<
   ir_version: 10,
   opset_import: ["" : 23],
   producer_name: "pytorch"
>
main_graph (int32[batch,77] text) => (float[batch,1024] text_embedding) 
   <
      float[batch,77,1024] add_1071
      float[batch,77,1024] add_1092
      float[batch,77,1024] add_119
      float[batch,77,1024] add_1207
      float[batch,77,1024] add_1228
      float[batch,77,1024] add_1343
      float[batch,77,1024] add_1364
      float[batch,77,1024] add_140
      float[batch,77,1024] add_1479
      float[batch,77,1024] add_1500
      float[batch,77,1024] add_1615
      float[batch,77,1024] add_1636
      float[batch,77,1024] add_1751
      float[batch,77,1024] add_1772
      float[batch,77,1024] add_1887
      float[batch,77,1024] add_1908
      float[batch,77,1024] add_2023
      float[batch,77,1024] add_2044
      float[batch,77,1024] add_2159
      float[batch,77,1024] add_2180
      float[batch,77,1024] add_2295
      float[batch,77,1024] add_2316
      float[batch,77,1024] add_2431
      float[batch,77,1024] add_2452
      float[batch,77,1024] add_255
      float[batch,77,1024] add_2567
      float[batch,77,1024] add_2588
      float[batch,77,1024] add_2703
      float[batch,77,1024] add_2724
      float[batch,77,1024] add_276
      float[batch,77,1024] add_2839
      float[batch,77,1024] add_2860
      float[batch,77,1024] add_2975
      float[batch,77,1024] add_2996
      float[batch,77,1024] add_3111
      float[batch,77,1024] add_3132
      float[batch,1,1024] add_3132_pooled
      float[batch,1,1024] add_3247
      float[batch,1,1024] add_3268
      float[batch,77,1024] add_391
      float[batch,77,1024] add_4
      float[batch,77,1024] add_412
      float[batch,77,1024] add_527
      float[batch,77,1024] add_548
      float[batch,77,1024] add_663
      float[batch,77,1024] add_684
      float[batch,77,1024] add_799
      float[batch,77,1024] add_820
      float[batch,77,1024] add_935
      float[batch,77,1024] add_956
      int64[batch] argmax
      float[batch,1] clamp_min
      float[batch,77,1024] embedding
      float[batch,77,4096] gelu
      float[batch,77,4096] gelu_1
      float[batch,77,4096] gelu_10
      float[batch,77,4096] gelu_11
      float[batch,77,4096] gelu_12
      float[batch,77,4096] gelu_13
      float[batch,77,4096] gelu_14
      float[batch,77,4096] gelu_15
      float[batch,77,4096] gelu_16
      float[batch,77,4096] gelu_17
      float[batch,77,4096] gelu_18
      float[batch,77,4096] gelu_19
      float[batch,77,4096] gelu_2
      float[batch,77,4096] gelu_20
      float[batch,77,4096] gelu_21
      float[batch,77,4096] gelu_22
      float[batch,1,4096] gelu_23
      float[batch,77,4096] gelu_3
      float[batch,77,4096] gelu_4
      float[batch,77,4096] gelu_5
      float[batch,77,4096] gelu_6
      float[batch,77,4096] gelu_7
      float[batch,77,4096] gelu_8
      float[batch,77,4096] gelu_9
      float[batch,77,1024] layer_norm
      float[batch,77,1024] layer_norm_1
      float[batch,77,1024] layer_norm_10
      float[batch,77,1024] layer_norm_11
      float[batch,77,1024] layer_norm_12
      float[batch,77,1024] layer_norm_13
      float[batch,77,1024] layer_norm_14
      float[batch,77,1024] layer_norm_15
      float[batch,77,1024] layer_norm_16
      float[batch,77,1024] layer_norm_17
      float[batch,77,1024] layer_norm_18
      float[batch,77,1024] layer_norm_19
      float[batch,77,1024] layer_norm_2
      float[batch,77,1024] layer_norm_20
      float[batch,77,1024] layer_norm_21
      float[batch,77,1024] layer_norm_22
      float[batch,77,1024] layer_norm_23
      float[batch,77,1024] layer_norm_24
      float[batch,77,1024] layer_norm_25
      float[batch,77,1024] layer_norm_26
      float[batch,77,1024] layer_norm_27
      float[batch,77,1024] layer_norm_28
      float[batch,77,1024] layer_norm_29
      float[batch,77,1024] layer_norm_3
      float[batch,77,1024] layer_norm_30
      float[batch,77,1024] layer_norm_31
      float[batch,77,1024] layer_norm_32
      float[batch,77,1024] layer_norm_33
      float[batch,77,1024] layer_norm_34
      float[batch,77,1024] layer_norm_35
      float[batch,77,1024] layer_norm_36
      float[batch,77,1024] layer_norm_37
      float[batch,77,1024] layer_norm_38
      float[batch,77,1024] layer_norm_39
      float[batch,77,1024] layer_norm_4
      float[batch,77,1024] layer_norm_40
      float[batch,77,1024] layer_norm_41
      float[batch,77,1024] layer_norm_42
      float[batch,77,1024] layer_norm_43
      float[batch,77,1024] layer_norm_44
      float[batch,77,1024] layer_norm_45
      float[batch,77,1024] layer_norm_46
      float[batch,1,1024] layer_norm_47
      float[batch,77,1024] layer_norm_5
      float[batch,77,1024] layer_norm_6
      float[batch,77,1024] layer_norm_7
      float[batch,77,1024] layer_norm_8
      float[batch,77,1024] layer_norm_9
      float[batch,1] linalg_vector_norm
      float[batch,77,4096] linear_10
      float[batch,77,1024] linear_11
      float[batch,77,4096] linear_14
      float[batch,77,1024] linear_15
      float[batch,77,4096] linear_18
      float[batch,77,1024] linear_19
      float[batch,77,4096] linear_2
      float[batch,77,4096] linear_22
      float[batch,77,1024] linear_23
      float[batch,77,4096] linear_26
      float[batch,77,1024] linear_27
      float[batch,77,1024] linear_3
      float[batch,77,4096] linear_30
      float[batch,77,1024] linear_31
      float[batch,77,4096] linear_34
      float[batch,77,1024] linear_35
      float[batch,77,4096] linear_38
      float[batch,77,1024] linear_39
      float[batch,77,4096] linear_42
      float[batch,77,1024] linear_43
      float[batch,77,4096] linear_46
      float[batch,77,1024] linear_47
      float[batch,77,4096] linear_50
      float[batch,77,1024] linear_51
      float[batch,77,4096] linear_54
      float[batch,77,1024] linear_55
      float[batch,77,4096] linear_58
      float[batch,77,1024] linear_59
      float[batch,77,4096] linear_6
      float[batch,77,4096] linear_62
      float[batch,77,1024] linear_63
      float[batch,77,4096] linear_66
      float[batch,77,1024] linear_67
      float[batch,77,1024] linear_7
      float[batch,77,4096] linear_70
      float[batch,77,1024] linear_71
      float[batch,77,4096] linear_74
      float[batch,77,1024] linear_75
      float[batch,77,4096] linear_78
      float[batch,77,1024] linear_79
      float[batch,77,4096] linear_82
      float[batch,77,1024] linear_83
      float[batch,77,4096] linear_86
      float[batch,77,1024] linear_87
      float[batch,77,4096] linear_90
      float[batch,77,1024] linear_91
      float[batch,1,4096] linear_94
      float[batch,1,1024] linear_95
      float[batch,1024] matmul
      float[batch,77,1024] node_scaled_dot_product_attention_10_k
      float[batch,77,1024] node_scaled_dot_product_attention_10_out
      float[batch,77,1024] node_scaled_dot_product_attention_10_out_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_10_q
      float[batch,77,3072] node_scaled_dot_product_attention_10_qkv
      float[batch,77,3072] node_scaled_dot_product_attention_10_qkv_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_10_v
      float[batch,77,1024] node_scaled_dot_product_attention_11_k
      float[batch,77,1024] node_scaled_dot_product_attention_11_out
      float[batch,77,1024] node_scaled_dot_product_attention_11_out_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_11_q
      float[batch,77,3072] node_scaled_dot_product_attention_11_qkv
      float[batch,77,3072] node_scaled_dot_product_attention_11_qkv_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_11_v
      float[batch,77,1024] node_scaled_dot_product_attention_12_k
      float[batch,77,1024] node_scaled_dot_product_attention_12_out
      float[batch,77,1024] node_scaled_dot_product_attention_12_out_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_12_q
      float[batch,77,3072] node_scaled_dot_product_attention_12_qkv
      float[batch,77,3072] node_scaled_dot_product_attention_12_qkv_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_12_v
      float[batch,77,1024] node_scaled_dot_product_attention_13_k
      float[batch,77,1024] node_scaled_dot_product_attention_13_out
      float[batch,77,1024] node_scaled_dot_product_attention_13_out_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_13_q
      float[batch,77,3072] node_scaled_dot_product_attention_13_qkv
      float[batch,77,3072] node_scaled_dot_product_attention_13_qkv_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_13_v
      float[batch,77,1024] node_scaled_dot_product_attention_14_k
      float[batch,77,1024] node_scaled_dot_product_attention_14_out
      float[batch,77,1024] node_scaled_dot_product_attention_14_out_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_14_q
      float[batch,77,3072] node_scaled_dot_product_attention_14_qkv
      float[batch,77,3072] node_scaled_dot_product_attention_14_qkv_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_14_v
      float[batch,77,1024] node_scaled_dot_product_attention_15_k
      float[batch,77,1024] node_scaled_dot_product_attention_15_out
      float[batch,77,1024] node_scaled_dot_product_attention_15_out_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_15_q
      float[batch,77,3072] node_scaled_dot_product_attention_15_qkv
      float[batch,77,3072] node_scaled_dot_product_attention_15_qkv_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_15_v
      float[batch,77,1024] node_scaled_dot_product_attention_16_k
      float[batch,77,1024] node_scaled_dot_product_attention_16_out
      float[batch,77,1024] node_scaled_dot_product_attention_16_out_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_16_q
      float[batch,77,3072] node_scaled_dot_product_attention_16_qkv
      float[batch,77,3072] node_scaled_dot_product_attention_16_qkv_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_16_v
      float[batch,77,1024] node_scaled_dot_product_attention_17_k
      float[batch,77,1024] node_scaled_dot_product_attention_17_out
      float[batch,77,1024] node_scaled_dot_product_attention_17_out_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_17_q
      float[batch,77,3072] node_scaled_dot_product_attention_17_qkv
      float[batch,77,3072] node_scaled_dot_product_attention_17_qkv_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_17_v
      float[batch,77,1024] node_scaled_dot_product_attention_18_k
      float[batch,77,1024] node_scaled_dot_product_attention_18_out
