_339 +axes + +hidden_states_axes_0* +gamma! + +model_transformer_ln_f_weight' +beta + +model_transformer_ln_f_bias +epsilon + + +var_26+ + hidden_states +  + +� +�*# +name + +" + hidden_states� +const! +var_3262_bias_0 +   +ш*$ +name + +" +op_3262_bias_0*? +val8 +   +ш*& +@model_path/weights/weight.bin��ã� +linear +x + + hidden_states* +weight + +model_transformer_wte_weight +bias + +var_3262_bias_0% +logits +  + +� +ш* +name + + " +op_3262"� + buildInfo�" + + +~"| +8 +! + +" +coremltools-version + + " +7.0b1 +@ +) +! +" +coremltools-component-torch + + " +2.0.1 \ No newline at end of file