<
   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_1012
      float[batch,77,1024] add_1127
      float[batch,77,1024] add_1156
      float[batch,77,1024] add_119
      float[batch,77,1024] add_1271
      float[batch,77,1024] add_1300
      float[batch,77,1024] add_1415
      float[batch,77,1024] add_1444
      float[batch,77,1024] add_148
      float[batch,77,1024] add_1559
      float[batch,77,1024] add_1588
      float[batch,77,1024] add_1703
      float[batch,77,1024] add_1732
      float[batch,77,1024] add_1847
      float[batch,77,1024] add_1876
      float[batch,77,1024] add_1991
      float[batch,77,1024] add_2020
      float[batch,77,1024] add_2135
      float[batch,77,1024] add_2164
      float[batch,77,1024] add_2279
      float[batch,77,1024] add_2308
      float[batch,77,1024] add_2423
      float[batch,77,1024] add_2452
      float[batch,77,1024] add_2567
      float[batch,77,1024] add_2596
      float[batch,77,1024] add_263
      float[batch,77,1024] add_2711
      float[batch,77,1024] add_2740
      float[batch,77,1024] add_2855
      float[batch,77,1024] add_2884
      float[batch,77,1024] add_292
      float[batch,77,1024] add_2999
      float[batch,77,1024] add_3028
      float[batch,77,1024] add_3143
      float[batch,77,1024] add_3172
      float[batch,77,1024] add_3287
      float[batch,77,1024] add_3316
      float[batch,1,1024] add_3316_pooled
      float[batch,1,1024] add_3431
      float[batch,1,1024] add_3460
      float[batch,77,1024] add_4
      float[batch,77,1024] add_407
      float[batch,77,1024] add_436
      float[batch,77,1024] add_551
      float[batch,77,1024] add_580
      float[batch,77,1024] add_695
      float[batch,77,1024] add_724
      float[batch,77,1024] add_839
      float[batch,77,1024] add_868
      float[batch,77,1024] add_983
      int64[batch] argmax
      float[batch,1] clamp_min
      float[batch,77,1024] embedding
      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,77,4096] mul_101
      float[batch,77,4096] mul_106
      float[batch,77,4096] mul_1064
      float[batch,77,4096] mul_1069
      float[batch,77,4096] mul_1171
      float[batch,77,4096] mul_1176
      float[batch,77,4096] mul_1278
      float[batch,77,4096] mul_1283
      float[batch,77,4096] mul_1385
      float[batch,77,4096] mul_1390
      float[batch,77,4096] mul_1492
      float[batch,77,4096] mul_1497
      float[batch,77,4096] mul_1599
      float[batch,77,4096] mul_1604
      float[batch,77,4096] mul_1706
      float[batch,77,4096] mul_1711
      float[batch,77,4096] mul_1813
      float[batch,77,4096] mul_1818
      float[batch,77,4096] mul_1920
      float[batch,77,4096] mul_1925
      float[batch,77,4096] mul_2027
      float[batch,77,4096] mul_2032
      float[batch,77,4096] mul_208
      float[batch,77,4096] mul_213
      float[batch,77,4096] mul_2134
      float[batch,77,4096] mul_2139
      float[batch,77,4096] mul_2241
      float[batch,77,4096] mul_2246
      float[batch,77,4096] mul_2348
      float[batch,77,4096] mul_2353
      float[batch,77,4096] mul_2455
      float[batch,77,4096] mul_2460
      float[batch,1,4096] mul_2562
      float[batch,1,4096] mul_2567
      float[batch,77,4096] mul_315
      float[batch,77,4096] mul_320
      float[batch,77,4096] mul_422
      float[batch,77,4096] mul_427
      float[batch,77,4096] mul_529
      float[batch,77,4096] mul_534
      float[batch,77,4096] mul_636
      float[batch,77,4096] mul_641
      float[batch,77,4096] mul_743
      float[batch,77,4096] mul_748
      float[batch,77,4096] mul_850
      float[batch,77,4096] mul_855
      float[batch,77,4096] mul_957