      float[batch,77,1024] node_scaled_dot_product_attention_18_out_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_18_q
      float[batch,77,3072] node_scaled_dot_product_attention_18_qkv
      float[batch,77,3072] node_scaled_dot_product_attention_18_qkv_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_18_v
      float[batch,77,1024] node_scaled_dot_product_attention_19_k
      float[batch,77,1024] node_scaled_dot_product_attention_19_out
      float[batch,77,1024] node_scaled_dot_product_attention_19_out_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_19_q
      float[batch,77,3072] node_scaled_dot_product_attention_19_qkv
      float[batch,77,3072] node_scaled_dot_product_attention_19_qkv_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_19_v
      float[batch,77,1024] node_scaled_dot_product_attention_1_k
      float[batch,77,1024] node_scaled_dot_product_attention_1_out
      float[batch,77,1024] node_scaled_dot_product_attention_1_out_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_1_q
      float[batch,77,3072] node_scaled_dot_product_attention_1_qkv
      float[batch,77,3072] node_scaled_dot_product_attention_1_qkv_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_1_v
      float[batch,77,1024] node_scaled_dot_product_attention_20_k
      float[batch,77,1024] node_scaled_dot_product_attention_20_out
      float[batch,77,1024] node_scaled_dot_product_attention_20_out_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_20_q
      float[batch,77,3072] node_scaled_dot_product_attention_20_qkv
      float[batch,77,3072] node_scaled_dot_product_attention_20_qkv_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_20_v
      float[batch,77,1024] node_scaled_dot_product_attention_21_k
      float[batch,77,1024] node_scaled_dot_product_attention_21_out
      float[batch,77,1024] node_scaled_dot_product_attention_21_out_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_21_q
      float[batch,77,3072] node_scaled_dot_product_attention_21_qkv
      float[batch,77,3072] node_scaled_dot_product_attention_21_qkv_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_21_v
      float[batch,77,1024] node_scaled_dot_product_attention_22_k
      float[batch,77,1024] node_scaled_dot_product_attention_22_out
      float[batch,77,1024] node_scaled_dot_product_attention_22_out_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_22_q
      float[batch,77,3072] node_scaled_dot_product_attention_22_qkv
      float[batch,77,3072] node_scaled_dot_product_attention_22_qkv_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_22_v
      float[batch,77,1024] node_scaled_dot_product_attention_23_k
      float[batch,1,1024] node_scaled_dot_product_attention_23_out
      float[batch,1,1024] node_scaled_dot_product_attention_23_out_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_23_q
      float[batch,77,3072] node_scaled_dot_product_attention_23_qkv
      float[batch,77,3072] node_scaled_dot_product_attention_23_qkv_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_23_v
      float[batch,77,1024] node_scaled_dot_product_attention_2_k
      float[batch,77,1024] node_scaled_dot_product_attention_2_out
      float[batch,77,1024] node_scaled_dot_product_attention_2_out_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_2_q
      float[batch,77,3072] node_scaled_dot_product_attention_2_qkv
      float[batch,77,3072] node_scaled_dot_product_attention_2_qkv_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_2_v
      float[batch,77,1024] node_scaled_dot_product_attention_3_k
      float[batch,77,1024] node_scaled_dot_product_attention_3_out
      float[batch,77,1024] node_scaled_dot_product_attention_3_out_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_3_q
      float[batch,77,3072] node_scaled_dot_product_attention_3_qkv
      float[batch,77,3072] node_scaled_dot_product_attention_3_qkv_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_3_v
      float[batch,77,1024] node_scaled_dot_product_attention_4_k
      float[batch,77,1024] node_scaled_dot_product_attention_4_out
      float[batch,77,1024] node_scaled_dot_product_attention_4_out_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_4_q
      float[batch,77,3072] node_scaled_dot_product_attention_4_qkv
      float[batch,77,3072] node_scaled_dot_product_attention_4_qkv_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_4_v
      float[batch,77,1024] node_scaled_dot_product_attention_5_k
      float[batch,77,1024] node_scaled_dot_product_attention_5_out
      float[batch,77,1024] node_scaled_dot_product_attention_5_out_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_5_q
      float[batch,77,3072] node_scaled_dot_product_attention_5_qkv
      float[batch,77,3072] node_scaled_dot_product_attention_5_qkv_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_5_v
      float[batch,77,1024] node_scaled_dot_product_attention_6_k
      float[batch,77,1024] node_scaled_dot_product_attention_6_out
      float[batch,77,1024] node_scaled_dot_product_attention_6_out_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_6_q
      float[batch,77,3072] node_scaled_dot_product_attention_6_qkv
      float[batch,77,3072] node_scaled_dot_product_attention_6_qkv_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_6_v
      float[batch,77,1024] node_scaled_dot_product_attention_7_k
      float[batch,77,1024] node_scaled_dot_product_attention_7_out
      float[batch,77,1024] node_scaled_dot_product_attention_7_out_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_7_q
      float[batch,77,3072] node_scaled_dot_product_attention_7_qkv
      float[batch,77,3072] node_scaled_dot_product_attention_7_qkv_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_7_v
      float[batch,77,1024] node_scaled_dot_product_attention_8_k
      float[batch,77,1024] node_scaled_dot_product_attention_8_out
      float[batch,77,1024] node_scaled_dot_product_attention_8_out_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_8_q
      float[batch,77,3072] node_scaled_dot_product_attention_8_qkv
      float[batch,77,3072] node_scaled_dot_product_attention_8_qkv_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_8_v
      float[batch,77,1024] node_scaled_dot_product_attention_9_k
      float[batch,77,1024] node_scaled_dot_product_attention_9_out
      float[batch,77,1024] node_scaled_dot_product_attention_9_out_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_9_q
      float[batch,77,3072] node_scaled_dot_product_attention_9_qkv
      float[batch,77,3072] node_scaled_dot_product_attention_9_qkv_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_9_v
      float[batch,77,1024] node_scaled_dot_product_attention_k
      float[batch,77,1024] node_scaled_dot_product_attention_out
      float[batch,77,1024] node_scaled_dot_product_attention_out_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_q
      float[batch,77,3072] node_scaled_dot_product_attention_qkv
      float[batch,77,3072] node_scaled_dot_product_attention_qkv_mm_out
      float[batch,77,1024] node_scaled_dot_product_attention_v
      float[batch,77,1024] scaled_dot_product_attention
      float[batch,77,1024] scaled_dot_product_attention_1
      float[batch,77,1024] scaled_dot_product_attention_10
      float[batch,77,1024] scaled_dot_product_attention_11
      float[batch,77,1024] scaled_dot_product_attention_12
      float[batch,77,1024] scaled_dot_product_attention_13
      float[batch,77,1024] scaled_dot_product_attention_14
      float[batch,77,1024] scaled_dot_product_attention_15
      float[batch,77,1024] scaled_dot_product_attention_16
      float[batch,77,1024] scaled_dot_product_attention_17
      float[batch,77,1024] scaled_dot_product_attention_18
      float[batch,77,1024] scaled_dot_product_attention_19
      float[batch,77,1024] scaled_dot_product_attention_2
      float[batch,77,1024] scaled_dot_product_attention_20
      float[batch,77,1024] scaled_dot_product_attention_21
      float[batch,77,1024] scaled_dot_product_attention_22
      float[batch,77,1024] scaled_dot_product_attention_23
      float[batch,1,1024] scaled_dot_product_attention_23_pooled
      float[batch,77,1024] scaled_dot_product_attention_3
      float[batch,77,1024] scaled_dot_product_attention_4
      float[batch,77,1024] scaled_dot_product_attention_5
      float[batch,77,1024] scaled_dot_product_attention_6
      float[batch,77,1024] scaled_dot_product_attention_7
      float[batch,77,1024] scaled_dot_product_attention_8
      float[batch,77,1024] scaled_dot_product_attention_9
      float[batch,77,1024] val_100
      float[batch,77,4096] val_101
      float[batch,77,1024] val_102
      float[batch,77,4096] val_103
      float[batch,77,1024] val_104
      float[batch,77,4096] val_105
      float[batch,77,1024] val_106
      float[batch,77,4096] val_107
      float[batch,77,1024] val_108
      float[batch,77,4096] val_109
      float[batch,77,1024] val_110
      float[batch,77,4096] val_111
      float[batch,77,1024] val_112
      float[batch,77,4096] val_113
      float[batch,77,1024] val_114
      float[batch,77,4096] val_115
      float[batch,77,1024] val_116
      float[batch,77,4096] val_117
      float[batch,77,1024] val_118
      float[batch,77,4096] val_119
      float[batch,77,1024] val_120
      float[batch,77] val_121
      float[batch,1,77] val_122
      float[batch,1,4096] val_123
      float[batch,1,1024] val_124
      float[batch,1,1024] val_125
      float[batch,1024] val_126
      float[batch,77,4096] val_75
      float[batch,77,1024] val_76
      float[batch,77,4096] val_77
      float[batch,77,1024] val_78
      float[batch,77,4096] val_79
      float[batch,77,1024] val_80
      float[batch,77,4096] val_81
      float[batch,77,1024] val_82
      float[batch,77,4096] val_83
      float[batch,77,1024] val_84
      float[batch,77,4096] val_85
      float[batch,77,1024] val_86
      float[batch,77,4096] val_87
      float[batch,77,1024] val_88
      float[batch,77,4096] val_89
      float[batch,77,1024] val_90
      float[batch,77,4096] val_91
      float[batch,77,1024] val_92
      float[batch,77,4096] val_93
      float[batch,77,1024] val_94
      float[batch,77,4096] val_95
      float[batch,77,1024] val_96
      float[batch,77,4096] val_97
      float[batch,77,1024] val_98
      float[batch,77,4096] val_99
   >
{
   val_74 = Gather <axis: int = 0> ("token_embedding.weight_fp16", text)
