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 +1,13736 @@ +� +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) * +���R@ +logits&Classification scores (before softmax)* +������� +4kykim/gpt3-kor-small_based_on_gpt2 (text-generation)�3 +#com.github.apple.coremltools.source torch==2.1.0�+ +$com.github.apple.coremltools.version7.1�: ++com.github.apple.coremltools.source_dialect TorchScript�C +co.huggingface.exporters.name"kykim/gpt3-kor-small_based_on_gpt2�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 +  +�� +�*2 +name* +" + " +model_transformer_wte_weight*B +val; +  +�� +�*" +@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���=� +const2 +!model_transformer_h_0_ln_1_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_0_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const7 +&model_transformer_h_0_attn_c_attn_bias +   +�*< +name4 +, +*"( +&model_transformer_h_0_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const7 +&model_transformer_h_0_attn_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_0_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const0 +model_transformer_h_0_ln_2_bias +   +�*5 +name- +% +#"! +model_transformer_h_0_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const2 +!model_transformer_h_0_ln_2_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_0_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +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���=� +const6 +%model_transformer_h_0_mlp_c_proj_bias +   +�*; +name3 ++ +)"' +%model_transformer_h_0_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const0 +model_transformer_h_1_ln_1_bias +   +�*5 +name- +% +#"! +model_transformer_h_1_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const2 +!model_transformer_h_1_ln_1_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_1_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const7 +&model_transformer_h_1_attn_c_attn_bias +   +�*< +name4 +, +*"( +&model_transformer_h_1_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const7 +&model_transformer_h_1_attn_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_1_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const0 +model_transformer_h_1_ln_2_bias +   +�*5 +name- +% +#"! +model_transformer_h_1_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const2 +!model_transformer_h_1_ln_2_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_1_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +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���=� +const6 +%model_transformer_h_1_mlp_c_proj_bias +   +�*; +name3 ++ +)"' +%model_transformer_h_1_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const0 +model_transformer_h_2_ln_1_bias +   +�*5 +name- +% +#"! +model_transformer_h_2_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const2 +!model_transformer_h_2_ln_1_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_2_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const7 +&model_transformer_h_2_attn_c_attn_bias +   +�*< +name4 +, +*"( +&model_transformer_h_2_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const7 +&model_transformer_h_2_attn_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_2_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const0 +model_transformer_h_2_ln_2_bias +   +�*5 +name- +% +#"! +model_transformer_h_2_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const2 +!model_transformer_h_2_ln_2_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_2_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +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���=� +const6 +%model_transformer_h_2_mlp_c_proj_bias +   +�*; +name3 ++ +)"' +%model_transformer_h_2_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const0 +model_transformer_h_3_ln_1_bias +   +�*5 +name- +% +#"! +model_transformer_h_3_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const2 +!model_transformer_h_3_ln_1_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_3_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const7 +&model_transformer_h_3_attn_c_attn_bias +   +�*< +name4 +, +*"( +&model_transformer_h_3_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const7 +&model_transformer_h_3_attn_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_3_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const0 +model_transformer_h_3_ln_2_bias +   +�*5 +name- +% +#"! +model_transformer_h_3_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const2 +!model_transformer_h_3_ln_2_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_3_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +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���=� +const6 +%model_transformer_h_3_mlp_c_proj_bias +   +�*; +name3 ++ +)"' +%model_transformer_h_3_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const0 +model_transformer_h_4_ln_1_bias +   +�*5 +name- +% +#"! +model_transformer_h_4_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const2 +!model_transformer_h_4_ln_1_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_4_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const7 +&model_transformer_h_4_attn_c_attn_bias +   +�*< +name4 +, +*"( +&model_transformer_h_4_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const7 +&model_transformer_h_4_attn_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_4_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const0 +model_transformer_h_4_ln_2_bias +   +�*5 +name- +% +#"! +model_transformer_h_4_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const2 +!model_transformer_h_4_ln_2_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_4_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +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���=� +const6 +%model_transformer_h_4_mlp_c_proj_bias +   +�*; +name3 ++ +)"' +%model_transformer_h_4_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const0 +model_transformer_h_5_ln_1_bias +   +�*5 +name- +% +#"! +model_transformer_h_5_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const2 +!model_transformer_h_5_ln_1_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_5_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const7 +&model_transformer_h_5_attn_c_attn_bias +   +�*< +name4 +, +*"( +&model_transformer_h_5_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const7 +&model_transformer_h_5_attn_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_5_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const0 +model_transformer_h_5_ln_2_bias +   +�*5 +name- +% +#"! +model_transformer_h_5_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const2 +!model_transformer_h_5_ln_2_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_5_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +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���=� +const6 +%model_transformer_h_5_mlp_c_proj_bias +   +�*; +name3 ++ +)"' +%model_transformer_h_5_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const0 +model_transformer_h_6_ln_1_bias +   +�*5 +name- +% +#"! +model_transformer_h_6_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const2 +!model_transformer_h_6_ln_1_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_6_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const7 +&model_transformer_h_6_attn_c_attn_bias +   +�*< +name4 +, +*"( +&model_transformer_h_6_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const7 +&model_transformer_h_6_attn_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_6_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const0 +model_transformer_h_6_ln_2_bias +   +�*5 +name- +% +#"! +model_transformer_h_6_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const2 +!model_transformer_h_6_ln_2_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_6_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +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���=� +const6 +%model_transformer_h_6_mlp_c_proj_bias +   +�*; +name3 ++ +)"' +%model_transformer_h_6_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const0 +model_transformer_h_7_ln_1_bias +   +�*5 +name- +% +#"! +model_transformer_h_7_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const2 +!model_transformer_h_7_ln_1_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_7_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const7 +&model_transformer_h_7_attn_c_attn_bias +   +�*< +name4 +, +*"( +&model_transformer_h_7_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const7 +&model_transformer_h_7_attn_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_7_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const0 +model_transformer_h_7_ln_2_bias +   +�*5 +name- +% +#"! +model_transformer_h_7_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const2 +!model_transformer_h_7_ln_2_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_7_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +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���=� +const6 +%model_transformer_h_7_mlp_c_proj_bias +   +�*; +name3 ++ +)"' +%model_transformer_h_7_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const0 +model_transformer_h_8_ln_1_bias +   +�*5 +name- +% +#"! +model_transformer_h_8_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const2 +!model_transformer_h_8_ln_1_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_8_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const7 +&model_transformer_h_8_attn_c_attn_bias +   +�*< +name4 +, +*"( +&model_transformer_h_8_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const7 +&model_transformer_h_8_attn_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_8_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const0 +model_transformer_h_8_ln_2_bias +   +�*5 +name- +% +#"! +model_transformer_h_8_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const2 +!model_transformer_h_8_ln_2_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_8_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +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���=� +const6 +%model_transformer_h_8_mlp_c_proj_bias +   +�*; +name3 ++ +)"' +%model_transformer_h_8_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const0 +model_transformer_h_9_ln_1_bias +   +�*5 +name- +% +#"! +model_transformer_h_9_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const2 +!model_transformer_h_9_ln_1_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_9_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const7 +&model_transformer_h_9_attn_c_attn_bias +   +�*< +name4 +, +*"( +&model_transformer_h_9_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const7 +&model_transformer_h_9_attn_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_9_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const0 +model_transformer_h_9_ln_2_bias +   +�*5 +name- +% +#"! +model_transformer_h_9_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const2 +!model_transformer_h_9_ln_2_weight +   +�*7 +name/ +' +%"# +!model_transformer_h_9_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +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���=� +const6 +%model_transformer_h_9_mlp_c_proj_bias +   +�*; +name3 ++ +)"' +%model_transformer_h_9_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const1 + model_transformer_h_10_ln_1_bias +   +�*6 +name. +& +$"" + model_transformer_h_10_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const3 +"model_transformer_h_10_ln_1_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_10_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const8 +'model_transformer_h_10_attn_c_attn_bias +   +�*= +name5 +- ++") +'model_transformer_h_10_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const8 +'model_transformer_h_10_attn_c_proj_bias +   +�*= +name5 +- ++") +'model_transformer_h_10_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const1 + model_transformer_h_10_ln_2_bias +   +�*6 +name. +& +$"" + model_transformer_h_10_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const3 +"model_transformer_h_10_ln_2_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_10_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const5 +$model_transformer_h_10_mlp_c_fc_bias +   +�*: +name2 +* +("& +$model_transformer_h_10_mlp_c_fc_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const7 +&model_transformer_h_10_mlp_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_10_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const1 + model_transformer_h_11_ln_1_bias +   +�*6 +name. +& +$"" + model_transformer_h_11_ln_1_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const3 +"model_transformer_h_11_ln_1_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_11_ln_1_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const8 +'model_transformer_h_11_attn_c_attn_bias +   +�*= +name5 +- ++") +'model_transformer_h_11_attn_c_attn_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const8 +'model_transformer_h_11_attn_c_proj_bias +   +�*= +name5 +- ++") +'model_transformer_h_11_attn_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const1 + model_transformer_h_11_ln_2_bias +   +�*6 +name. +& +$"" + model_transformer_h_11_ln_2_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const3 +"model_transformer_h_11_ln_2_weight +   +�*8 +name0 +( +&"$ +"model_transformer_h_11_ln_2_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const5 +$model_transformer_h_11_mlp_c_fc_bias +   +�*: +name2 +* +("& +$model_transformer_h_11_mlp_c_fc_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const7 +&model_transformer_h_11_mlp_c_proj_bias +   +�*< +name4 +, +*"( +&model_transformer_h_11_mlp_c_proj_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const, +model_transformer_ln_f_bias +   +�*1 +name) +! +" +model_transformer_ln_f_bias*= +val6 +   +�*% +@model_path/weights/weight.bin���=� +const. +model_transformer_ln_f_weight +   +�*3 +name+ +# +!" +model_transformer_ln_f_weight*= +val6 +   +�*% +@model_path/weights/weight.bin���=S +const +var_5 +* +name + + +" +op_5* +val + + + +���������O +const +var_15 + * +name + + " +op_15* +val +  + + + +�?O +const +var_26 + * +name + + " +op_26* +val +  + + + +��'7j +const + var_70_axes_0 + + +*" +name + +" + op_70_axes_0* +val + + + + +� + expand_dims +x + +attention_mask +axes + + var_70_axes_0# +var_70 + + + +�* +name + + " +op_70j +const + var_71_axes_0 + + +*" +name + +" + op_71_axes_0* +val + + + + +� + expand_dims +x + + +var_70 +axes + + var_71_axes_0) +var_71 + + + + +�* +name + + " +op_71_ +const +var_73_dtype_0 +*# +name + +" + op_73_dtype_0* +val + + +" +fp32� +cast +x + + +var_71 +dtype + +var_73_dtype_0+ +cast_197 +  + + + +�* +name + + " + +cast_197q +sub +x + + +var_15 +y + + +cast_197) +var_74 +  + + + +�* +name + + " +op_74O +const +var_75 + * +name + + " +op_75* +val +  + + + +���� +mul +x + + +var_74 +y + + +var_753 +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���=} +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_100 + + +* +name + + +" +op_100*) +val" + + + + + ����������l +reshape +x + +x_1 +shape + +var_100 +var_101 +  +� +�* +name + + +" +op_101� +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���=� +linear +x + +var_101A +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_103 + + +* +name + + +" +op_103*" +val + + + +  +��r +reshape +x + +x_3 +shape + +var_103% +var_104 +  + +� +�* +name + + +" +op_104b +const +tile_0 + + +* +name + + +" +tile_0*# +val + + + + + +���\ +const +var_105_axis_0 +*# +name + +" + op_105_axis_0* +val + + +� +split +x + +var_104 + split_sizes + + +tile_0 +axis + +var_105_axis_0' + var_105_0 +  + +� +�' + var_105_1 +  + +� +�' + var_105_2 +  + +� +�* +name + + +" +op_105b +const +var_111 + + +* +name + + +" +op_111*" +val + + + +  +� @� +reshape +x + + var_105_0 +shape + +var_111+ +tensor_3 +  + +� +  +@* +name + + " + +tensor_3b +const +var_117 + + +* +name + + +" +op_117*" +val + + + +  +� @� +reshape +x + + var_105_1 +shape + +var_117+ +tensor_7 +  + +� +  +@* +name + + " + +tensor_7b +const +var_123 + + +* +name + + +" +op_123*" +val + + + +  +� @� +reshape +x + + var_105_2 +shape + +var_123, + tensor_11 +  + +� +  +@* +name + + " + tensor_11a +const +var_125 + + +* +name + + +" +op_125*! +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_36_perm_0 + + +*) +name! + +" +transpose_36_perm_0*! +val + + + + + +z +const# +transpose_37_perm_0 + + +*) +name! + +" +transpose_37_perm_0*! +val + + + + + +� + transpose +x + + +tensor_7 +perm + +transpose_37_perm_00 + transpose_105 +  + +  +@ +�*# +name + +" + transpose_105� + transpose +x + + +tensor_3 +perm + +transpose_36_perm_00 + transpose_106 +  + +  +� +@*# +name + +" + transpose_106� +matmul +x + + transpose_106 +y + + transpose_105/ + 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� +const6 +attn_weights_7_x_0 +  + +  +� +�*( +name  + +" +attn_weights_7_x_0*P +valI +  + +  +� +�*% +@model_path/weights/weight.bin���A� +add +x + +attn_weights_7_x_0! +y + +_inversed_attn_weights_32 +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_52 +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_1250 + transpose_107 +  + +  +� +@*# +name + +" + transpose_107� +matmul +x + +attn_weights_9 +y + + transpose_107* + transpose_x + +tensor_13_transpose_x_0* + transpose_y + +tensor_13_transpose_y_0, + tensor_13 +  + +  +� +@* +name + + " + tensor_13a +const +var_156 + + +* +name + + +" +op_156*! +val + + + + + +i +const +var_168 + + +* +name + + +" +op_168*) +val" + + + + + ����������� + transpose +x + + tensor_13 +perm + +var_1560 + transpose_104 +  + +� +  +@*# +name + +" + transpose_104v +reshape +x + + transpose_104 +shape + +var_168 +var_169 +  +� +�* +name + + +" +op_169� +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���A� +linear +x + +var_169A +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_171 + + +* +name + + +" +op_171*" +val + + + +  +��s +reshape +x + +x_7 +shape + +var_171% +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_186 + + +* +name + + +" +op_186*) +val" + + + + + ����������l +reshape +x + +x_9 +shape + +var_186 +var_187 +  +� +�* +name + + +" +op_187� +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���B� +linear +x + +var_187> +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_189 + + +* +name + + +" +op_189*" +val + + + +  +��v +reshape +x + +x_11 +shape + +var_189& +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_208 + + +* +name + + +" +op_208*) +val" + + + + + ����������m +reshape +x + +x_13 +shape + +var_208 +var_209 +  +� +�* +name + + +" +op_209� +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���G� +linear +x + +var_209@ +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_211 + + +* +name + + +" +op_211*" +val + + + +  +��v +reshape +x + +x_15 +shape + +var_211& +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_231 + + +* +name + + +" +op_231*) +val" + + + + + ����������m +reshape +x + +x_17 +shape + +var_231 +var_232 +  +� +�* +name + + +" +op_232� +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���K� +linear +x + +var_232A +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_234 + + +* +name + + +" +op_234*" +val + + + +  +��s +reshape +x + +x_19 +shape + +var_234% +var_235 +  + +� +�* +name + + +" +op_235b +const +tile_1 + + +* +name + + +" +tile_1*# +val + + + + + +���\ +const +var_236_axis_0 +*# +name + +" + op_236_axis_0* +val + + +� +split +x + +var_235 + split_sizes + + +tile_1 +axis + +var_236_axis_0' + var_236_0 +  + +� +�' + var_236_1 +  + +� +�' + var_236_2 +  + +� +�* +name + + +" +op_236b +const +var_242 + + +* +name + + +" +op_242*" +val + + + +  +� @� +reshape +x + + var_236_0 +shape + +var_242, + tensor_19 +  + +� +  +@* +name + + " + tensor_19b +const +var_248 + + +* +name + + +" +op_248*" +val + + + +  +� @� +reshape +x + + var_236_1 +shape + +var_248, + tensor_23 +  + +� +  +@* +name + + " + tensor_23b +const +var_254 + + +* +name + + +" +op_254*" +val + + + +  +� @� +reshape +x + + var_236_2 +shape + +var_254, + tensor_27 +  + +� +  +@* +name + + " + tensor_27a +const +var_256 + + +* +name + + +" +op_256*! +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_38_perm_0 + + +*) +name! + +" +transpose_38_perm_0*! +val + + + + + +z +const# +transpose_39_perm_0 + + +*) +name! + +" +transpose_39_perm_0*! +val + + + + + +� + transpose +x + + tensor_23 +perm + +transpose_39_perm_00 + transpose_101 +  + +  +@ +�*# +name + +" + transpose_101� + transpose +x + + tensor_19 +perm + +transpose_38_perm_00 + transpose_102 +  + +  +� +@*# +name + +" + transpose_102� +matmul +x + + transpose_102 +y + + transpose_1010 + 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� +add +x + +attn_weights_7_x_0" +y + +_inversed_attn_weights_153 +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_53 +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_2560 + transpose_103 +  + +  +� +@*# +name + +" + transpose_103� +matmul +x + +attn_weights_21 +y + + transpose_103* + transpose_x + +tensor_29_transpose_x_0* + transpose_y + +tensor_29_transpose_y_0, + tensor_29 +  + +  +� +@* +name + + " + tensor_29a +const +var_287 + + +* +name + + +" +op_287*! +val + + + + + +i +const +var_299 + + +* +name + + +" +op_299*) +val" + + + + + ����������� + transpose +x + + tensor_29 +perm + +var_2870 + transpose_100 +  + +� +  +@*# +name + +" + transpose_100v +reshape +x + + transpose_100 +shape + +var_299 +var_300 +  +� +�* +name + + +" +op_300� +constK +3model_transformer_h_1_attn_c_proj_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_1_attn_c_proj_weight_transposed*D +val= +  +� +�*% +@model_path/weights/weight.bin���O� +linear +x + +var_300A +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_302 + + +* +name + + +" +op_302*" +val + + + +  +��v +reshape +x + +x_23 +shape + +var_302& +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_317 + + +* +name + + +" +op_317*) +val" + + + + + ����������m +reshape +x + +x_25 +shape + +var_317 +var_318 +  +� +�* +name + + +" +op_318� +constH +0model_transformer_h_1_mlp_c_fc_weight_transposed +  +� +�*F +name> +6 +4"2 +0model_transformer_h_1_mlp_c_fc_weight_transposed*D +val= +  +� +�*% +@model_path/weights/weight.bin���P� +linear +x + +var_318> +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_320 + + +* +name + + +" +op_320*" +val + + + +  +��v +reshape +x + +x_27 +shape + +var_320& +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_339 + + +* +name + + +" +op_339*) +val" + + + + + ����������m +reshape +x + +x_29 +shape + +var_339 +var_340 +  +� +�* +name + + +" +op_340� +constJ +2model_transformer_h_1_mlp_c_proj_weight_transposed +  +� +�*H +name@ +8 +6"4 +2model_transformer_h_1_mlp_c_proj_weight_transposed*D +val= +  +� +�*% +@model_path/weights/weight.bin���T� +linear +x + +var_340@ +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_342 + + +* +name + + +" +op_342*" +val + + + +  +��v +reshape +x + +x_31 +shape + +var_342& +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_362 + + +* +name + + +" +op_362*) +val" + + + + + ����������m +reshape +x + +x_33 +shape + +var_362 +var_363 +  +� +�* +name + + +" +op_363� +constK +3model_transformer_h_2_attn_c_attn_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_2_attn_c_attn_weight_transposed*D +val= +  +� +�*% +@model_path/weights/weight.bin���Y� +linear +x + +var_363A +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_365 + + +* +name + + +" +op_365*" +val + + + +  +��s +reshape +x + +x_35 +shape + +var_365% +var_366 +  + +� +�* +name + + +" +op_366b +const +tile_2 + + +* +name + + +" +tile_2*# +val + + + + + +���\ +const +var_367_axis_0 +*# +name + +" + op_367_axis_0* +val + + +� +split +x + +var_366 + split_sizes + + +tile_2 +axis + +var_367_axis_0' + var_367_0 +  + +� +�' + var_367_1 +  + +� +�' + var_367_2 +  + +� +�* +name + + +" +op_367b +const +var_373 + + +* +name + + +" +op_373*" +val + + + +  +� @� +reshape +x + + var_367_0 +shape + +var_373, + tensor_35 +  + +� +  +@* +name + + " + tensor_35b +const +var_379 + + +* +name + + +" +op_379*" +val + + + +  +� @� +reshape +x + + var_367_1 +shape + +var_379, + tensor_39 +  + +� +  +@* +name + + " + tensor_39b +const +var_385 + + +* +name + + +" +op_385*" +val + + + +  +� @� +reshape +x + + var_367_2 +shape + +var_385, + tensor_43 +  + +� +  +@* +name + + " + tensor_43a +const +var_387 + + +* +name + + +" +op_387*! +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_40_perm_0 + + +*) +name! + +" +transpose_40_perm_0*! +val + + + + + +z +const# +transpose_41_perm_0 + + +*) +name! + +" +transpose_41_perm_0*! +val + + + + + +� + transpose +x + + tensor_39 +perm + +transpose_41_perm_0/ + transpose_97 +  + +  +@ +�*" +name + +" + transpose_97� + transpose +x + + tensor_35 +perm + +transpose_40_perm_0/ + transpose_98 +  + +  +� +@*" +name + +" + transpose_98� +matmul +x + + transpose_98 +y + + transpose_970 + 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� +add +x + +attn_weights_7_x_0" +y + +_inversed_attn_weights_273 +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_53 +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_387/ + transpose_99 +  + +  +� +@*" +name + +" + transpose_99� +matmul +x + +attn_weights_33 +y + + transpose_99* + transpose_x + +tensor_45_transpose_x_0* + transpose_y + +tensor_45_transpose_y_0, + tensor_45 +  + +  +� +@* +name + + " + tensor_45a +const +var_418 + + +* +name + + +" +op_418*! +val + + + + + +i +const +var_430 + + +* +name + + +" +op_430*) +val" + + + + + ����������� + transpose +x + + tensor_45 +perm + +var_418/ + transpose_96 +  + +� +  +@*" +name + +" + transpose_96u +reshape +x + + transpose_96 +shape + +var_430 +var_431 +  +� +�* +name + + +" +op_431� +constK +3model_transformer_h_2_attn_c_proj_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_2_attn_c_proj_weight_transposed*D +val= +  +� +�*% +@model_path/weights/weight.bin���\� +linear +x + +var_431A +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_433 + + +* +name + + +" +op_433*" +val + + + +  +��v +reshape +x + +x_39 +shape + +var_433& +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_448 + + +* +name + + +" +op_448*) +val" + + + + + ����������m +reshape +x + +x_41 +shape + +var_448 +var_449 +  +� +�* +name + + +" +op_449� +constH +0model_transformer_h_2_mlp_c_fc_weight_transposed +  +� +�*F +name> +6 +4"2 +0model_transformer_h_2_mlp_c_fc_weight_transposed*D +val= +  +� +�*% +@model_path/weights/weight.bin���]� +linear +x + +var_449> +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_451 + + +* +name + + +" +op_451*" +val + + + +  +��v +reshape +x + +x_43 +shape + +var_451& +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_470 + + +* +name + + +" +op_470*) +val" + + + + + ����������m +reshape +x + +x_45 +shape + +var_470 +var_471 +  +� +�* +name + + +" +op_471� +constJ +2model_transformer_h_2_mlp_c_proj_weight_transposed +  +� +�*H +name@ +8 +6"4 +2model_transformer_h_2_mlp_c_proj_weight_transposed*D +val= +  +� +�*% +@model_path/weights/weight.bin���b� +linear +x + +var_471@ +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_473 + + +* +name + + +" +op_473*" +val + + + +  +��v +reshape +x + +x_47 +shape + +var_473& +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_493 + + +* +name + + +" +op_493*) +val" + + + + + ����������m +reshape +x + +x_49 +shape + +var_493 +var_494 +  +� +�* +name + + +" +op_494� +constK +3model_transformer_h_3_attn_c_attn_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_3_attn_c_attn_weight_transposed*D +val= +  +� +�*% +@model_path/weights/weight.bin���f� +linear +x + +var_494A +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_496 + + +* +name + + +" +op_496*" +val + + + +  +��s +reshape +x + +x_51 +shape + +var_496% +var_497 +  + +� +�* +name + + +" +op_497b +const +tile_3 + + +* +name + + +" +tile_3*# +val + + + + + +���\ +const +var_498_axis_0 +*# +name + +" + op_498_axis_0* +val + + +� +split +x + +var_497 + split_sizes + + +tile_3 +axis + +var_498_axis_0' + var_498_0 +  + +� +�' + var_498_1 +  + +� +�' + var_498_2 +  + +� +�* +name + + +" +op_498b +const +var_504 + + +* +name + + +" +op_504*" +val + + + +  +� @� +reshape +x + + var_498_0 +shape + +var_504, + tensor_51 +  + +� +  +@* +name + + " + tensor_51b +const +var_510 + + +* +name + + +" +op_510*" +val + + + +  +� @� +reshape +x + + var_498_1 +shape + +var_510, + tensor_55 +  + +� +  +@* +name + + " + tensor_55b +const +var_516 + + +* +name + + +" +op_516*" +val + + + +  +� @� +reshape +x + + var_498_2 +shape + +var_516, + tensor_59 +  + +� +  +@* +name + + " + tensor_59a +const +var_518 + + +* +name + + +" +op_518*! +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_42_perm_0 + + +*) +name! + +" +transpose_42_perm_0*! +val + + + + + +z +const# +transpose_43_perm_0 + + +*) +name! + +" +transpose_43_perm_0*! +val + + + + + +� + transpose +x + + tensor_55 +perm + +transpose_43_perm_0/ + transpose_93 +  + +  +@ +�*" +name + +" + transpose_93� + transpose +x + + tensor_51 +perm + +transpose_42_perm_0/ + transpose_94 +  + +  +� +@*" +name + +" + transpose_94� +matmul +x + + transpose_94 +y + + transpose_930 + 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� +add +x + +attn_weights_7_x_0" +y + +_inversed_attn_weights_393 +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_53 +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_518/ + transpose_95 +  + +  +� +@*" +name + +" + transpose_95� +matmul +x + +attn_weights_45 +y + + transpose_95* + transpose_x + +tensor_61_transpose_x_0* + transpose_y + +tensor_61_transpose_y_0, + tensor_61 +  + +  +� +@* +name + + " + tensor_61a +const +var_549 + + +* +name + + +" +op_549*! +val + + + + + +i +const +var_561 + + +* +name + + +" +op_561*) +val" + + + + + ����������� + transpose +x + + tensor_61 +perm + +var_549/ + transpose_92 +  + +� +  +@*" +name + +" + transpose_92u +reshape +x + + transpose_92 +shape + +var_561 +var_562 +  +� +�* +name + + +" +op_562� +constK +3model_transformer_h_3_attn_c_proj_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_3_attn_c_proj_weight_transposed*D +val= +  +� +�*% +@model_path/weights/weight.bin���j� +linear +x + +var_562A +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_564 + + +* +name + + +" +op_564*" +val + + + +  +��v +reshape +x + +x_55 +shape + +var_564& +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_579 + + +* +name + + +" +op_579*) +val" + + + + + ����������m +reshape +x + +x_57 +shape + +var_579 +var_580 +  +� +�* +name + + +" +op_580� +constH +0model_transformer_h_3_mlp_c_fc_weight_transposed +  +� +�*F +name> +6 +4"2 +0model_transformer_h_3_mlp_c_fc_weight_transposed*D +val= +  +� +�*% +@model_path/weights/weight.bin���k� +linear +x + +var_580> +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_582 + + +* +name + + +" +op_582*" +val + + + +  +��v +reshape +x + +x_59 +shape + +var_582& +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_601 + + +* +name + + +" +op_601*) +val" + + + + + ����������m +reshape +x + +x_61 +shape + +var_601 +var_602 +  +� +�* +name + + +" +op_602� +constJ +2model_transformer_h_3_mlp_c_proj_weight_transposed +  +� +�*H +name@ +8 +6"4 +2model_transformer_h_3_mlp_c_proj_weight_transposed*D +val= +  +� +�*% +@model_path/weights/weight.bin���o� +linear +x + +var_602@ +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_604 + + +* +name + + +" +op_604*" +val + + + +  +��v +reshape +x + +x_63 +shape + +var_604& +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_624 + + +* +name + + +" +op_624*) +val" + + + + + ����������m +reshape +x + +x_65 +shape + +var_624 +var_625 +  +� +�* +name + + +" +op_625� +constK +3model_transformer_h_4_attn_c_attn_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_4_attn_c_attn_weight_transposed*D +val= +  +� +�*% +@model_path/weights/weight.bin���t� +linear +x + +var_625A +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_627 + + +* +name + + +" +op_627*" +val + + + +  +��s +reshape +x + +x_67 +shape + +var_627% +var_628 +  + +� +�* +name + + +" +op_628b +const +tile_4 + + +* +name + + +" +tile_4*# +val + + + + + +���\ +const +var_629_axis_0 +*# +name + +" + op_629_axis_0* +val + + +� +split +x + +var_628 + split_sizes + + +tile_4 +axis + +var_629_axis_0' + var_629_0 +  + +� +�' + var_629_1 +  + +� +�' + var_629_2 +  + +� +�* +name + + +" +op_629b +const +var_635 + + +* +name + + +" +op_635*" +val + + + +  +� @� +reshape +x + + var_629_0 +shape + +var_635, + tensor_67 +  + +� +  +@* +name + + " + tensor_67b +const +var_641 + + +* +name + + +" +op_641*" +val + + + +  +� @� +reshape +x + + var_629_1 +shape + +var_641, + tensor_71 +  + +� +  +@* +name + + " + tensor_71b +const +var_647 + + +* +name + + +" +op_647*" +val + + + +  +� @� +reshape +x + + var_629_2 +shape + +var_647, + tensor_75 +  + +� +  +@* +name + + " + tensor_75a +const +var_649 + + +* +name + + +" +op_649*! +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_44_perm_0 + + +*) +name! + +" +transpose_44_perm_0*! +val + + + + + +z +const# +transpose_45_perm_0 + + +*) +name! + +" +transpose_45_perm_0*! +val + + + + + +� + transpose +x + + tensor_71 +perm + +transpose_45_perm_0/ + transpose_89 +  + +  +@ +�*" +name + +" + transpose_89� + transpose +x + + tensor_67 +perm + +transpose_44_perm_0/ + transpose_90 +  + +  +� +@*" +name + +" + transpose_90� +matmul +x + + transpose_90 +y + + transpose_890 + 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� +add +x + +attn_weights_7_x_0" +y + +_inversed_attn_weights_513 +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_53 +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_649/ + transpose_91 +  + +  +� +@*" +name + +" + transpose_91� +matmul +x + +attn_weights_57 +y + + transpose_91* + transpose_x + +tensor_77_transpose_x_0* + transpose_y + +tensor_77_transpose_y_0, + tensor_77 +  + +  +� +@* +name + + " + tensor_77a +const +var_680 + + +* +name + + +" +op_680*! +val + + + + + +i +const +var_692 + + +* +name + + +" +op_692*) +val" + + + + + ����������� + transpose +x + + tensor_77 +perm + +var_680/ + transpose_88 +  + +� +  +@*" +name + +" + transpose_88u +reshape +x + + transpose_88 +shape + +var_692 +var_693 +  +� +�* +name + + +" +op_693� +constK +3model_transformer_h_4_attn_c_proj_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_4_attn_c_proj_weight_transposed*D +val= +  +� +�*% +@model_path/weights/weight.bin���w� +linear +x + +var_693A +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_695 + + +* +name + + +" +op_695*" +val + + + +  +��v +reshape +x + +x_71 +shape + +var_695& +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_710 + + +* +name + + +" +op_710*) +val" + + + + + ����������m +reshape +x + +x_73 +shape + +var_710 +var_711 +  +� +�* +name + + +" +op_711� +constH +0model_transformer_h_4_mlp_c_fc_weight_transposed +  +� +�*F +name> +6 +4"2 +0model_transformer_h_4_mlp_c_fc_weight_transposed*D +val= +  +� +�*% +@model_path/weights/weight.bin���x� +linear +x + +var_711> +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_713 + + +* +name + + +" +op_713*" +val + + + +  +��v +reshape +x + +x_75 +shape + +var_713& +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_732 + + +* +name + + +" +op_732*) +val" + + + + + ����������m +reshape +x + +x_77 +shape + +var_732 +var_733 +  +� +�* +name + + +" +op_733� +constJ +2model_transformer_h_4_mlp_c_proj_weight_transposed +  +� +�*H +name@ +8 +6"4 +2model_transformer_h_4_mlp_c_proj_weight_transposed*D +val= +  +� +�*% +@model_path/weights/weight.bin���}� +linear +x + +var_733@ +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_735 + + +* +name + + +" +op_735*" +val + + + +  +��v +reshape +x + +x_79 +shape + +var_735& +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_755 + + +* +name + + +" +op_755*) +val" + + + + + ����������m +reshape +x + +x_81 +shape + +var_755 +var_756 +  +� +�* +name + + +" +op_756� +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_756A +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_758 + + +* +name + + +" +op_758*" +val + + + +  +��s +reshape +x + +x_83 +shape + +var_758% +var_759 +  + +� +�* +name + + +" +op_759b +const +tile_5 + + +* +name + + +" +tile_5*# +val + + + + + +���\ +const +var_760_axis_0 +*# +name + +" + op_760_axis_0* +val + + +� +split +x + +var_759 + split_sizes + + +tile_5 +axis + +var_760_axis_0' + var_760_0 +  + +� +�' + var_760_1 +  + +� +�' + var_760_2 +  + +� +�* +name + + +" +op_760b +const +var_766 + + +* +name + + +" +op_766*" +val + + + +  +� @� +reshape +x + + var_760_0 +shape + +var_766, + tensor_83 +  + +� +  +@* +name + + " + tensor_83b +const +var_772 + + +* +name + + +" +op_772*" +val + + + +  +� @� +reshape +x + + var_760_1 +shape + +var_772, + tensor_87 +  + +� +  +@* +name + + " + tensor_87b +const +var_778 + + +* +name + + +" +op_778*" +val + + + +  +� @� +reshape +x + + var_760_2 +shape + +var_778, + tensor_91 +  + +� +  +@* +name + + " + tensor_91a +const +var_780 + + +* +name + + +" +op_780*! +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_46_perm_0 + + +*) +name! + +" +transpose_46_perm_0*! +val + + + + + +z +const# +transpose_47_perm_0 + + +*) +name! + +" +transpose_47_perm_0*! +val + + + + + +� + transpose +x + + tensor_87 +perm + +transpose_47_perm_0/ + transpose_85 +  + +  +@ +�*" +name + +" + transpose_85� + transpose +x + + tensor_83 +perm + +transpose_46_perm_0/ + transpose_86 +  + +  +� +@*" +name + +" + transpose_86� +matmul +x + + transpose_86 +y + + transpose_850 + 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� +add +x + +attn_weights_7_x_0" +y + +_inversed_attn_weights_633 +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_53 +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_780/ + transpose_87 +  + +  +� +@*" +name + +" + transpose_87� +matmul +x + +attn_weights_69 +y + + transpose_87* + transpose_x + +tensor_93_transpose_x_0* + transpose_y + +tensor_93_transpose_y_0, + tensor_93 +  + +  +� +@* +name + + " + tensor_93a +const +var_811 + + +* +name + + +" +op_811*! +val + + + + + +i +const +var_823 + + +* +name + + +" +op_823*) +val" + + + + + ����������� + transpose +x + + tensor_93 +perm + +var_811/ + transpose_84 +  + +� +  +@*" +name + +" + transpose_84u +reshape +x + + transpose_84 +shape + +var_823 +var_824 +  +� +�* +name + + +" +op_824� +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_824A +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_826 + + +* +name + + +" +op_826*" +val + + + +  +��v +reshape +x + +x_87 +shape + +var_826& +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_841 + + +* +name + + +" +op_841*) +val" + + + + + ����������m +reshape +x + +x_89 +shape + +var_841 +var_842 +  +� +�* +name + + +" +op_842� +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_842> +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_844 + + +* +name + + +" +op_844*" +val + + + +  +��v +reshape +x + +x_91 +shape + +var_844& +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_863 + + +* +name + + +" +op_863*) +val" + + + + + ����������m +reshape +x + +x_93 +shape + +var_863 +var_864 +  +� +�* +name + + +" +op_864� +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_864@ +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_866 + + +* +name + + +" +op_866*" +val + + + +  +��v +reshape +x + +x_95 +shape + +var_866& +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_886 + + +* +name + + +" +op_886*) +val" + + + + + ����������m +reshape +x + +x_97 +shape + +var_886 +var_887 +  +� +�* +name + + +" +op_887� +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_887A +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_889 + + +* +name + + +" +op_889*" +val + + + +  +��s +reshape +x + +x_99 +shape + +var_889% +var_890 +  + +� +�* +name + + +" +op_890b +const +tile_6 + + +* +name + + +" +tile_6*# +val + + + + + +���\ +const +var_891_axis_0 +*# +name + +" + op_891_axis_0* +val + + +� +split +x + +var_890 + split_sizes + + +tile_6 +axis + +var_891_axis_0' + var_891_0 +  + +� +�' + var_891_1 +  + +� +�' + var_891_2 +  + +� +�* +name + + +" +op_891b +const +var_897 + + +* +name + + +" +op_897*" +val + + + +  +� @� +reshape +x + + var_891_0 +shape + +var_897, + tensor_99 +  + +� +  +@* +name + + " + tensor_99b +const +var_903 + + +* +name + + +" +op_903*" +val + + + +  +� @� +reshape +x + + var_891_1 +shape + +var_903- + +tensor_103 +  + +� +  +@* +name + +" + +tensor_103b +const +var_909 + + +* +name + + +" +op_909*" +val + + + +  +� @� +reshape +x + + var_891_2 +shape + +var_909- + +tensor_107 +  + +� +  +@* +name + +" + +tensor_107a +const +var_911 + + +* +name + + +" +op_911*! +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_48_perm_0 + + +*) +name! + +" +transpose_48_perm_0*! +val + + + + + +z +const# +transpose_49_perm_0 + + +*) +name! + +" +transpose_49_perm_0*! +val + + + + + +� + transpose +x + + +tensor_103 +perm + +transpose_49_perm_0/ + transpose_81 +  + +  +@ +�*" +name + +" + transpose_81� + transpose +x + + tensor_99 +perm + +transpose_48_perm_0/ + transpose_82 +  + +  +� +@*" +name + +" + transpose_82� +matmul +x + + transpose_82 +y + + transpose_810 + 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� +add +x + +attn_weights_7_x_0" +y + +_inversed_attn_weights_753 +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_53 +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_911/ + transpose_83 +  + +  +� +@*" +name + +" + transpose_83� +matmul +x + +attn_weights_81 +y + + transpose_83+ + transpose_x + +tensor_109_transpose_x_0+ + transpose_y + +tensor_109_transpose_y_0- + +tensor_109 +  + +  +� +@* +name + +" + +tensor_109a +const +var_942 + + +* +name + + +" +op_942*! +val + + + + + +i +const +var_954 + + +* +name + + +" +op_954*) +val" + + + + + ����������� + transpose +x + + +tensor_109 +perm + +var_942/ + transpose_80 +  + +� +  +@*" +name + +" + transpose_80u +reshape +x + + transpose_80 +shape + +var_954 +var_955 +  +� +�* +name + + +" +op_955� +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_955A +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_957 + + +* +name + + +" +op_957*" +val + + + +  +��w +reshape +x + +x_103 +shape + +var_957& +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_972 + + +* +name + + +" +op_972*) +val" + + + + + ����������n +reshape +x + +x_105 +shape + +var_972 +var_973 +  +� +�* +name + + +" +op_973� +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_973> +weight4 +2 +0model_transformer_h_6_mlp_c_fc_weight_transposed/ +bias' +% +#model_transformer_h_6_mlp_c_fc_bias +x_107 +  +� +�* +name + + " +x_107b +const +var_975 + + +* +name + + +" +op_975*" +val + + + +  +��w +reshape +x + +x_107 +shape + +var_975& +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_109i +const +var_994 + + +* +name + + +" +op_994*) +val" + + + + + ����������n +reshape +x + +x_109 +shape + +var_994 +var_995 +  +� +�* +name + + +" +op_995� +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����� +linear +x + +var_995@ +weight6 +4 +2model_transformer_h_6_mlp_c_proj_weight_transposed1 +bias) +' +%model_transformer_h_6_mlp_c_proj_bias +x_111 +  +� +�* +name + + " +x_111b +const +var_997 + + +* +name + + +" +op_997*" +val + + + +  +��w +reshape +x + +x_111 +shape + +var_997& +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_1017 + + +* +name + + " +op_1017*) +val" + + + + + ����������q +reshape +x + +x_113 +shape + + +var_1017 +var_1018 +  +� +�* +name + + " +op_1018� +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���� +linear +x + + +var_1018A +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_1020 + + +* +name + + " +op_1020*" +val + + + +  +��w +reshape +x + +x_115 +shape + + +var_1020& +var_1021 +  + +� +�* +name + + " +op_1021b +const +tile_7 + + +* +name + + +" +tile_7*# +val + + + + + +���^ +const +var_1022_axis_0 +*$ +name + +" +op_1022_axis_0* +val + + +� +split +x + + +var_1021 + split_sizes + + +tile_7 +axis + +var_1022_axis_0( + +var_1022_0 +  + +� +�( + +var_1022_1 +  + +� +�( + +var_1022_2 +  + +� +�* +name + + " +op_1022d +const +var_1028 + + +* +name + + " +op_1028*" +val + + + +  +� @� +reshape +x + + +var_1022_0 +shape + + +var_1028- + +tensor_115 +  + +� +  +@* +name + +" + +tensor_115d +const +var_1034 + + +* +name + + " +op_1034*" +val + + + +  +� @� +reshape +x + + +var_1022_1 +shape + + +var_1034- + +tensor_119 +  + +� +  +@* +name + +" + +tensor_119d +const +var_1040 + + +* +name + + " +op_1040*" +val + + + +  +� @� +reshape +x + + +var_1022_2 +shape + + +var_1040- + +tensor_123 +  + +� +  +@* +name + +" + +tensor_123c +const +var_1042 + + +* +name + + " +op_1042*! +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_50_perm_0 + + +*) +name! + +" +transpose_50_perm_0*! +val + + + + + +z +const# +transpose_51_perm_0 + + +*) +name! + +" +transpose_51_perm_0*! +val + + + + + +� + transpose +x + + +tensor_119 +perm + +transpose_51_perm_0/ + transpose_77 +  + +  +@ +�*" +name + +" + transpose_77� + transpose +x + + +tensor_115 +perm + +transpose_50_perm_0/ + transpose_78 +  + +  +� +@*" +name + +" + transpose_78� +matmul +x + + transpose_78 +y + + transpose_770 + 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� +add +x + +attn_weights_7_x_0" +y + +_inversed_attn_weights_873 +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_53 +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_1042/ + transpose_79 +  + +  +� +@*" +name + +" + transpose_79� +matmul +x + +attn_weights_93 +y + + transpose_79+ + transpose_x + +tensor_125_transpose_x_0+ + transpose_y + +tensor_125_transpose_y_0- + +tensor_125 +  + +  +� +@* +name + +" + +tensor_125c +const +var_1073 + + +* +name + + " +op_1073*! +val + + + + + +k +const +var_1085 + + +* +name + + " +op_1085*) +val" + + + + + ����������� + transpose +x + + +tensor_125 +perm + + +var_1073/ + transpose_76 +  + +� +  +@*" +name + +" + transpose_76x +reshape +x + + transpose_76 +shape + + +var_1085 +var_1086 +  +� +�* +name + + " +op_1086� +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����� +linear +x + + +var_1086A +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_1088 + + +* +name + + " +op_1088*" +val + + + +  +��z +reshape +x + +x_119 +shape + + +var_1088' + 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_1103 + + +* +name + + " +op_1103*) +val" + + + + + ����������q +reshape +x + +x_121 +shape + + +var_1103 +var_1104 +  +� +�* +name + + " +op_1104� +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����� +linear +x + + +var_1104> +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_1106 + + +* +name + + " +op_1106*" +val + + + +  +��z +reshape +x + +x_123 +shape + + +var_1106' + 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_1125 + + +* +name + + " +op_1125*) +val" + + + + + ����������q +reshape +x + +x_125 +shape + + +var_1125 +var_1126 +  +� +�* +name + + " +op_1126� +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���� +linear +x + + +var_1126@ +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_1128 + + +* +name + + " +op_1128*" +val + + + +  +��z +reshape +x + +x_127 +shape + + +var_1128' + 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_1148 + + +* +name + + " +op_1148*) +val" + + + + + ����������q +reshape +x + +x_129 +shape + + +var_1148 +var_1149 +  +� +�* +name + + " +op_1149� +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����� +linear +x + + +var_1149A +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_1151 + + +* +name + + " +op_1151*" +val + + + +  +��w +reshape +x + +x_131 +shape + + +var_1151& +var_1152 +  + +� +�* +name + + " +op_1152b +const +tile_8 + + +* +name + + +" +tile_8*# +val + + + + + +���^ +const +var_1153_axis_0 +*$ +name + +" +op_1153_axis_0* +val + + +� +split +x + + +var_1152 + split_sizes + + +tile_8 +axis + +var_1153_axis_0( + +var_1153_0 +  + +� +�( + +var_1153_1 +  + +� +�( + +var_1153_2 +  + +� +�* +name + + " +op_1153d +const +var_1159 + + +* +name + + " +op_1159*" +val + + + +  +� @� +reshape +x + + +var_1153_0 +shape + + +var_1159- + +tensor_131 +  + +� +  +@* +name + +" + +tensor_131d +const +var_1165 + + +* +name + + " +op_1165*" +val + + + +  +� @� +reshape +x + + +var_1153_1 +shape + + +var_1165- + +tensor_135 +  + +� +  +@* +name + +" + +tensor_135d +const +var_1171 + + +* +name + + " +op_1171*" +val + + + +  +� @� +reshape +x + + +var_1153_2 +shape + + +var_1171- + +tensor_139 +  + +� +  +@* +name + +" + +tensor_139c +const +var_1173 + + +* +name + + " +op_1173*! +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_52_perm_0 + + +*) +name! + +" +transpose_52_perm_0*! +val + + + + + +z +const# +transpose_53_perm_0 + + +*) +name! + +" +transpose_53_perm_0*! +val + + + + + +� + transpose +x + + +tensor_135 +perm + +transpose_53_perm_0/ + transpose_73 +  + +  +@ +�*" +name + +" + transpose_73� + transpose +x + + +tensor_131 +perm + +transpose_52_perm_0/ + transpose_74 +  + +  +� +@*" +name + +" + transpose_74� +matmul +x + + transpose_74 +y + + transpose_730 + 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� +add +x + +attn_weights_7_x_0" +y + +_inversed_attn_weights_994 +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_54 +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_1173/ + transpose_75 +  + +  +� +@*" +name + +" + transpose_75� +matmul +x + +attn_weights_105 +y + + transpose_75+ + transpose_x + +tensor_141_transpose_x_0+ + transpose_y + +tensor_141_transpose_y_0- + +tensor_141 +  + +  +� +@* +name + +" + +tensor_141c +const +var_1204 + + +* +name + + " +op_1204*! +val + + + + + +k +const +var_1216 + + +* +name + + " +op_1216*) +val" + + + + + ����������� + transpose +x + + +tensor_141 +perm + + +var_1204/ + transpose_72 +  + +� +  +@*" +name + +" + transpose_72x +reshape +x + + transpose_72 +shape + + +var_1216 +var_1217 +  +� +�* +name + + " +op_1217� +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��٭� +linear +x + + +var_1217A +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_1219 + + +* +name + + " +op_1219*" +val + + + +  +��z +reshape +x + +x_135 +shape + + +var_1219' + 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_1234 + + +* +name + + " +op_1234*) +val" + + + + + ����������q +reshape +x + +x_137 +shape + + +var_1234 +var_1235 +  +� +�* +name + + " +op_1235� +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���� +linear +x + + +var_1235> +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_1237 + + +* +name + + " +op_1237*" +val + + + +  +��z +reshape +x + +x_139 +shape + + +var_1237' + 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_1256 + + +* +name + + " +op_1256*) +val" + + + + + ����������q +reshape +x + +x_141 +shape + + +var_1256 +var_1257 +  +� +�* +name + + " +op_1257� +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����� +linear +x + + +var_1257@ +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_1259 + + +* +name + + " +op_1259*" +val + + + +  +��z +reshape +x + +x_143 +shape + + +var_1259' + 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_1279 + + +* +name + + " +op_1279*) +val" + + + + + ����������q +reshape +x + +x_145 +shape + + +var_1279 +var_1280 +  +� +�* +name + + " +op_1280� +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���� +linear +x + + +var_1280A +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_1282 + + +* +name + + " +op_1282*" +val + + + +  +��w +reshape +x + +x_147 +shape + + +var_1282& +var_1283 +  + +� +�* +name + + " +op_1283b +const +tile_9 + + +* +name + + +" +tile_9*# +val + + + + + +���^ +const +var_1284_axis_0 +*$ +name + +" +op_1284_axis_0* +val + + +� +split +x + + +var_1283 + split_sizes + + +tile_9 +axis + +var_1284_axis_0( + +var_1284_0 +  + +� +�( + +var_1284_1 +  + +� +�( + +var_1284_2 +  + +� +�* +name + + " +op_1284d +const +var_1290 + + +* +name + + " +op_1290*" +val + + + +  +� @� +reshape +x + + +var_1284_0 +shape + + +var_1290- + +tensor_147 +  + +� +  +@* +name + +" + +tensor_147d +const +var_1296 + + +* +name + + " +op_1296*" +val + + + +  +� @� +reshape +x + + +var_1284_1 +shape + + +var_1296- + +tensor_151 +  + +� +  +@* +name + +" + +tensor_151d +const +var_1302 + + +* +name + + " +op_1302*" +val + + + +  +� @� +reshape +x + + +var_1284_2 +shape + + +var_1302- + +tensor_155 +  + +� +  +@* +name + +" + +tensor_155c +const +var_1304 + + +* +name + + " +op_1304*! +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_54_perm_0 + + +*) +name! + +" +transpose_54_perm_0*! +val + + + + + +z +const# +transpose_55_perm_0 + + +*) +name! + +" +transpose_55_perm_0*! +val + + + + + +� + transpose +x + + +tensor_151 +perm + +transpose_55_perm_0/ + transpose_69 +  + +  +@ +�*" +name + +" + transpose_69� + transpose +x + + +tensor_147 +perm + +transpose_54_perm_0/ + transpose_70 +  + +  +� +@*" +name + +" + transpose_70� +matmul +x + + transpose_70 +y + + transpose_691 + 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� +add +x + +attn_weights_7_x_0# +y + +_inversed_attn_weights_1114 +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_54 +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_1304/ + transpose_71 +  + +  +� +@*" +name + +" + transpose_71� +matmul +x + +attn_weights_117 +y + + transpose_71+ + transpose_x + +tensor_157_transpose_x_0+ + transpose_y + +tensor_157_transpose_y_0- + +tensor_157 +  + +  +� +@* +name + +" + +tensor_157c +const +var_1335 + + +* +name + + " +op_1335*! +val + + + + + +k +const +var_1347 + + +* +name + + " +op_1347*) +val" + + + + + ����������� + transpose +x + + +tensor_157 +perm + + +var_1335/ + transpose_68 +  + +� +  +@*" +name + +" + transpose_68x +reshape +x + + transpose_68 +shape + + +var_1347 +var_1348 +  +� +�* +name + + " +op_1348� +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_1348A +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_1350 + + +* +name + + " +op_1350*" +val + + + +  +��z +reshape +x + +x_151 +shape + + +var_1350' + 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_1365 + + +* +name + + " +op_1365*) +val" + + + + + ����������q +reshape +x + +x_153 +shape + + +var_1365 +var_1366 +  +� +�* +name + + " +op_1366� +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_1366> +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_1368 + + +* +name + + " +op_1368*" +val + + + +  +��z +reshape +x + +x_155 +shape + + +var_1368' + 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_1387 + + +* +name + + " +op_1387*) +val" + + + + + ����������q +reshape +x + +x_157 +shape + + +var_1387 +var_1388 +  +� +�* +name + + " +op_1388� +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_1388@ +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_1390 + + +* +name + + " +op_1390*" +val + + + +  +��z +reshape +x + +x_159 +shape + + +var_1390' + 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_1410 + + +* +name + + " +op_1410*) +val" + + + + + ����������q +reshape +x + +x_161 +shape + + +var_1410 +var_1411 +  +� +�* +name + + " +op_1411� +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_1411B +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_1413 + + +* +name + + " +op_1413*" +val + + + +  +��w +reshape +x + +x_163 +shape + + +var_1413& +var_1414 +  + +� +�* +name + + " +op_1414d +const +tile_10 + + +* +name + + " +tile_10*# +val + + + + + +���^ +const +var_1415_axis_0 +*$ +name + +" +op_1415_axis_0* +val + + +� +split +x + + +var_1414 + split_sizes + +tile_10 +axis + +var_1415_axis_0( + +var_1415_0 +  + +� +�( + +var_1415_1 +  + +� +�( + +var_1415_2 +  + +� +�* +name + + " +op_1415d +const +var_1421 + + +* +name + + " +op_1421*" +val + + + +  +� @� +reshape +x + + +var_1415_0 +shape + + +var_1421- + +tensor_163 +  + +� +  +@* +name + +" + +tensor_163d +const +var_1427 + + +* +name + + " +op_1427*" +val + + + +  +� @� +reshape +x + + +var_1415_1 +shape + + +var_1427- + +tensor_167 +  + +� +  +@* +name + +" + +tensor_167d +const +var_1433 + + +* +name + + " +op_1433*" +val + + + +  +� @� +reshape +x + + +var_1415_2 +shape + + +var_1433- + +tensor_171 +  + +� +  +@* +name + +" + +tensor_171c +const +var_1435 + + +* +name + + " +op_1435*! +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_56_perm_0 + + +*) +name! + +" +transpose_56_perm_0*! +val + + + + + +z +const# +transpose_57_perm_0 + + +*) +name! + +" +transpose_57_perm_0*! +val + + + + + +� + transpose +x + + +tensor_167 +perm + +transpose_57_perm_0/ + transpose_65 +  + +  +@ +�*" +name + +" + transpose_65� + transpose +x + + +tensor_163 +perm + +transpose_56_perm_0/ + transpose_66 +  + +  +� +@*" +name + +" + transpose_66� +matmul +x + + transpose_66 +y + + transpose_651 + 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� +add +x + +attn_weights_7_x_0# +y + +_inversed_attn_weights_1234 +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_54 +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_1435/ + transpose_67 +  + +  +� +@*" +name + +" + transpose_67� +matmul +x + +attn_weights_129 +y + + transpose_67+ + transpose_x + +tensor_173_transpose_x_0+ + transpose_y + +tensor_173_transpose_y_0- + +tensor_173 +  + +  +� +@* +name + +" + +tensor_173c +const +var_1466 + + +* +name + + " +op_1466*! +val + + + + + +k +const +var_1478 + + +* +name + + " +op_1478*) +val" + + + + + ����������� + transpose +x + + +tensor_173 +perm + + +var_1466/ + transpose_64 +  + +� +  +@*" +name + +" + transpose_64x +reshape +x + + transpose_64 +shape + + +var_1478 +var_1479 +  +� +�* +name + + " +op_1479� +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_1479B +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_1481 + + +* +name + + " +op_1481*" +val + + + +  +��z +reshape +x + +x_167 +shape + + +var_1481' + 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_1496 + + +* +name + + " +op_1496*) +val" + + + + + ����������q +reshape +x + +x_169 +shape + + +var_1496 +var_1497 +  +� +�* +name + + " +op_1497� +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_1497? +weight5 +3 +1model_transformer_h_10_mlp_c_fc_weight_transposed0 +bias( +& +$model_transformer_h_10_mlp_c_fc_bias +x_171 +  +� +�* +name + + " +x_171d +const +var_1499 + + +* +name + + " +op_1499*" +val + + + +  +��z +reshape +x + +x_171 +shape + + +var_1499' + input_153 +  + +� +�* +name + + " + input_153j +const + x_173_mode_0 +*" +name + +" + x_173_mode_0*' +val  + +" +TANH_APPROXIMATIONv +gelu +x + + input_153 +mode + + x_173_mode_0# +x_173 +  + +� +�* +name + + " +x_173k +const +var_1518 + + +* +name + + " +op_1518*) +val" + + + + + ����������q +reshape +x + +x_173 +shape + + +var_1518 +var_1519 +  +� +�* +name + + " +op_1519� +constK +3model_transformer_h_10_mlp_c_proj_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_10_mlp_c_proj_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_1519A +weight7 +5 +3model_transformer_h_10_mlp_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_10_mlp_c_proj_bias +x_175 +  +� +�* +name + + " +x_175d +const +var_1521 + + +* +name + + " +op_1521*" +val + + + +  +��z +reshape +x + +x_175 +shape + + +var_1521' + input_155 +  + +� +�* +name + + " + input_155w +add +x + + input_151 +y + + input_155' + input_157 +  + +� +�* +name + + " + input_157r +const + x_177_axes_0 + + +*" +name + +" + x_177_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_157 +axes + + x_177_axes_0/ +gamma& +$ +"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_1541 + + +* +name + + " +op_1541*) +val" + + + + + ����������q +reshape +x + +x_177 +shape + + +var_1541 +var_1542 +  +� +�* +name + + " +op_1542� +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_1542B +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_1544 + + +* +name + + " +op_1544*" +val + + + +  +��w +reshape +x + +x_179 +shape + + +var_1544& +var_1545 +  + +� +�* +name + + " +op_1545d +const +tile_11 + + +* +name + + " +tile_11*# +val + + + + + +���^ +const +var_1546_axis_0 +*$ +name + +" +op_1546_axis_0* +val + + +� +split +x + + +var_1545 + split_sizes + +tile_11 +axis + +var_1546_axis_0( + +var_1546_0 +  + +� +�( + +var_1546_1 +  + +� +�( + +var_1546_2 +  + +� +�* +name + + " +op_1546d +const +var_1552 + + +* +name + + " +op_1552*" +val + + + +  +� @� +reshape +x + + +var_1546_0 +shape + + +var_1552- + +tensor_179 +  + +� +  +@* +name + +" + +tensor_179d +const +var_1558 + + +* +name + + " +op_1558*" +val + + + +  +� @� +reshape +x + + +var_1546_1 +shape + + +var_1558- + +tensor_183 +  + +� +  +@* +name + +" + +tensor_183d +const +var_1564 + + +* +name + + " +op_1564*" +val + + + +  +� @� +reshape +x + + +var_1546_2 +shape + + +var_1564- + +tensor_187 +  + +� +  +@* +name + +" + +tensor_187c +const +var_1566 + + +* +name + + " +op_1566*! +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_58_perm_0 + + +*) +name! + +" +transpose_58_perm_0*! +val + + + + + +z +const# +transpose_59_perm_0 + + +*) +name! + +" +transpose_59_perm_0*! +val + + + + + +� + transpose +x + + +tensor_183 +perm + +transpose_59_perm_0/ + transpose_61 +  + +  +@ +�*" +name + +" + transpose_61� + transpose +x + + +tensor_179 +perm + +transpose_58_perm_0/ + transpose_62 +  + +  +� +@*" +name + +" + transpose_62� +matmul +x + + transpose_62 +y + + transpose_611 + 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� +add +x + +attn_weights_7_x_0# +y + +_inversed_attn_weights_1354 +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_54 +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_1566/ + transpose_63 +  + +  +� +@*" +name + +" + transpose_63� +matmul +x + +attn_weights_141 +y + + transpose_63+ + transpose_x + +tensor_189_transpose_x_0+ + transpose_y + +tensor_189_transpose_y_0- + +tensor_189 +  + +  +� +@* +name + +" + +tensor_189c +const +var_1597 + + +* +name + + " +op_1597*! +val + + + + + +k +const +var_1609 + + +* +name + + " +op_1609*) +val" + + + + + ����������� + transpose +x + + +tensor_189 +perm + + +var_1597/ + transpose_60 +  + +� +  +@*" +name + +" + transpose_60x +reshape +x + + transpose_60 +shape + + +var_1609 +var_1610 +  +� +�* +name + + " +op_1610� +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_1610B +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_1612 + + +* +name + + " +op_1612*" +val + + + +  +��z +reshape +x + +x_183 +shape + + +var_1612' + 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_1627 + + +* +name + + " +op_1627*) +val" + + + + + ����������q +reshape +x + +x_185 +shape + + +var_1627 +var_1628 +  +� +�* +name + + " +op_1628� +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_1628? +weight5 +3 +1model_transformer_h_11_mlp_c_fc_weight_transposed0 +bias( +& +$model_transformer_h_11_mlp_c_fc_bias +x_187 +  +� +�* +name + + " +x_187d +const +var_1630 + + +* +name + + " +op_1630*" +val + + + +  +��z +reshape +x + +x_187 +shape + + +var_1630' + input_167 +  + +� +�* +name + + " + input_167j +const + x_189_mode_0 +*" +name + +" + x_189_mode_0*' +val  + +" +TANH_APPROXIMATIONv +gelu +x + + input_167 +mode + + x_189_mode_0# +x_189 +  + +� +�* +name + + " +x_189k +const +var_1649 + + +* +name + + " +op_1649*) +val" + + + + + ����������q +reshape +x + +x_189 +shape + + +var_1649 +var_1650 +  +� +�* +name + + " +op_1650� +constK +3model_transformer_h_11_mlp_c_proj_weight_transposed +  +� +�*I +nameA +9 +7"5 +3model_transformer_h_11_mlp_c_proj_weight_transposed*E +val> +  +� +�*& +@model_path/weights/weight.bin����� +linear +x + + +var_1650A +weight7 +5 +3model_transformer_h_11_mlp_c_proj_weight_transposed2 +bias* +( +&model_transformer_h_11_mlp_c_proj_bias +x +  +� +�* +name + +" +xd +const +var_1652 + + +* +name + + " +op_1652*" +val + + + +  +��v +reshape + +x + +x +shape + + +var_1652' + input_169 +  + +� +�* +name + + " + input_169w +add +x + + input_165 +y + + input_169' + input_171 +  + +� +�* +name + + " + input_171� +const$ +hidden_states_axes_0 + + +** +name" + +" +hidden_states_axes_0*' +val  + + + + + +���������� + +layer_norm +x + + input_171 +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! +linear_0_bias_0 +   +��*% +name + +" +linear_0_bias_0*? +val8 +   +��*& +@model_path/weights/weight.bin����� +linear +x + + hidden_states* +weight + +model_transformer_wte_weight +bias + +linear_0_bias_0% +logits +  + +� +��* +name + + " + +linear_0"� + buildInfo�" + + +�"� +6 +! + +" +coremltools-version + +" +7.1 +@ +) +! +" +coremltools-component-torch + + " +2.1.0 +E +( + +" +coremltools-source-dialect + +" + TorchScript \ No newline at end of file