      float[batch,77,4096] mul_962
      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,77,4096] sigmoid
      float[batch,77,4096] sigmoid_1
      float[batch,77,4096] sigmoid_10
      float[batch,77,4096] sigmoid_11
      float[batch,77,4096] sigmoid_12
      float[batch,77,4096] sigmoid_13
      float[batch,77,4096] sigmoid_14
      float[batch,77,4096] sigmoid_15
      float[batch,77,4096] sigmoid_16
      float[batch,77,4096] sigmoid_17
      float[batch,77,4096] sigmoid_18
      float[batch,77,4096] sigmoid_19
      float[batch,77,4096] sigmoid_2
      float[batch,77,4096] sigmoid_20
      float[batch,77,4096] sigmoid_21
      float[batch,77,4096] sigmoid_22
      float[batch,1,4096] sigmoid_23
      float[batch,77,4096] sigmoid_3
      float[batch,77,4096] sigmoid_4
      float[batch,77,4096] sigmoid_5
      float[batch,77,4096] sigmoid_6
      float[batch,77,4096] sigmoid_7
      float[batch,77,4096] sigmoid_8
      float[batch,77,4096] sigmoid_9
      float[batch,77,4096] val_100
      float[batch,77,1024] val_101
      float[batch,77,4096] val_102
      float[batch,77,1024] val_103
      float[batch,77,4096] val_104
      float[batch,77,1024] val_105
      float[batch,77,4096] val_106
      float[batch,77,1024] val_107
      float[batch,77,4096] val_108
      float[batch,77,1024] val_109
      float[batch,77,4096] val_110
      float[batch,77,1024] val_111
      float[batch,77,4096] val_112
      float[batch,77,1024] val_113
      float[batch,77,4096] val_114
      float[batch,77,1024] val_115
      float[batch,77,4096] val_116
      float[batch,77,1024] val_117
      float[batch,77,4096] val_118
      float[batch,77,1024] val_119
      float[batch,77,4096] val_120
      float[batch,77,1024] val_121
      float[batch,77] val_122
      float[batch,1,77] val_123
      float[batch,1,4096] val_124
      float[batch,1,1024] val_125
      float[batch,1,1024] val_126
      float[batch,1024] val_127
      float[batch,77,4096] val_76
      float[batch,77,1024] val_77
      float[batch,77,4096] val_78
      float[batch,77,1024] val_79
      float[batch,77,4096] val_80
      float[batch,77,1024] val_81
      float[batch,77,4096] val_82
      float[batch,77,1024] val_83
      float[batch,77,4096] val_84
      float[batch,77,1024] val_85
      float[batch,77,4096] val_86
      float[batch,77,1024] val_87
      float[batch,77,4096] val_88
      float[batch,77,1024] val_89
      float[batch,77,4096] val_90
      float[batch,77,1024] val_91
      float[batch,77,4096] val_92
      float[batch,77,1024] val_93
      float[batch,77,4096] val_94
      float[batch,77,1024] val_95
      float[batch,77,4096] val_96
      float[batch,77,1024] val_97
      float[batch,77,4096] val_98
      float[batch,77,1024] val_99
   >
{
   val_75 = Gather <axis: int = 0> ("token_embedding.weight_fp16", text)
   embedding = Cast <to: int = 1> (val_75)
   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_3)
   [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_76 = MatMul (layer_norm_1, val_4)
   linear_2 = Add (val_76, "transformer.resblocks.0.mlp.c_fc.bias")
   mul_101 = Mul (linear_2, val_2)
   [node_sigmoid] sigmoid = Sigmoid (mul_101)
   mul_106 = Mul (linear_2, sigmoid)
   val_77 = MatMul (mul_106, val_5)
   linear_3 = Add (val_77, "transformer.resblocks.0.mlp.c_proj.bias")
   add_148 = Add (add_119, linear_3)
   layer_norm_2 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_148, "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_6)
   [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_263 = Add (add_148, node_scaled_dot_product_attention_1_out)