   embedding = Cast <to: int = 1> (val_74)
   add_4 = Add (embedding, positional_embedding)
   [node_layer_norm] layer_norm = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_4, "transformer.resblocks.0.ln_1.weight", "transformer.resblocks.0.ln_1.bias")
   [node_scaled_dot_product_attention_qkv_mm] node_scaled_dot_product_attention_qkv_mm_out = MatMul (layer_norm, val_2)
   [node_scaled_dot_product_attention_qkv_bias] node_scaled_dot_product_attention_qkv = Add (node_scaled_dot_product_attention_qkv_mm_out, "transformer.resblocks.0.attn.in_proj_bias")
   [node_scaled_dot_product_attention_qkv_split] node_scaled_dot_product_attention_q, node_scaled_dot_product_attention_k, node_scaled_dot_product_attention_v = Split <axis: int = -1> (node_scaled_dot_product_attention_qkv, attn3d_split_3x1024)
   scaled_dot_product_attention = Attention <is_causal: int = 1, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_q, node_scaled_dot_product_attention_k, node_scaled_dot_product_attention_v)
   [node_scaled_dot_product_attention_out_mm] node_scaled_dot_product_attention_out_mm_out = MatMul (scaled_dot_product_attention, node_scaled_dot_product_attention_wo_t)
   [node_scaled_dot_product_attention_out_bias] node_scaled_dot_product_attention_out = Add (node_scaled_dot_product_attention_out_mm_out, "transformer.resblocks.0.attn.out_proj.bias")
   add_119 = Add (add_4, node_scaled_dot_product_attention_out)
   layer_norm_1 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_119, "transformer.resblocks.0.ln_2.weight", "transformer.resblocks.0.ln_2.bias")
   val_75 = MatMul (layer_norm_1, val_3)
   linear_2 = Add (val_75, "transformer.resblocks.0.mlp.c_fc.bias")
   [node_gelu] gelu = Gelu <approximate: string = "none"> (linear_2)
   val_76 = MatMul (gelu, val_4)
   linear_3 = Add (val_76, "transformer.resblocks.0.mlp.c_proj.bias")
   add_140 = Add (add_119, linear_3)
   layer_norm_2 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_140, "transformer.resblocks.1.ln_1.weight", "transformer.resblocks.1.ln_1.bias")
   [node_scaled_dot_product_attention_1_qkv_mm] node_scaled_dot_product_attention_1_qkv_mm_out = MatMul (layer_norm_2, val_5)
   [node_scaled_dot_product_attention_1_qkv_bias] node_scaled_dot_product_attention_1_qkv = Add (node_scaled_dot_product_attention_1_qkv_mm_out, "transformer.resblocks.1.attn.in_proj_bias")
   [node_scaled_dot_product_attention_1_qkv_split] node_scaled_dot_product_attention_1_q, node_scaled_dot_product_attention_1_k, node_scaled_dot_product_attention_1_v = Split <axis: int = -1> (node_scaled_dot_product_attention_1_qkv, attn3d_split_3x1024)
   scaled_dot_product_attention_1 = Attention <is_causal: int = 1, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_1_q, node_scaled_dot_product_attention_1_k, node_scaled_dot_product_attention_1_v)
   [node_scaled_dot_product_attention_1_out_mm] node_scaled_dot_product_attention_1_out_mm_out = MatMul (scaled_dot_product_attention_1, node_scaled_dot_product_attention_1_wo_t)
   [node_scaled_dot_product_attention_1_out_bias] node_scaled_dot_product_attention_1_out = Add (node_scaled_dot_product_attention_1_out_mm_out, "transformer.resblocks.1.attn.out_proj.bias")
   add_255 = Add (add_140, node_scaled_dot_product_attention_1_out)
   layer_norm_3 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_255, "transformer.resblocks.1.ln_2.weight", "transformer.resblocks.1.ln_2.bias")
   val_77 = MatMul (layer_norm_3, val_6)
   linear_6 = Add (val_77, "transformer.resblocks.1.mlp.c_fc.bias")
   gelu_1 = Gelu <approximate: string = "none"> (linear_6)
   val_78 = MatMul (gelu_1, val_7)
   linear_7 = Add (val_78, "transformer.resblocks.1.mlp.c_proj.bias")
   add_276 = Add (add_255, linear_7)
   layer_norm_4 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_276, "transformer.resblocks.2.ln_1.weight", "transformer.resblocks.2.ln_1.bias")
   [node_scaled_dot_product_attention_2_qkv_mm] node_scaled_dot_product_attention_2_qkv_mm_out = MatMul (layer_norm_4, val_8)
   [node_scaled_dot_product_attention_2_qkv_bias] node_scaled_dot_product_attention_2_qkv = Add (node_scaled_dot_product_attention_2_qkv_mm_out, "transformer.resblocks.2.attn.in_proj_bias")
   [node_scaled_dot_product_attention_2_qkv_split] node_scaled_dot_product_attention_2_q, node_scaled_dot_product_attention_2_k, node_scaled_dot_product_attention_2_v = Split <axis: int = -1> (node_scaled_dot_product_attention_2_qkv, attn3d_split_3x1024)
   scaled_dot_product_attention_2 = Attention <is_causal: int = 1, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_2_q, node_scaled_dot_product_attention_2_k, node_scaled_dot_product_attention_2_v)
   [node_scaled_dot_product_attention_2_out_mm] node_scaled_dot_product_attention_2_out_mm_out = MatMul (scaled_dot_product_attention_2, node_scaled_dot_product_attention_2_wo_t)
   [node_scaled_dot_product_attention_2_out_bias] node_scaled_dot_product_attention_2_out = Add (node_scaled_dot_product_attention_2_out_mm_out, "transformer.resblocks.2.attn.out_proj.bias")
   add_391 = Add (add_276, node_scaled_dot_product_attention_2_out)
   layer_norm_5 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_391, "transformer.resblocks.2.ln_2.weight", "transformer.resblocks.2.ln_2.bias")
   val_79 = MatMul (layer_norm_5, val_9)
   linear_10 = Add (val_79, "transformer.resblocks.2.mlp.c_fc.bias")
   gelu_2 = Gelu <approximate: string = "none"> (linear_10)
   val_80 = MatMul (gelu_2, val_10)
   linear_11 = Add (val_80, "transformer.resblocks.2.mlp.c_proj.bias")
   add_412 = Add (add_391, linear_11)
   layer_norm_6 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_412, "transformer.resblocks.3.ln_1.weight", "transformer.resblocks.3.ln_1.bias")
   [node_scaled_dot_product_attention_3_qkv_mm] node_scaled_dot_product_attention_3_qkv_mm_out = MatMul (layer_norm_6, val_11)
   [node_scaled_dot_product_attention_3_qkv_bias] node_scaled_dot_product_attention_3_qkv = Add (node_scaled_dot_product_attention_3_qkv_mm_out, "transformer.resblocks.3.attn.in_proj_bias")
   [node_scaled_dot_product_attention_3_qkv_split] node_scaled_dot_product_attention_3_q, node_scaled_dot_product_attention_3_k, node_scaled_dot_product_attention_3_v = Split <axis: int = -1> (node_scaled_dot_product_attention_3_qkv, attn3d_split_3x1024)
   scaled_dot_product_attention_3 = Attention <is_causal: int = 1, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_3_q, node_scaled_dot_product_attention_3_k, node_scaled_dot_product_attention_3_v)
   [node_scaled_dot_product_attention_3_out_mm] node_scaled_dot_product_attention_3_out_mm_out = MatMul (scaled_dot_product_attention_3, node_scaled_dot_product_attention_3_wo_t)
   [node_scaled_dot_product_attention_3_out_bias] node_scaled_dot_product_attention_3_out = Add (node_scaled_dot_product_attention_3_out_mm_out, "transformer.resblocks.3.attn.out_proj.bias")
   add_527 = Add (add_412, node_scaled_dot_product_attention_3_out)
   layer_norm_7 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_527, "transformer.resblocks.3.ln_2.weight", "transformer.resblocks.3.ln_2.bias")
   val_81 = MatMul (layer_norm_7, val_12)
   linear_14 = Add (val_81, "transformer.resblocks.3.mlp.c_fc.bias")
   gelu_3 = Gelu <approximate: string = "none"> (linear_14)
   val_82 = MatMul (gelu_3, val_13)
   linear_15 = Add (val_82, "transformer.resblocks.3.mlp.c_proj.bias")
   add_548 = Add (add_527, linear_15)
   layer_norm_8 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_548, "transformer.resblocks.4.ln_1.weight", "transformer.resblocks.4.ln_1.bias")
   [node_scaled_dot_product_attention_4_qkv_mm] node_scaled_dot_product_attention_4_qkv_mm_out = MatMul (layer_norm_8, val_14)
   [node_scaled_dot_product_attention_4_qkv_bias] node_scaled_dot_product_attention_4_qkv = Add (node_scaled_dot_product_attention_4_qkv_mm_out, "transformer.resblocks.4.attn.in_proj_bias")
   [node_scaled_dot_product_attention_4_qkv_split] node_scaled_dot_product_attention_4_q, node_scaled_dot_product_attention_4_k, node_scaled_dot_product_attention_4_v = Split <axis: int = -1> (node_scaled_dot_product_attention_4_qkv, attn3d_split_3x1024)
   scaled_dot_product_attention_4 = Attention <is_causal: int = 1, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_4_q, node_scaled_dot_product_attention_4_k, node_scaled_dot_product_attention_4_v)
   [node_scaled_dot_product_attention_4_out_mm] node_scaled_dot_product_attention_4_out_mm_out = MatMul (scaled_dot_product_attention_4, node_scaled_dot_product_attention_4_wo_t)
   [node_scaled_dot_product_attention_4_out_bias] node_scaled_dot_product_attention_4_out = Add (node_scaled_dot_product_attention_4_out_mm_out, "transformer.resblocks.4.attn.out_proj.bias")
   add_663 = Add (add_548, node_scaled_dot_product_attention_4_out)
   layer_norm_9 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_663, "transformer.resblocks.4.ln_2.weight", "transformer.resblocks.4.ln_2.bias")
   val_83 = MatMul (layer_norm_9, val_15)
   linear_18 = Add (val_83, "transformer.resblocks.4.mlp.c_fc.bias")
   gelu_4 = Gelu <approximate: string = "none"> (linear_18)
   val_84 = MatMul (gelu_4, val_16)
   linear_19 = Add (val_84, "transformer.resblocks.4.mlp.c_proj.bias")
   add_684 = Add (add_663, linear_19)
   layer_norm_10 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_684, "transformer.resblocks.5.ln_1.weight", "transformer.resblocks.5.ln_1.bias")
   [node_scaled_dot_product_attention_5_qkv_mm] node_scaled_dot_product_attention_5_qkv_mm_out = MatMul (layer_norm_10, val_17)
   [node_scaled_dot_product_attention_5_qkv_bias] node_scaled_dot_product_attention_5_qkv = Add (node_scaled_dot_product_attention_5_qkv_mm_out, "transformer.resblocks.5.attn.in_proj_bias")
   [node_scaled_dot_product_attention_5_qkv_split] node_scaled_dot_product_attention_5_q, node_scaled_dot_product_attention_5_k, node_scaled_dot_product_attention_5_v = Split <axis: int = -1> (node_scaled_dot_product_attention_5_qkv, attn3d_split_3x1024)
   scaled_dot_product_attention_5 = Attention <is_causal: int = 1, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_5_q, node_scaled_dot_product_attention_5_k, node_scaled_dot_product_attention_5_v)
   [node_scaled_dot_product_attention_5_out_mm] node_scaled_dot_product_attention_5_out_mm_out = MatMul (scaled_dot_product_attention_5, node_scaled_dot_product_attention_5_wo_t)
   [node_scaled_dot_product_attention_5_out_bias] node_scaled_dot_product_attention_5_out = Add (node_scaled_dot_product_attention_5_out_mm_out, "transformer.resblocks.5.attn.out_proj.bias")
   add_799 = Add (add_684, node_scaled_dot_product_attention_5_out)
   layer_norm_11 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_799, "transformer.resblocks.5.ln_2.weight", "transformer.resblocks.5.ln_2.bias")
   val_85 = MatMul (layer_norm_11, val_18)
   linear_22 = Add (val_85, "transformer.resblocks.5.mlp.c_fc.bias")
   gelu_5 = Gelu <approximate: string = "none"> (linear_22)
   val_86 = MatMul (gelu_5, val_19)
   linear_23 = Add (val_86, "transformer.resblocks.5.mlp.c_proj.bias")
   add_820 = Add (add_799, linear_23)
   layer_norm_12 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_820, "transformer.resblocks.6.ln_1.weight", "transformer.resblocks.6.ln_1.bias")
   [node_scaled_dot_product_attention_6_qkv_mm] node_scaled_dot_product_attention_6_qkv_mm_out = MatMul (layer_norm_12, val_20)
   [node_scaled_dot_product_attention_6_qkv_bias] node_scaled_dot_product_attention_6_qkv = Add (node_scaled_dot_product_attention_6_qkv_mm_out, "transformer.resblocks.6.attn.in_proj_bias")
   [node_scaled_dot_product_attention_6_qkv_split] node_scaled_dot_product_attention_6_q, node_scaled_dot_product_attention_6_k, node_scaled_dot_product_attention_6_v = Split <axis: int = -1> (node_scaled_dot_product_attention_6_qkv, attn3d_split_3x1024)
   scaled_dot_product_attention_6 = Attention <is_causal: int = 1, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_6_q, node_scaled_dot_product_attention_6_k, node_scaled_dot_product_attention_6_v)
   [node_scaled_dot_product_attention_6_out_mm] node_scaled_dot_product_attention_6_out_mm_out = MatMul (scaled_dot_product_attention_6, node_scaled_dot_product_attention_6_wo_t)
   [node_scaled_dot_product_attention_6_out_bias] node_scaled_dot_product_attention_6_out = Add (node_scaled_dot_product_attention_6_out_mm_out, "transformer.resblocks.6.attn.out_proj.bias")
   add_935 = Add (add_820, node_scaled_dot_product_attention_6_out)
   layer_norm_13 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_935, "transformer.resblocks.6.ln_2.weight", "transformer.resblocks.6.ln_2.bias")
   val_87 = MatMul (layer_norm_13, val_21)
   linear_26 = Add (val_87, "transformer.resblocks.6.mlp.c_fc.bias")
   gelu_6 = Gelu <approximate: string = "none"> (linear_26)
   val_88 = MatMul (gelu_6, val_22)
   linear_27 = Add (val_88, "transformer.resblocks.6.mlp.c_proj.bias")
   add_956 = Add (add_935, linear_27)
   layer_norm_14 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_956, "transformer.resblocks.7.ln_1.weight", "transformer.resblocks.7.ln_1.bias")
   [node_scaled_dot_product_attention_7_qkv_mm] node_scaled_dot_product_attention_7_qkv_mm_out = MatMul (layer_norm_14, val_23)
   [node_scaled_dot_product_attention_7_qkv_bias] node_scaled_dot_product_attention_7_qkv = Add (node_scaled_dot_product_attention_7_qkv_mm_out, "transformer.resblocks.7.attn.in_proj_bias")
   [node_scaled_dot_product_attention_7_qkv_split] node_scaled_dot_product_attention_7_q, node_scaled_dot_product_attention_7_k, node_scaled_dot_product_attention_7_v = Split <axis: int = -1> (node_scaled_dot_product_attention_7_qkv, attn3d_split_3x1024)
   scaled_dot_product_attention_7 = Attention <is_causal: int = 1, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_7_q, node_scaled_dot_product_attention_7_k, node_scaled_dot_product_attention_7_v)
   [node_scaled_dot_product_attention_7_out_mm] node_scaled_dot_product_attention_7_out_mm_out = MatMul (scaled_dot_product_attention_7, node_scaled_dot_product_attention_7_wo_t)
   [node_scaled_dot_product_attention_7_out_bias] node_scaled_dot_product_attention_7_out = Add (node_scaled_dot_product_attention_7_out_mm_out, "transformer.resblocks.7.attn.out_proj.bias")
   add_1071 = Add (add_956, node_scaled_dot_product_attention_7_out)
   layer_norm_15 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1071, "transformer.resblocks.7.ln_2.weight", "transformer.resblocks.7.ln_2.bias")
   val_89 = MatMul (layer_norm_15, val_24)
   linear_30 = Add (val_89, "transformer.resblocks.7.mlp.c_fc.bias")
   gelu_7 = Gelu <approximate: string = "none"> (linear_30)
   val_90 = MatMul (gelu_7, val_25)
   linear_31 = Add (val_90, "transformer.resblocks.7.mlp.c_proj.bias")
   add_1092 = Add (add_1071, linear_31)
   layer_norm_16 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1092, "transformer.resblocks.8.ln_1.weight", "transformer.resblocks.8.ln_1.bias")
   [node_scaled_dot_product_attention_8_qkv_mm] node_scaled_dot_product_attention_8_qkv_mm_out = MatMul (layer_norm_16, val_26)
   [node_scaled_dot_product_attention_8_qkv_bias] node_scaled_dot_product_attention_8_qkv = Add (node_scaled_dot_product_attention_8_qkv_mm_out, "transformer.resblocks.8.attn.in_proj_bias")
   [node_scaled_dot_product_attention_8_qkv_split] node_scaled_dot_product_attention_8_q, node_scaled_dot_product_attention_8_k, node_scaled_dot_product_attention_8_v = Split <axis: int = -1> (node_scaled_dot_product_attention_8_qkv, attn3d_split_3x1024)
   scaled_dot_product_attention_8 = Attention <is_causal: int = 1, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_8_q, node_scaled_dot_product_attention_8_k, node_scaled_dot_product_attention_8_v)
   [node_scaled_dot_product_attention_8_out_mm] node_scaled_dot_product_attention_8_out_mm_out = MatMul (scaled_dot_product_attention_8, node_scaled_dot_product_attention_8_wo_t)
   [node_scaled_dot_product_attention_8_out_bias] node_scaled_dot_product_attention_8_out = Add (node_scaled_dot_product_attention_8_out_mm_out, "transformer.resblocks.8.attn.out_proj.bias")
   add_1207 = Add (add_1092, node_scaled_dot_product_attention_8_out)
   layer_norm_17 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1207, "transformer.resblocks.8.ln_2.weight", "transformer.resblocks.8.ln_2.bias")
   val_91 = MatMul (layer_norm_17, val_27)
   linear_34 = Add (val_91, "transformer.resblocks.8.mlp.c_fc.bias")
   gelu_8 = Gelu <approximate: string = "none"> (linear_34)
   val_92 = MatMul (gelu_8, val_28)
   linear_35 = Add (val_92, "transformer.resblocks.8.mlp.c_proj.bias")
   add_1228 = Add (add_1207, linear_35)
   layer_norm_18 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1228, "transformer.resblocks.9.ln_1.weight", "transformer.resblocks.9.ln_1.bias")
   [node_scaled_dot_product_attention_9_qkv_mm] node_scaled_dot_product_attention_9_qkv_mm_out = MatMul (layer_norm_18, val_29)
   [node_scaled_dot_product_attention_9_qkv_bias] node_scaled_dot_product_attention_9_qkv = Add (node_scaled_dot_product_attention_9_qkv_mm_out, "transformer.resblocks.9.attn.in_proj_bias")
   [node_scaled_dot_product_attention_9_qkv_split] node_scaled_dot_product_attention_9_q, node_scaled_dot_product_attention_9_k, node_scaled_dot_product_attention_9_v = Split <axis: int = -1> (node_scaled_dot_product_attention_9_qkv, attn3d_split_3x1024)
   scaled_dot_product_attention_9 = Attention <is_causal: int = 1, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_9_q, node_scaled_dot_product_attention_9_k, node_scaled_dot_product_attention_9_v)
   [node_scaled_dot_product_attention_9_out_mm] node_scaled_dot_product_attention_9_out_mm_out = MatMul (scaled_dot_product_attention_9, node_scaled_dot_product_attention_9_wo_t)
   [node_scaled_dot_product_attention_9_out_bias] node_scaled_dot_product_attention_9_out = Add (node_scaled_dot_product_attention_9_out_mm_out, "transformer.resblocks.9.attn.out_proj.bias")
   add_1343 = Add (add_1228, node_scaled_dot_product_attention_9_out)
   layer_norm_19 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1343, "transformer.resblocks.9.ln_2.weight", "transformer.resblocks.9.ln_2.bias")
   val_93 = MatMul (layer_norm_19, val_30)
   linear_38 = Add (val_93, "transformer.resblocks.9.mlp.c_fc.bias")
   gelu_9 = Gelu <approximate: string = "none"> (linear_38)
   val_94 = MatMul (gelu_9, val_31)
   linear_39 = Add (val_94, "transformer.resblocks.9.mlp.c_proj.bias")
   add_1364 = Add (add_1343, linear_39)
   layer_norm_20 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1364, "transformer.resblocks.10.ln_1.weight", "transformer.resblocks.10.ln_1.bias")
   [node_scaled_dot_product_attention_10_qkv_mm] node_scaled_dot_product_attention_10_qkv_mm_out = MatMul (layer_norm_20, val_32)
   [node_scaled_dot_product_attention_10_qkv_bias] node_scaled_dot_product_attention_10_qkv = Add (node_scaled_dot_product_attention_10_qkv_mm_out, "transformer.resblocks.10.attn.in_proj_bias")
   [node_scaled_dot_product_attention_10_qkv_split] node_scaled_dot_product_attention_10_q, node_scaled_dot_product_attention_10_k, node_scaled_dot_product_attention_10_v = Split <axis: int = -1> (node_scaled_dot_product_attention_10_qkv, attn3d_split_3x1024)
   scaled_dot_product_attention_10 = Attention <is_causal: int = 1, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_10_q, node_scaled_dot_product_attention_10_k, node_scaled_dot_product_attention_10_v)
   [node_scaled_dot_product_attention_10_out_mm] node_scaled_dot_product_attention_10_out_mm_out = MatMul (scaled_dot_product_attention_10, node_scaled_dot_product_attention_10_wo_t)
   [node_scaled_dot_product_attention_10_out_bias] node_scaled_dot_product_attention_10_out = Add (node_scaled_dot_product_attention_10_out_mm_out, "transformer.resblocks.10.attn.out_proj.bias")
   add_1479 = Add (add_1364, node_scaled_dot_product_attention_10_out)
   layer_norm_21 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1479, "transformer.resblocks.10.ln_2.weight", "transformer.resblocks.10.ln_2.bias")
   val_95 = MatMul (layer_norm_21, val_33)
   linear_42 = Add (val_95, "transformer.resblocks.10.mlp.c_fc.bias")
   gelu_10 = Gelu <approximate: string = "none"> (linear_42)
   val_96 = MatMul (gelu_10, val_34)
   linear_43 = Add (val_96, "transformer.resblocks.10.mlp.c_proj.bias")
   add_1500 = Add (add_1479, linear_43)
   layer_norm_22 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1500, "transformer.resblocks.11.ln_1.weight", "transformer.resblocks.11.ln_1.bias")
   [node_scaled_dot_product_attention_11_qkv_mm] node_scaled_dot_product_attention_11_qkv_mm_out = MatMul (layer_norm_22, val_35)
   [node_scaled_dot_product_attention_11_qkv_bias] node_scaled_dot_product_attention_11_qkv = Add (node_scaled_dot_product_attention_11_qkv_mm_out, "transformer.resblocks.11.attn.in_proj_bias")