   layer_norm_3 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_263, "transformer.resblocks.1.ln_2.weight", "transformer.resblocks.1.ln_2.bias")
   val_78 = MatMul (layer_norm_3, val_7)
   linear_6 = Add (val_78, "transformer.resblocks.1.mlp.c_fc.bias")
   mul_208 = Mul (linear_6, val_2)
   sigmoid_1 = Sigmoid (mul_208)
   mul_213 = Mul (linear_6, sigmoid_1)
   val_79 = MatMul (mul_213, val_8)
   linear_7 = Add (val_79, "transformer.resblocks.1.mlp.c_proj.bias")
   add_292 = Add (add_263, linear_7)
   layer_norm_4 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_292, "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_9)
   [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_407 = Add (add_292, node_scaled_dot_product_attention_2_out)
   layer_norm_5 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_407, "transformer.resblocks.2.ln_2.weight", "transformer.resblocks.2.ln_2.bias")
   val_80 = MatMul (layer_norm_5, val_10)
   linear_10 = Add (val_80, "transformer.resblocks.2.mlp.c_fc.bias")
   mul_315 = Mul (linear_10, val_2)
   sigmoid_2 = Sigmoid (mul_315)
   mul_320 = Mul (linear_10, sigmoid_2)
   val_81 = MatMul (mul_320, val_11)
   linear_11 = Add (val_81, "transformer.resblocks.2.mlp.c_proj.bias")
   add_436 = Add (add_407, linear_11)
   layer_norm_6 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_436, "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_12)
   [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_551 = Add (add_436, node_scaled_dot_product_attention_3_out)
   layer_norm_7 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_551, "transformer.resblocks.3.ln_2.weight", "transformer.resblocks.3.ln_2.bias")
   val_82 = MatMul (layer_norm_7, val_13)
   linear_14 = Add (val_82, "transformer.resblocks.3.mlp.c_fc.bias")
   mul_422 = Mul (linear_14, val_2)
   sigmoid_3 = Sigmoid (mul_422)
   mul_427 = Mul (linear_14, sigmoid_3)
   val_83 = MatMul (mul_427, val_14)
   linear_15 = Add (val_83, "transformer.resblocks.3.mlp.c_proj.bias")
   add_580 = Add (add_551, linear_15)
   layer_norm_8 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_580, "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_15)
   [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_695 = Add (add_580, node_scaled_dot_product_attention_4_out)
   layer_norm_9 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_695, "transformer.resblocks.4.ln_2.weight", "transformer.resblocks.4.ln_2.bias")
   val_84 = MatMul (layer_norm_9, val_16)
   linear_18 = Add (val_84, "transformer.resblocks.4.mlp.c_fc.bias")
   mul_529 = Mul (linear_18, val_2)
   sigmoid_4 = Sigmoid (mul_529)
   mul_534 = Mul (linear_18, sigmoid_4)
   val_85 = MatMul (mul_534, val_17)
   linear_19 = Add (val_85, "transformer.resblocks.4.mlp.c_proj.bias")
   add_724 = Add (add_695, linear_19)
   layer_norm_10 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_724, "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_18)
   [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_839 = Add (add_724, node_scaled_dot_product_attention_5_out)
   layer_norm_11 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_839, "transformer.resblocks.5.ln_2.weight", "transformer.resblocks.5.ln_2.bias")
   val_86 = MatMul (layer_norm_11, val_19)
   linear_22 = Add (val_86, "transformer.resblocks.5.mlp.c_fc.bias")
   mul_636 = Mul (linear_22, val_2)
   sigmoid_5 = Sigmoid (mul_636)
   mul_641 = Mul (linear_22, sigmoid_5)
   val_87 = MatMul (mul_641, val_20)
   linear_23 = Add (val_87, "transformer.resblocks.5.mlp.c_proj.bias")
   add_868 = Add (add_839, linear_23)