   [node_scaled_dot_product_attention_11_qkv_split] node_scaled_dot_product_attention_11_q, node_scaled_dot_product_attention_11_k, node_scaled_dot_product_attention_11_v = Split <axis: int = -1> (node_scaled_dot_product_attention_11_qkv, attn3d_split_3x1024)
   scaled_dot_product_attention_11 = Attention <is_causal: int = 1, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_11_q, node_scaled_dot_product_attention_11_k, node_scaled_dot_product_attention_11_v)
   [node_scaled_dot_product_attention_11_out_mm] node_scaled_dot_product_attention_11_out_mm_out = MatMul (scaled_dot_product_attention_11, node_scaled_dot_product_attention_11_wo_t)
   [node_scaled_dot_product_attention_11_out_bias] node_scaled_dot_product_attention_11_out = Add (node_scaled_dot_product_attention_11_out_mm_out, "transformer.resblocks.11.attn.out_proj.bias")
   add_1615 = Add (add_1500, node_scaled_dot_product_attention_11_out)
   layer_norm_23 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1615, "transformer.resblocks.11.ln_2.weight", "transformer.resblocks.11.ln_2.bias")
   val_97 = MatMul (layer_norm_23, val_36)
   linear_46 = Add (val_97, "transformer.resblocks.11.mlp.c_fc.bias")
   gelu_11 = Gelu <approximate: string = "none"> (linear_46)
   val_98 = MatMul (gelu_11, val_37)
   linear_47 = Add (val_98, "transformer.resblocks.11.mlp.c_proj.bias")
   add_1636 = Add (add_1615, linear_47)
   layer_norm_24 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1636, "transformer.resblocks.12.ln_1.weight", "transformer.resblocks.12.ln_1.bias")
   [node_scaled_dot_product_attention_12_qkv_mm] node_scaled_dot_product_attention_12_qkv_mm_out = MatMul (layer_norm_24, val_38)
   [node_scaled_dot_product_attention_12_qkv_bias] node_scaled_dot_product_attention_12_qkv = Add (node_scaled_dot_product_attention_12_qkv_mm_out, "transformer.resblocks.12.attn.in_proj_bias")
   [node_scaled_dot_product_attention_12_qkv_split] node_scaled_dot_product_attention_12_q, node_scaled_dot_product_attention_12_k, node_scaled_dot_product_attention_12_v = Split <axis: int = -1> (node_scaled_dot_product_attention_12_qkv, attn3d_split_3x1024)
   scaled_dot_product_attention_12 = Attention <is_causal: int = 1, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_12_q, node_scaled_dot_product_attention_12_k, node_scaled_dot_product_attention_12_v)
   [node_scaled_dot_product_attention_12_out_mm] node_scaled_dot_product_attention_12_out_mm_out = MatMul (scaled_dot_product_attention_12, node_scaled_dot_product_attention_12_wo_t)
   [node_scaled_dot_product_attention_12_out_bias] node_scaled_dot_product_attention_12_out = Add (node_scaled_dot_product_attention_12_out_mm_out, "transformer.resblocks.12.attn.out_proj.bias")
   add_1751 = Add (add_1636, node_scaled_dot_product_attention_12_out)
   layer_norm_25 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1751, "transformer.resblocks.12.ln_2.weight", "transformer.resblocks.12.ln_2.bias")
   val_99 = MatMul (layer_norm_25, val_39)
   linear_50 = Add (val_99, "transformer.resblocks.12.mlp.c_fc.bias")
   gelu_12 = Gelu <approximate: string = "none"> (linear_50)
   val_100 = MatMul (gelu_12, val_40)
   linear_51 = Add (val_100, "transformer.resblocks.12.mlp.c_proj.bias")
   add_1772 = Add (add_1751, linear_51)
   layer_norm_26 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1772, "transformer.resblocks.13.ln_1.weight", "transformer.resblocks.13.ln_1.bias")
   [node_scaled_dot_product_attention_13_qkv_mm] node_scaled_dot_product_attention_13_qkv_mm_out = MatMul (layer_norm_26, val_41)
   [node_scaled_dot_product_attention_13_qkv_bias] node_scaled_dot_product_attention_13_qkv = Add (node_scaled_dot_product_attention_13_qkv_mm_out, "transformer.resblocks.13.attn.in_proj_bias")
   [node_scaled_dot_product_attention_13_qkv_split] node_scaled_dot_product_attention_13_q, node_scaled_dot_product_attention_13_k, node_scaled_dot_product_attention_13_v = Split <axis: int = -1> (node_scaled_dot_product_attention_13_qkv, attn3d_split_3x1024)
   scaled_dot_product_attention_13 = Attention <is_causal: int = 1, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_13_q, node_scaled_dot_product_attention_13_k, node_scaled_dot_product_attention_13_v)
   [node_scaled_dot_product_attention_13_out_mm] node_scaled_dot_product_attention_13_out_mm_out = MatMul (scaled_dot_product_attention_13, node_scaled_dot_product_attention_13_wo_t)
   [node_scaled_dot_product_attention_13_out_bias] node_scaled_dot_product_attention_13_out = Add (node_scaled_dot_product_attention_13_out_mm_out, "transformer.resblocks.13.attn.out_proj.bias")
   add_1887 = Add (add_1772, node_scaled_dot_product_attention_13_out)
   layer_norm_27 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1887, "transformer.resblocks.13.ln_2.weight", "transformer.resblocks.13.ln_2.bias")
   val_101 = MatMul (layer_norm_27, val_42)
   linear_54 = Add (val_101, "transformer.resblocks.13.mlp.c_fc.bias")
   gelu_13 = Gelu <approximate: string = "none"> (linear_54)
   val_102 = MatMul (gelu_13, val_43)
   linear_55 = Add (val_102, "transformer.resblocks.13.mlp.c_proj.bias")
   add_1908 = Add (add_1887, linear_55)
   layer_norm_28 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1908, "transformer.resblocks.14.ln_1.weight", "transformer.resblocks.14.ln_1.bias")
   [node_scaled_dot_product_attention_14_qkv_mm] node_scaled_dot_product_attention_14_qkv_mm_out = MatMul (layer_norm_28, val_44)
   [node_scaled_dot_product_attention_14_qkv_bias] node_scaled_dot_product_attention_14_qkv = Add (node_scaled_dot_product_attention_14_qkv_mm_out, "transformer.resblocks.14.attn.in_proj_bias")
   [node_scaled_dot_product_attention_14_qkv_split] node_scaled_dot_product_attention_14_q, node_scaled_dot_product_attention_14_k, node_scaled_dot_product_attention_14_v = Split <axis: int = -1> (node_scaled_dot_product_attention_14_qkv, attn3d_split_3x1024)
   scaled_dot_product_attention_14 = Attention <is_causal: int = 1, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_14_q, node_scaled_dot_product_attention_14_k, node_scaled_dot_product_attention_14_v)
   [node_scaled_dot_product_attention_14_out_mm] node_scaled_dot_product_attention_14_out_mm_out = MatMul (scaled_dot_product_attention_14, node_scaled_dot_product_attention_14_wo_t)
   [node_scaled_dot_product_attention_14_out_bias] node_scaled_dot_product_attention_14_out = Add (node_scaled_dot_product_attention_14_out_mm_out, "transformer.resblocks.14.attn.out_proj.bias")
   add_2023 = Add (add_1908, node_scaled_dot_product_attention_14_out)
   layer_norm_29 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2023, "transformer.resblocks.14.ln_2.weight", "transformer.resblocks.14.ln_2.bias")
   val_103 = MatMul (layer_norm_29, val_45)
   linear_58 = Add (val_103, "transformer.resblocks.14.mlp.c_fc.bias")
   gelu_14 = Gelu <approximate: string = "none"> (linear_58)
   val_104 = MatMul (gelu_14, val_46)
   linear_59 = Add (val_104, "transformer.resblocks.14.mlp.c_proj.bias")
   add_2044 = Add (add_2023, linear_59)
   layer_norm_30 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2044, "transformer.resblocks.15.ln_1.weight", "transformer.resblocks.15.ln_1.bias")
   [node_scaled_dot_product_attention_15_qkv_mm] node_scaled_dot_product_attention_15_qkv_mm_out = MatMul (layer_norm_30, val_47)
   [node_scaled_dot_product_attention_15_qkv_bias] node_scaled_dot_product_attention_15_qkv = Add (node_scaled_dot_product_attention_15_qkv_mm_out, "transformer.resblocks.15.attn.in_proj_bias")
   [node_scaled_dot_product_attention_15_qkv_split] node_scaled_dot_product_attention_15_q, node_scaled_dot_product_attention_15_k, node_scaled_dot_product_attention_15_v = Split <axis: int = -1> (node_scaled_dot_product_attention_15_qkv, attn3d_split_3x1024)
   scaled_dot_product_attention_15 = Attention <is_causal: int = 1, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_15_q, node_scaled_dot_product_attention_15_k, node_scaled_dot_product_attention_15_v)
   [node_scaled_dot_product_attention_15_out_mm] node_scaled_dot_product_attention_15_out_mm_out = MatMul (scaled_dot_product_attention_15, node_scaled_dot_product_attention_15_wo_t)
   [node_scaled_dot_product_attention_15_out_bias] node_scaled_dot_product_attention_15_out = Add (node_scaled_dot_product_attention_15_out_mm_out, "transformer.resblocks.15.attn.out_proj.bias")
   add_2159 = Add (add_2044, node_scaled_dot_product_attention_15_out)
   layer_norm_31 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2159, "transformer.resblocks.15.ln_2.weight", "transformer.resblocks.15.ln_2.bias")
   val_105 = MatMul (layer_norm_31, val_48)
   linear_62 = Add (val_105, "transformer.resblocks.15.mlp.c_fc.bias")
   gelu_15 = Gelu <approximate: string = "none"> (linear_62)
   val_106 = MatMul (gelu_15, val_49)
   linear_63 = Add (val_106, "transformer.resblocks.15.mlp.c_proj.bias")
   add_2180 = Add (add_2159, linear_63)
   layer_norm_32 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2180, "transformer.resblocks.16.ln_1.weight", "transformer.resblocks.16.ln_1.bias")
   [node_scaled_dot_product_attention_16_qkv_mm] node_scaled_dot_product_attention_16_qkv_mm_out = MatMul (layer_norm_32, val_50)
   [node_scaled_dot_product_attention_16_qkv_bias] node_scaled_dot_product_attention_16_qkv = Add (node_scaled_dot_product_attention_16_qkv_mm_out, "transformer.resblocks.16.attn.in_proj_bias")
   [node_scaled_dot_product_attention_16_qkv_split] node_scaled_dot_product_attention_16_q, node_scaled_dot_product_attention_16_k, node_scaled_dot_product_attention_16_v = Split <axis: int = -1> (node_scaled_dot_product_attention_16_qkv, attn3d_split_3x1024)
   scaled_dot_product_attention_16 = Attention <is_causal: int = 1, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_16_q, node_scaled_dot_product_attention_16_k, node_scaled_dot_product_attention_16_v)
   [node_scaled_dot_product_attention_16_out_mm] node_scaled_dot_product_attention_16_out_mm_out = MatMul (scaled_dot_product_attention_16, node_scaled_dot_product_attention_16_wo_t)
   [node_scaled_dot_product_attention_16_out_bias] node_scaled_dot_product_attention_16_out = Add (node_scaled_dot_product_attention_16_out_mm_out, "transformer.resblocks.16.attn.out_proj.bias")
   add_2295 = Add (add_2180, node_scaled_dot_product_attention_16_out)
   layer_norm_33 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2295, "transformer.resblocks.16.ln_2.weight", "transformer.resblocks.16.ln_2.bias")
   val_107 = MatMul (layer_norm_33, val_51)
   linear_66 = Add (val_107, "transformer.resblocks.16.mlp.c_fc.bias")
   gelu_16 = Gelu <approximate: string = "none"> (linear_66)
   val_108 = MatMul (gelu_16, val_52)
   linear_67 = Add (val_108, "transformer.resblocks.16.mlp.c_proj.bias")
   add_2316 = Add (add_2295, linear_67)
   layer_norm_34 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2316, "transformer.resblocks.17.ln_1.weight", "transformer.resblocks.17.ln_1.bias")
   [node_scaled_dot_product_attention_17_qkv_mm] node_scaled_dot_product_attention_17_qkv_mm_out = MatMul (layer_norm_34, val_53)
   [node_scaled_dot_product_attention_17_qkv_bias] node_scaled_dot_product_attention_17_qkv = Add (node_scaled_dot_product_attention_17_qkv_mm_out, "transformer.resblocks.17.attn.in_proj_bias")
   [node_scaled_dot_product_attention_17_qkv_split] node_scaled_dot_product_attention_17_q, node_scaled_dot_product_attention_17_k, node_scaled_dot_product_attention_17_v = Split <axis: int = -1> (node_scaled_dot_product_attention_17_qkv, attn3d_split_3x1024)
   scaled_dot_product_attention_17 = Attention <is_causal: int = 1, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_17_q, node_scaled_dot_product_attention_17_k, node_scaled_dot_product_attention_17_v)
   [node_scaled_dot_product_attention_17_out_mm] node_scaled_dot_product_attention_17_out_mm_out = MatMul (scaled_dot_product_attention_17, node_scaled_dot_product_attention_17_wo_t)
   [node_scaled_dot_product_attention_17_out_bias] node_scaled_dot_product_attention_17_out = Add (node_scaled_dot_product_attention_17_out_mm_out, "transformer.resblocks.17.attn.out_proj.bias")
   add_2431 = Add (add_2316, node_scaled_dot_product_attention_17_out)
   layer_norm_35 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2431, "transformer.resblocks.17.ln_2.weight", "transformer.resblocks.17.ln_2.bias")
   val_109 = MatMul (layer_norm_35, val_54)
   linear_70 = Add (val_109, "transformer.resblocks.17.mlp.c_fc.bias")
   gelu_17 = Gelu <approximate: string = "none"> (linear_70)
   val_110 = MatMul (gelu_17, val_55)
   linear_71 = Add (val_110, "transformer.resblocks.17.mlp.c_proj.bias")
   add_2452 = Add (add_2431, linear_71)
   layer_norm_36 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2452, "transformer.resblocks.18.ln_1.weight", "transformer.resblocks.18.ln_1.bias")
   [node_scaled_dot_product_attention_18_qkv_mm] node_scaled_dot_product_attention_18_qkv_mm_out = MatMul (layer_norm_36, val_56)
   [node_scaled_dot_product_attention_18_qkv_bias] node_scaled_dot_product_attention_18_qkv = Add (node_scaled_dot_product_attention_18_qkv_mm_out, "transformer.resblocks.18.attn.in_proj_bias")
   [node_scaled_dot_product_attention_18_qkv_split] node_scaled_dot_product_attention_18_q, node_scaled_dot_product_attention_18_k, node_scaled_dot_product_attention_18_v = Split <axis: int = -1> (node_scaled_dot_product_attention_18_qkv, attn3d_split_3x1024)
   scaled_dot_product_attention_18 = Attention <is_causal: int = 1, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_18_q, node_scaled_dot_product_attention_18_k, node_scaled_dot_product_attention_18_v)
   [node_scaled_dot_product_attention_18_out_mm] node_scaled_dot_product_attention_18_out_mm_out = MatMul (scaled_dot_product_attention_18, node_scaled_dot_product_attention_18_wo_t)
   [node_scaled_dot_product_attention_18_out_bias] node_scaled_dot_product_attention_18_out = Add (node_scaled_dot_product_attention_18_out_mm_out, "transformer.resblocks.18.attn.out_proj.bias")
   add_2567 = Add (add_2452, node_scaled_dot_product_attention_18_out)
   layer_norm_37 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2567, "transformer.resblocks.18.ln_2.weight", "transformer.resblocks.18.ln_2.bias")
   val_111 = MatMul (layer_norm_37, val_57)
   linear_74 = Add (val_111, "transformer.resblocks.18.mlp.c_fc.bias")
   gelu_18 = Gelu <approximate: string = "none"> (linear_74)
   val_112 = MatMul (gelu_18, val_58)
   linear_75 = Add (val_112, "transformer.resblocks.18.mlp.c_proj.bias")
   add_2588 = Add (add_2567, linear_75)
   layer_norm_38 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2588, "transformer.resblocks.19.ln_1.weight", "transformer.resblocks.19.ln_1.bias")
   [node_scaled_dot_product_attention_19_qkv_mm] node_scaled_dot_product_attention_19_qkv_mm_out = MatMul (layer_norm_38, val_59)
   [node_scaled_dot_product_attention_19_qkv_bias] node_scaled_dot_product_attention_19_qkv = Add (node_scaled_dot_product_attention_19_qkv_mm_out, "transformer.resblocks.19.attn.in_proj_bias")
   [node_scaled_dot_product_attention_19_qkv_split] node_scaled_dot_product_attention_19_q, node_scaled_dot_product_attention_19_k, node_scaled_dot_product_attention_19_v = Split <axis: int = -1> (node_scaled_dot_product_attention_19_qkv, attn3d_split_3x1024)
   scaled_dot_product_attention_19 = Attention <is_causal: int = 1, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_19_q, node_scaled_dot_product_attention_19_k, node_scaled_dot_product_attention_19_v)
   [node_scaled_dot_product_attention_19_out_mm] node_scaled_dot_product_attention_19_out_mm_out = MatMul (scaled_dot_product_attention_19, node_scaled_dot_product_attention_19_wo_t)
   [node_scaled_dot_product_attention_19_out_bias] node_scaled_dot_product_attention_19_out = Add (node_scaled_dot_product_attention_19_out_mm_out, "transformer.resblocks.19.attn.out_proj.bias")
   add_2703 = Add (add_2588, node_scaled_dot_product_attention_19_out)
   layer_norm_39 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2703, "transformer.resblocks.19.ln_2.weight", "transformer.resblocks.19.ln_2.bias")
   val_113 = MatMul (layer_norm_39, val_60)
   linear_78 = Add (val_113, "transformer.resblocks.19.mlp.c_fc.bias")
   gelu_19 = Gelu <approximate: string = "none"> (linear_78)
   val_114 = MatMul (gelu_19, val_61)
   linear_79 = Add (val_114, "transformer.resblocks.19.mlp.c_proj.bias")
   add_2724 = Add (add_2703, linear_79)
   layer_norm_40 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2724, "transformer.resblocks.20.ln_1.weight", "transformer.resblocks.20.ln_1.bias")
   [node_scaled_dot_product_attention_20_qkv_mm] node_scaled_dot_product_attention_20_qkv_mm_out = MatMul (layer_norm_40, val_62)
   [node_scaled_dot_product_attention_20_qkv_bias] node_scaled_dot_product_attention_20_qkv = Add (node_scaled_dot_product_attention_20_qkv_mm_out, "transformer.resblocks.20.attn.in_proj_bias")
   [node_scaled_dot_product_attention_20_qkv_split] node_scaled_dot_product_attention_20_q, node_scaled_dot_product_attention_20_k, node_scaled_dot_product_attention_20_v = Split <axis: int = -1> (node_scaled_dot_product_attention_20_qkv, attn3d_split_3x1024)
   scaled_dot_product_attention_20 = Attention <is_causal: int = 1, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_20_q, node_scaled_dot_product_attention_20_k, node_scaled_dot_product_attention_20_v)
   [node_scaled_dot_product_attention_20_out_mm] node_scaled_dot_product_attention_20_out_mm_out = MatMul (scaled_dot_product_attention_20, node_scaled_dot_product_attention_20_wo_t)
   [node_scaled_dot_product_attention_20_out_bias] node_scaled_dot_product_attention_20_out = Add (node_scaled_dot_product_attention_20_out_mm_out, "transformer.resblocks.20.attn.out_proj.bias")
   add_2839 = Add (add_2724, node_scaled_dot_product_attention_20_out)
   layer_norm_41 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2839, "transformer.resblocks.20.ln_2.weight", "transformer.resblocks.20.ln_2.bias")
   val_115 = MatMul (layer_norm_41, val_63)
   linear_82 = Add (val_115, "transformer.resblocks.20.mlp.c_fc.bias")
   gelu_20 = Gelu <approximate: string = "none"> (linear_82)
   val_116 = MatMul (gelu_20, val_64)
   linear_83 = Add (val_116, "transformer.resblocks.20.mlp.c_proj.bias")
   add_2860 = Add (add_2839, linear_83)
   layer_norm_42 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2860, "transformer.resblocks.21.ln_1.weight", "transformer.resblocks.21.ln_1.bias")
   [node_scaled_dot_product_attention_21_qkv_mm] node_scaled_dot_product_attention_21_qkv_mm_out = MatMul (layer_norm_42, val_65)
   [node_scaled_dot_product_attention_21_qkv_bias] node_scaled_dot_product_attention_21_qkv = Add (node_scaled_dot_product_attention_21_qkv_mm_out, "transformer.resblocks.21.attn.in_proj_bias")
   [node_scaled_dot_product_attention_21_qkv_split] node_scaled_dot_product_attention_21_q, node_scaled_dot_product_attention_21_k, node_scaled_dot_product_attention_21_v = Split <axis: int = -1> (node_scaled_dot_product_attention_21_qkv, attn3d_split_3x1024)
   scaled_dot_product_attention_21 = Attention <is_causal: int = 1, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_21_q, node_scaled_dot_product_attention_21_k, node_scaled_dot_product_attention_21_v)
   [node_scaled_dot_product_attention_21_out_mm] node_scaled_dot_product_attention_21_out_mm_out = MatMul (scaled_dot_product_attention_21, node_scaled_dot_product_attention_21_wo_t)
   [node_scaled_dot_product_attention_21_out_bias] node_scaled_dot_product_attention_21_out = Add (node_scaled_dot_product_attention_21_out_mm_out, "transformer.resblocks.21.attn.out_proj.bias")
   add_2975 = Add (add_2860, node_scaled_dot_product_attention_21_out)
   layer_norm_43 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2975, "transformer.resblocks.21.ln_2.weight", "transformer.resblocks.21.ln_2.bias")
   val_117 = MatMul (layer_norm_43, val_66)
   linear_86 = Add (val_117, "transformer.resblocks.21.mlp.c_fc.bias")
   gelu_21 = Gelu <approximate: string = "none"> (linear_86)
   val_118 = MatMul (gelu_21, val_67)
   linear_87 = Add (val_118, "transformer.resblocks.21.mlp.c_proj.bias")
   add_2996 = Add (add_2975, linear_87)
   layer_norm_44 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2996, "transformer.resblocks.22.ln_1.weight", "transformer.resblocks.22.ln_1.bias")
   [node_scaled_dot_product_attention_22_qkv_mm] node_scaled_dot_product_attention_22_qkv_mm_out = MatMul (layer_norm_44, val_68)
   [node_scaled_dot_product_attention_22_qkv_bias] node_scaled_dot_product_attention_22_qkv = Add (node_scaled_dot_product_attention_22_qkv_mm_out, "transformer.resblocks.22.attn.in_proj_bias")
   [node_scaled_dot_product_attention_22_qkv_split] node_scaled_dot_product_attention_22_q, node_scaled_dot_product_attention_22_k, node_scaled_dot_product_attention_22_v = Split <axis: int = -1> (node_scaled_dot_product_attention_22_qkv, attn3d_split_3x1024)
   scaled_dot_product_attention_22 = Attention <is_causal: int = 1, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_22_q, node_scaled_dot_product_attention_22_k, node_scaled_dot_product_attention_22_v)
   [node_scaled_dot_product_attention_22_out_mm] node_scaled_dot_product_attention_22_out_mm_out = MatMul (scaled_dot_product_attention_22, node_scaled_dot_product_attention_22_wo_t)
   [node_scaled_dot_product_attention_22_out_bias] node_scaled_dot_product_attention_22_out = Add (node_scaled_dot_product_attention_22_out_mm_out, "transformer.resblocks.22.attn.out_proj.bias")
   add_3111 = Add (add_2996, node_scaled_dot_product_attention_22_out)
   layer_norm_45 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_3111, "transformer.resblocks.22.ln_2.weight", "transformer.resblocks.22.ln_2.bias")
   val_119 = MatMul (layer_norm_45, val_69)
   linear_90 = Add (val_119, "transformer.resblocks.22.mlp.c_fc.bias")
   gelu_22 = Gelu <approximate: string = "none"> (linear_90)
   val_120 = MatMul (gelu_22, val_70)
   linear_91 = Add (val_120, "transformer.resblocks.22.mlp.c_proj.bias")
   add_3132 = Add (add_3111, linear_91)
   layer_norm_46 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_3132, "transformer.resblocks.23.ln_1.weight", "transformer.resblocks.23.ln_1.bias")
   [node_scaled_dot_product_attention_23_qkv_mm] node_scaled_dot_product_attention_23_qkv_mm_out = MatMul (layer_norm_46, val_71)
   [node_scaled_dot_product_attention_23_qkv_bias] node_scaled_dot_product_attention_23_qkv = Add (node_scaled_dot_product_attention_23_qkv_mm_out, "transformer.resblocks.23.attn.in_proj_bias")
   [node_scaled_dot_product_attention_23_qkv_split] node_scaled_dot_product_attention_23_q, node_scaled_dot_product_attention_23_k, node_scaled_dot_product_attention_23_v = Split <axis: int = -1> (node_scaled_dot_product_attention_23_qkv, attn3d_split_3x1024)
   scaled_dot_product_attention_23 = Attention <is_causal: int = 1, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_23_q, node_scaled_dot_product_attention_23_k, node_scaled_dot_product_attention_23_v)
   [node_argmax] argmax = ArgMax <axis: int = -1, keepdims: int = 0, select_last_index: int = 0> (text)
   val_121 = Gather <axis: int = 0> (val_1487_eye, argmax)
   val_122 = Unsqueeze (val_121, val_1487_axes1)
   [pool_hoist_scaled_dot_product_attention_23] scaled_dot_product_attention_23_pooled = MatMul (val_122, scaled_dot_product_attention_23)
   [node_scaled_dot_product_attention_23_out_mm] node_scaled_dot_product_attention_23_out_mm_out = MatMul (scaled_dot_product_attention_23_pooled, node_scaled_dot_product_attention_23_wo_t)
   [node_scaled_dot_product_attention_23_out_bias] node_scaled_dot_product_attention_23_out = Add (node_scaled_dot_product_attention_23_out_mm_out, "transformer.resblocks.23.attn.out_proj.bias")
   [pool_hoist_add_3132] add_3132_pooled = MatMul (val_122, add_3132)
   add_3247 = Add (add_3132_pooled, node_scaled_dot_product_attention_23_out)
   layer_norm_47 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_3247, "transformer.resblocks.23.ln_2.weight", "transformer.resblocks.23.ln_2.bias")
   val_123 = MatMul (layer_norm_47, val_72)
   linear_94 = Add (val_123, "transformer.resblocks.23.mlp.c_fc.bias")
   gelu_23 = Gelu <approximate: string = "none"> (linear_94)
   val_124 = MatMul (gelu_23, val_73)
   linear_95 = Add (val_124, "transformer.resblocks.23.mlp.c_proj.bias")
   add_3268 = Add (add_3247, linear_95)
   val_125 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_3268, "ln_final.weight", "ln_final.bias")
   val_126 = Squeeze (val_125, val_1487_axes1)
   [node_matmul] matmul = MatMul (val_126, text_projection)
   [node_linalg_vector_norm] linalg_vector_norm = ReduceL2 <keepdims: int = 1, noop_with_empty_axes: int = 0> (matmul, val_0)
   [node_clamp_min] clamp_min = Clip (linalg_vector_norm, val_1)
   [node_div] text_embedding = Div (matmul, clamp_min)
}

weights:
attn3d_split_3x1024 INT64[3] 0ec6f5651fd5
ln_final.bias FLOAT[1024] f3ed701c93f4
ln_final.weight FLOAT[1024] 7303250c4ca6
node_scaled_dot_product_attention_10_wo_t FLOAT[1024,1024] b63feb6dc349
node_scaled_dot_product_attention_11_wo_t FLOAT[1024,1024] fe1c05f7f443
node_scaled_dot_product_attention_12_wo_t FLOAT[1024,1024] 5ea1277d07d3
node_scaled_dot_product_attention_13_wo_t FLOAT[1024,1024] 01fa6b46a58f
node_scaled_dot_product_attention_14_wo_t FLOAT[1024,1024] d20106abb2cd
node_scaled_dot_product_attention_15_wo_t FLOAT[1024,1024] 87e13de0631a
node_scaled_dot_product_attention_16_wo_t FLOAT[1024,1024] c1ae7c3934fd
node_scaled_dot_product_attention_17_wo_t FLOAT[1024,1024] 44f139d812e4
node_scaled_dot_product_attention_18_wo_t FLOAT[1024,1024] ef2d84bbab29
node_scaled_dot_product_attention_19_wo_t FLOAT[1024,1024] 789540b88058
node_scaled_dot_product_attention_1_wo_t FLOAT[1024,1024] 7cc3038e572b
node_scaled_dot_product_attention_20_wo_t FLOAT[1024,1024] 7beb618fdd69
node_scaled_dot_product_attention_21_wo_t FLOAT[1024,1024] 8fce02e4d901
node_scaled_dot_product_attention_22_wo_t FLOAT[1024,1024] 19a7c9c4cf92
node_scaled_dot_product_attention_23_wo_t FLOAT[1024,1024] 0cc299633f1f
node_scaled_dot_product_attention_2_wo_t FLOAT[1024,1024] 6206b6dc8b2e
node_scaled_dot_product_attention_3_wo_t FLOAT[1024,1024] 3dbc7b6c58aa
node_scaled_dot_product_attention_4_wo_t FLOAT[1024,1024] db768f97c342
node_scaled_dot_product_attention_5_wo_t FLOAT[1024,1024] c2f533def169
node_scaled_dot_product_attention_6_wo_t FLOAT[1024,1024] 5fd77b00c87e
node_scaled_dot_product_attention_7_wo_t FLOAT[1024,1024] f4be2b4f2226
node_scaled_dot_product_attention_8_wo_t FLOAT[1024,1024] 7fb57fac5adc
node_scaled_dot_product_attention_9_wo_t FLOAT[1024,1024] 3c58a5da32be
node_scaled_dot_product_attention_wo_t FLOAT[1024,1024] 3fc8b164cbc1
positional_embedding FLOAT[77,1024] b809ce3bedea
text_projection FLOAT[1024,1024] 82cbe5c68519
token_embedding.weight_fp16 FLOAT16[49408,1024] f17e12a3bb28
transformer.resblocks.0.attn.in_proj_bias FLOAT[3072] ba94abb93a8b
transformer.resblocks.0.attn.out_proj.bias FLOAT[1024] 801db8f29af9
transformer.resblocks.0.ln_1.bias FLOAT[1024] 08400718e35f
transformer.resblocks.0.ln_1.weight FLOAT[1024] 7e8cb64ed94e
transformer.resblocks.0.ln_2.bias FLOAT[1024] 568a6072ca54
transformer.resblocks.0.ln_2.weight FLOAT[1024] 3e5ffaab71a9
transformer.resblocks.0.mlp.c_fc.bias FLOAT[4096] 5cec1460bd7e
transformer.resblocks.0.mlp.c_proj.bias FLOAT[1024] 01fa56e8dedb
transformer.resblocks.1.attn.in_proj_bias FLOAT[3072] 21336a5ae5dc
transformer.resblocks.1.attn.out_proj.bias FLOAT[1024] f8ddcb38ddfb
transformer.resblocks.1.ln_1.bias FLOAT[1024] 1d2fc16c7af5
transformer.resblocks.1.ln_1.weight FLOAT[1024] c55605f84e03
transformer.resblocks.1.ln_2.bias FLOAT[1024] 38bb96c197df
transformer.resblocks.1.ln_2.weight FLOAT[1024] 7d5a5bfe811c
transformer.resblocks.1.mlp.c_fc.bias FLOAT[4096] 4b3b32d1fb09
transformer.resblocks.1.mlp.c_proj.bias FLOAT[1024] 9306468b2d4c
transformer.resblocks.10.attn.in_proj_bias FLOAT[3072] 646c393eafab
transformer.resblocks.10.attn.out_proj.bias FLOAT[1024] e1de46476167
transformer.resblocks.10.ln_1.bias FLOAT[1024] 71e173d369f0
transformer.resblocks.10.ln_1.weight FLOAT[1024] 821d9a1ed983
transformer.resblocks.10.ln_2.bias FLOAT[1024] 118ac7042326
transformer.resblocks.10.ln_2.weight FLOAT[1024] 8901d343fd5a
transformer.resblocks.10.mlp.c_fc.bias FLOAT[4096] 93b5912c5e66
transformer.resblocks.10.mlp.c_proj.bias FLOAT[1024] d0a7f084757b
transformer.resblocks.11.attn.in_proj_bias FLOAT[3072] c279d9cf0b8c
transformer.resblocks.11.attn.out_proj.bias FLOAT[1024] c7390f69dcf0
transformer.resblocks.11.ln_1.bias FLOAT[1024] c20c95a88297
transformer.resblocks.11.ln_1.weight FLOAT[1024] 42eb1cb7e823
transformer.resblocks.11.ln_2.bias FLOAT[1024] e0e271884022
transformer.resblocks.11.ln_2.weight FLOAT[1024] 6f0dac3b74b6
transformer.resblocks.11.mlp.c_fc.bias FLOAT[4096] a63ae60f0a09
transformer.resblocks.11.mlp.c_proj.bias FLOAT[1024] 034c033255a7
transformer.resblocks.12.attn.in_proj_bias FLOAT[3072] 824596392b10
transformer.resblocks.12.attn.out_proj.bias FLOAT[1024] d289801be047
transformer.resblocks.12.ln_1.bias FLOAT[1024] b38c55ee7b04
transformer.resblocks.12.ln_1.weight FLOAT[1024] fe4cc9e35e54
transformer.resblocks.12.ln_2.bias FLOAT[1024] c9a5befb08b1
transformer.resblocks.12.ln_2.weight FLOAT[1024] ccf5e085fd2f
transformer.resblocks.12.mlp.c_fc.bias FLOAT[4096] 8d092e08bf94
transformer.resblocks.12.mlp.c_proj.bias FLOAT[1024] 8b1d07075fad
transformer.resblocks.13.attn.in_proj_bias FLOAT[3072] 4da8b942e1f9
transformer.resblocks.13.attn.out_proj.bias FLOAT[1024] ecd183ff6d75
transformer.resblocks.13.ln_1.bias FLOAT[1024] 42bd00e90fe4
transformer.resblocks.13.ln_1.weight FLOAT[1024] 22e25af34c38
transformer.resblocks.13.ln_2.bias FLOAT[1024] ba82eeef4258
transformer.resblocks.13.ln_2.weight FLOAT[1024] 1abadd70378e
transformer.resblocks.13.mlp.c_fc.bias FLOAT[4096] f92ed72f31d0
transformer.resblocks.13.mlp.c_proj.bias FLOAT[1024] 53e84edb9adf
transformer.resblocks.14.attn.in_proj_bias FLOAT[3072] 5415f15ca5be
transformer.resblocks.14.attn.out_proj.bias FLOAT[1024] bacc731f1343
transformer.resblocks.14.ln_1.bias FLOAT[1024] 8e5c0379bed2
transformer.resblocks.14.ln_1.weight FLOAT[1024] bd68e8a7841f
transformer.resblocks.14.ln_2.bias FLOAT[1024] 1362504647c5
transformer.resblocks.14.ln_2.weight FLOAT[1024] d31befd3a8ba
transformer.resblocks.14.mlp.c_fc.bias FLOAT[4096] 5784def37856
transformer.resblocks.14.mlp.c_proj.bias FLOAT[1024] 9529e51441c5
transformer.resblocks.15.attn.in_proj_bias FLOAT[3072] 6ae44b0db03c
transformer.resblocks.15.attn.out_proj.bias FLOAT[1024] 98e2e30e8e94
transformer.resblocks.15.ln_1.bias FLOAT[1024] 3a2cbb6e573a
transformer.resblocks.15.ln_1.weight FLOAT[1024] 3d018ac0421c
transformer.resblocks.15.ln_2.bias FLOAT[1024] 596be965c249
transformer.resblocks.15.ln_2.weight FLOAT[1024] ba71c272d8dc
transformer.resblocks.15.mlp.c_fc.bias FLOAT[4096] ac6def7b3fe9
transformer.resblocks.15.mlp.c_proj.bias FLOAT[1024] 558441cf7732
transformer.resblocks.16.attn.in_proj_bias FLOAT[3072] fb6a3dc38d83
transformer.resblocks.16.attn.out_proj.bias FLOAT[1024] 8ab97ab3b69d
transformer.resblocks.16.ln_1.bias FLOAT[1024] 7f0b95aa1e97
transformer.resblocks.16.ln_1.weight FLOAT[1024] 2631157e94cb
transformer.resblocks.16.ln_2.bias FLOAT[1024] 333c980d5a9a
transformer.resblocks.16.ln_2.weight FLOAT[1024] 706e5f610679
transformer.resblocks.16.mlp.c_fc.bias FLOAT[4096] 940316f6c8ca
transformer.resblocks.16.mlp.c_proj.bias FLOAT[1024] 48ec5dba7b5b
transformer.resblocks.17.attn.in_proj_bias FLOAT[3072] 979f789a0943
transformer.resblocks.17.attn.out_proj.bias FLOAT[1024] 7922cf416599
transformer.resblocks.17.ln_1.bias FLOAT[1024] d1ab860eb24c
transformer.resblocks.17.ln_1.weight FLOAT[1024] 56f3639307e8
transformer.resblocks.17.ln_2.bias FLOAT[1024] 0307ea9f23aa
transformer.resblocks.17.ln_2.weight FLOAT[1024] 5bebdfeb586b
transformer.resblocks.17.mlp.c_fc.bias FLOAT[4096] 318c122e9d2a
transformer.resblocks.17.mlp.c_proj.bias FLOAT[1024] 019f037f0f02
transformer.resblocks.18.attn.in_proj_bias FLOAT[3072] e2cba69c98d9
transformer.resblocks.18.attn.out_proj.bias FLOAT[1024] f48cb7e3dbc2
transformer.resblocks.18.ln_1.bias FLOAT[1024] b2adcc587dab
transformer.resblocks.18.ln_1.weight FLOAT[1024] e461ff20e3ab
transformer.resblocks.18.ln_2.bias FLOAT[1024] 061e14add6b4
transformer.resblocks.18.ln_2.weight FLOAT[1024] ec44efbe0d65
transformer.resblocks.18.mlp.c_fc.bias FLOAT[4096] 7d9139349450
transformer.resblocks.18.mlp.c_proj.bias FLOAT[1024] 8305243e939f
transformer.resblocks.19.attn.in_proj_bias FLOAT[3072] 0130a941a76a
transformer.resblocks.19.attn.out_proj.bias FLOAT[1024] 438287b08cb6
transformer.resblocks.19.ln_1.bias FLOAT[1024] f523d5a297a2
transformer.resblocks.19.ln_1.weight FLOAT[1024] 6870d3e8de8e
transformer.resblocks.19.ln_2.bias FLOAT[1024] 7c84f52deb9d
transformer.resblocks.19.ln_2.weight FLOAT[1024] de7d0948a39c
transformer.resblocks.19.mlp.c_fc.bias FLOAT[4096] e1b578dfc3d1
transformer.resblocks.19.mlp.c_proj.bias FLOAT[1024] 84b55350d34f
transformer.resblocks.2.attn.in_proj_bias FLOAT[3072] 4a65fa8cf38b
transformer.resblocks.2.attn.out_proj.bias FLOAT[1024] 5541f352aa9f
transformer.resblocks.2.ln_1.bias FLOAT[1024] c38f681a465a
transformer.resblocks.2.ln_1.weight FLOAT[1024] 63c1afc221b0
transformer.resblocks.2.ln_2.bias FLOAT[1024] 6af458122c77
transformer.resblocks.2.ln_2.weight FLOAT[1024] 987195821706
transformer.resblocks.2.mlp.c_fc.bias FLOAT[4096] 3b21ab7653a9
transformer.resblocks.2.mlp.c_proj.bias FLOAT[1024] 8de21e7e267d
transformer.resblocks.20.attn.in_proj_bias FLOAT[3072] 658c0a099a81
transformer.resblocks.20.attn.out_proj.bias FLOAT[1024] f433957e902c
transformer.resblocks.20.ln_1.bias FLOAT[1024] 37c9ed08a8a0
transformer.resblocks.20.ln_1.weight FLOAT[1024] f866a69b5c87
transformer.resblocks.20.ln_2.bias FLOAT[1024] 9bef8467fbd0
transformer.resblocks.20.ln_2.weight FLOAT[1024] 93e3730a5012
transformer.resblocks.20.mlp.c_fc.bias FLOAT[4096] 46ee3e619c7b
transformer.resblocks.20.mlp.c_proj.bias FLOAT[1024] 0a0ea83cee98
transformer.resblocks.21.attn.in_proj_bias FLOAT[3072] 4776052baac8
transformer.resblocks.21.attn.out_proj.bias FLOAT[1024] c22390af9be2
transformer.resblocks.21.ln_1.bias FLOAT[1024] d523b075267e
transformer.resblocks.21.ln_1.weight FLOAT[1024] 192c0b3d0131
transformer.resblocks.21.ln_2.bias FLOAT[1024] d17f10fd8917
transformer.resblocks.21.ln_2.weight FLOAT[1024] 0caa675f094a
transformer.resblocks.21.mlp.c_fc.bias FLOAT[4096] 79e7b93f5bfe
transformer.resblocks.21.mlp.c_proj.bias FLOAT[1024] fe66218a3607
transformer.resblocks.22.attn.in_proj_bias FLOAT[3072] cd8adde4def0
transformer.resblocks.22.attn.out_proj.bias FLOAT[1024] 6b8e5abb0fa6
transformer.resblocks.22.ln_1.bias FLOAT[1024] 7feb891f74bc
transformer.resblocks.22.ln_1.weight FLOAT[1024] 54c8b1d18543
transformer.resblocks.22.ln_2.bias FLOAT[1024] a55f871c1d3d
transformer.resblocks.22.ln_2.weight FLOAT[1024] 43f95d9aa4e0
transformer.resblocks.22.mlp.c_fc.bias FLOAT[4096] 0fe344681d18
transformer.resblocks.22.mlp.c_proj.bias FLOAT[1024] d4794a9fd8b3
transformer.resblocks.23.attn.in_proj_bias FLOAT[3072] 8f246cf26dfa
transformer.resblocks.23.attn.out_proj.bias FLOAT[1024] a0b0189da929
transformer.resblocks.23.ln_1.bias FLOAT[1024] e17aaaefb4c6
transformer.resblocks.23.ln_1.weight FLOAT[1024] 6bdd9734c444
transformer.resblocks.23.ln_2.bias FLOAT[1024] 00deafb11c8d
transformer.resblocks.23.ln_2.weight FLOAT[1024] fb60775f6316
transformer.resblocks.23.mlp.c_fc.bias FLOAT[4096] 5179589c2a2a
transformer.resblocks.23.mlp.c_proj.bias FLOAT[1024] 7dac19b8c07c
transformer.resblocks.3.attn.in_proj_bias FLOAT[3072] 0e3beb4e1e0e
transformer.resblocks.3.attn.out_proj.bias FLOAT[1024] 6f3f2c773791
transformer.resblocks.3.ln_1.bias FLOAT[1024] 5b6e5f460578
transformer.resblocks.3.ln_1.weight FLOAT[1024] 5420a928ac8d
transformer.resblocks.3.ln_2.bias FLOAT[1024] 54920fbcd16d
transformer.resblocks.3.ln_2.weight FLOAT[1024] b6d18a12a28b
transformer.resblocks.3.mlp.c_fc.bias FLOAT[4096] 91161a8e561f
transformer.resblocks.3.mlp.c_proj.bias FLOAT[1024] d57e45922f9a
transformer.resblocks.4.attn.in_proj_bias FLOAT[3072] 049514746e67
transformer.resblocks.4.attn.out_proj.bias FLOAT[1024] a85989a9dbc9
transformer.resblocks.4.ln_1.bias FLOAT[1024] 584b4ec83e7f
transformer.resblocks.4.ln_1.weight FLOAT[1024] 84745e3a3a9f
transformer.resblocks.4.ln_2.bias FLOAT[1024] 01fd54c2625f
transformer.resblocks.4.ln_2.weight FLOAT[1024] 018c8cac9cae
transformer.resblocks.4.mlp.c_fc.bias FLOAT[4096] 358389813084
transformer.resblocks.4.mlp.c_proj.bias FLOAT[1024] 517c1a4c19a7
transformer.resblocks.5.attn.in_proj_bias FLOAT[3072] bd80786b88d8
transformer.resblocks.5.attn.out_proj.bias FLOAT[1024] 853d0f352051
transformer.resblocks.5.ln_1.bias FLOAT[1024] e6a9272b8e07
transformer.resblocks.5.ln_1.weight FLOAT[1024] abdba5892302
transformer.resblocks.5.ln_2.bias FLOAT[1024] 3d2d5b61d1c7
transformer.resblocks.5.ln_2.weight FLOAT[1024] e4a36f4f6daa
transformer.resblocks.5.mlp.c_fc.bias FLOAT[4096] 54de17359a33
transformer.resblocks.5.mlp.c_proj.bias FLOAT[1024] cc1216f31163
transformer.resblocks.6.attn.in_proj_bias FLOAT[3072] de4a1584344d
transformer.resblocks.6.attn.out_proj.bias FLOAT[1024] 0eae01d5f485
transformer.resblocks.6.ln_1.bias FLOAT[1024] de39101c5cd8
transformer.resblocks.6.ln_1.weight FLOAT[1024] 6e4726e41e4a
transformer.resblocks.6.ln_2.bias FLOAT[1024] efae6790626f
transformer.resblocks.6.ln_2.weight FLOAT[1024] 387be8217e30
transformer.resblocks.6.mlp.c_fc.bias FLOAT[4096] 5e89e3199b02
transformer.resblocks.6.mlp.c_proj.bias FLOAT[1024] 443e5461fd33
transformer.resblocks.7.attn.in_proj_bias FLOAT[3072] 4cebd7547344
transformer.resblocks.7.attn.out_proj.bias FLOAT[1024] c2a659e1e29b
transformer.resblocks.7.ln_1.bias FLOAT[1024] b8ae51ef5d72
transformer.resblocks.7.ln_1.weight FLOAT[1024] 87c752a116b5
transformer.resblocks.7.ln_2.bias FLOAT[1024] 8bca31f4100c
transformer.resblocks.7.ln_2.weight FLOAT[1024] b9e14f9470bd
transformer.resblocks.7.mlp.c_fc.bias FLOAT[4096] 83ab20679224
transformer.resblocks.7.mlp.c_proj.bias FLOAT[1024] 23d296abb10c
transformer.resblocks.8.attn.in_proj_bias FLOAT[3072] 1fa2e562e6c6
transformer.resblocks.8.attn.out_proj.bias FLOAT[1024] 6d0d1053c54c
transformer.resblocks.8.ln_1.bias FLOAT[1024] 9b1160b23238
transformer.resblocks.8.ln_1.weight FLOAT[1024] 51334c72c30d
transformer.resblocks.8.ln_2.bias FLOAT[1024] 208953d3b4be
transformer.resblocks.8.ln_2.weight FLOAT[1024] 87997de12ed1
transformer.resblocks.8.mlp.c_fc.bias FLOAT[4096] c3ced7290d0c
transformer.resblocks.8.mlp.c_proj.bias FLOAT[1024] 0575e9494da6
transformer.resblocks.9.attn.in_proj_bias FLOAT[3072] 3e00968e8b03
transformer.resblocks.9.attn.out_proj.bias FLOAT[1024] fc598c6aee0e
transformer.resblocks.9.ln_1.bias FLOAT[1024] 61169a5fcc0c
transformer.resblocks.9.ln_1.weight FLOAT[1024] 01f5786cb78f
transformer.resblocks.9.ln_2.bias FLOAT[1024] b5c5844193c3
transformer.resblocks.9.ln_2.weight FLOAT[1024] c8e763e620c9
transformer.resblocks.9.mlp.c_fc.bias FLOAT[4096] 407a38fef072
transformer.resblocks.9.mlp.c_proj.bias FLOAT[1024] 0adce4f099ad
val_0 INT64[1] 12a3ae445661
val_1 FLOAT[] 6708d9be4956
val_10 FLOAT[4096,1024] a2d93b20b202
val_11 FLOAT[1024,3072] c3573bff887d
val_12 FLOAT[1024,4096] c4cf7c360c89
val_13 FLOAT[4096,1024] 4268560cfb19
val_14 FLOAT[1024,3072] f8869a2ff3e8
val_1487_axes1 INT64[1] 7c9fa136d441
val_1487_eye FLOAT[77,77] 39ccea14e08c
val_15 FLOAT[1024,4096] 980e1c9bfb40
val_16 FLOAT[4096,1024] 4a0f32b2c394
val_17 FLOAT[1024,3072] 98ed7dd51e18
val_18 FLOAT[1024,4096] 681950f57c2d
val_19 FLOAT[4096,1024] 08af247b6c83
val_2 FLOAT[1024,3072] 064503b0c348
val_20 FLOAT[1024,3072] 5119cb35dca4
val_21 FLOAT[1024,4096] b5a2e5bb3313
val_22 FLOAT[4096,1024] 3a145309fad6
val_23 FLOAT[1024,3072] 1eb1651af62d
val_24 FLOAT[1024,4096] 56c2888dbb66
val_25 FLOAT[4096,1024] 3c015c442277
val_26 FLOAT[1024,3072] 2fec8fb0891a
val_27 FLOAT[1024,4096] 5da28d022079
val_28 FLOAT[4096,1024] e4b1b068f36e
val_29 FLOAT[1024,3072] a8a3621b3ad8
val_3 FLOAT[1024,4096] 61415faf29b5
val_30 FLOAT[1024,4096] 5b61b9b5e8c3
val_31 FLOAT[4096,1024] dc9c4582e2f7
val_32 FLOAT[1024,3072] be31ec8ee68c
val_33 FLOAT[1024,4096] 7160b5f0b71c
val_34 FLOAT[4096,1024] 3402979eb90c
val_35 FLOAT[1024,3072] 5c8f713c0f1e
val_36 FLOAT[1024,4096] 28861dce6b31
val_37 FLOAT[4096,1024] ebdf9de1ed2d
val_38 FLOAT[1024,3072] b75e86d55a4b
val_39 FLOAT[1024,4096] 51420c00742f
val_4 FLOAT[4096,1024] 6e5ae412a24c
val_40 FLOAT[4096,1024] b7cc6264e9c5
val_41 FLOAT[1024,3072] 4ef899e695c8
val_42 FLOAT[1024,4096] 69e7af62c055
val_43 FLOAT[4096,1024] 036bc269d2b6
val_44 FLOAT[1024,3072] d945bcfd93a9
val_45 FLOAT[1024,4096] e9b7ff2b3cb5
val_46 FLOAT[4096,1024] e672f016f602
val_47 FLOAT[1024,3072] 7639e838e5b8
val_48 FLOAT[1024,4096] b8b4445b3d14
val_49 FLOAT[4096,1024] b8dabdddb024
val_5 FLOAT[1024,3072] f6a968f2d407
val_50 FLOAT[1024,3072] 75685f7e3e22
val_51 FLOAT[1024,4096] d45dbf80d4cd
val_52 FLOAT[4096,1024] 54546767e96b
val_53 FLOAT[1024,3072] 86e2eb625021
val_54 FLOAT[1024,4096] 342c4d089961
val_55 FLOAT[4096,1024] 1a87899063f1
val_56 FLOAT[1024,3072] 12bf4844113a
val_57 FLOAT[1024,4096] 37c59a8bcd9f
val_58 FLOAT[4096,1024] 32620d28e9a1
val_59 FLOAT[1024,3072] 195e835ee930
val_6 FLOAT[1024,4096] 9aa58efb6f46
val_60 FLOAT[1024,4096] 88a6ddaef7c3
val_61 FLOAT[4096,1024] 9e37967a4ade
val_62 FLOAT[1024,3072] f0df899a3d4e
val_63 FLOAT[1024,4096] 176197355b9a
val_64 FLOAT[4096,1024] 0b9e8274e60e
val_65 FLOAT[1024,3072] 5e026056ca99
val_66 FLOAT[1024,4096] 7ebfd9b896e2
val_67 FLOAT[4096,1024] 85f6bb83db2a
val_68 FLOAT[1024,3072] 6d16469cdb53
val_69 FLOAT[1024,4096] 7084d9787954
val_7 FLOAT[4096,1024] 767b245a7d97
val_70 FLOAT[4096,1024] 89a1a950c61b
val_71 FLOAT[1024,3072] ca4fe23ae8d8
val_72 FLOAT[1024,4096] 2d68207ff3d8
val_73 FLOAT[4096,1024] 8c627527b033
val_8 FLOAT[1024,3072] cafd15b98a6f
val_9 FLOAT[1024,4096] a6f0f76d5fbf