   layer_norm_12 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_868, "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_21)
   [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_983 = Add (add_868, node_scaled_dot_product_attention_6_out)
   layer_norm_13 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_983, "transformer.resblocks.6.ln_2.weight", "transformer.resblocks.6.ln_2.bias")
   val_88 = MatMul (layer_norm_13, val_22)
   linear_26 = Add (val_88, "transformer.resblocks.6.mlp.c_fc.bias")
   mul_743 = Mul (linear_26, val_2)
   sigmoid_6 = Sigmoid (mul_743)
   mul_748 = Mul (linear_26, sigmoid_6)
   val_89 = MatMul (mul_748, val_23)
   linear_27 = Add (val_89, "transformer.resblocks.6.mlp.c_proj.bias")
   add_1012 = Add (add_983, linear_27)
   layer_norm_14 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1012, "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_24)
   [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_1127 = Add (add_1012, node_scaled_dot_product_attention_7_out)
   layer_norm_15 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1127, "transformer.resblocks.7.ln_2.weight", "transformer.resblocks.7.ln_2.bias")
   val_90 = MatMul (layer_norm_15, val_25)
   linear_30 = Add (val_90, "transformer.resblocks.7.mlp.c_fc.bias")
   mul_850 = Mul (linear_30, val_2)
   sigmoid_7 = Sigmoid (mul_850)
   mul_855 = Mul (linear_30, sigmoid_7)
   val_91 = MatMul (mul_855, val_26)
   linear_31 = Add (val_91, "transformer.resblocks.7.mlp.c_proj.bias")
   add_1156 = Add (add_1127, linear_31)
   layer_norm_16 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1156, "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_27)
   [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_1271 = Add (add_1156, node_scaled_dot_product_attention_8_out)
   layer_norm_17 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1271, "transformer.resblocks.8.ln_2.weight", "transformer.resblocks.8.ln_2.bias")
   val_92 = MatMul (layer_norm_17, val_28)
   linear_34 = Add (val_92, "transformer.resblocks.8.mlp.c_fc.bias")
   mul_957 = Mul (linear_34, val_2)
   sigmoid_8 = Sigmoid (mul_957)
   mul_962 = Mul (linear_34, sigmoid_8)
   val_93 = MatMul (mul_962, val_29)
   linear_35 = Add (val_93, "transformer.resblocks.8.mlp.c_proj.bias")
   add_1300 = Add (add_1271, linear_35)
   layer_norm_18 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1300, "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_30)
   [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_1415 = Add (add_1300, node_scaled_dot_product_attention_9_out)
   layer_norm_19 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1415, "transformer.resblocks.9.ln_2.weight", "transformer.resblocks.9.ln_2.bias")
   val_94 = MatMul (layer_norm_19, val_31)
   linear_38 = Add (val_94, "transformer.resblocks.9.mlp.c_fc.bias")
   mul_1064 = Mul (linear_38, val_2)
   sigmoid_9 = Sigmoid (mul_1064)
   mul_1069 = Mul (linear_38, sigmoid_9)
   val_95 = MatMul (mul_1069, val_32)
   linear_39 = Add (val_95, "transformer.resblocks.9.mlp.c_proj.bias")
   add_1444 = Add (add_1415, linear_39)
   layer_norm_20 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1444, "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_33)
   [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_1559 = Add (add_1444, node_scaled_dot_product_attention_10_out)
   layer_norm_21 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1559, "transformer.resblocks.10.ln_2.weight", "transformer.resblocks.10.ln_2.bias")
   val_96 = MatMul (layer_norm_21, val_34)
   linear_42 = Add (val_96, "transformer.resblocks.10.mlp.c_fc.bias")
   mul_1171 = Mul (linear_42, val_2)
   sigmoid_10 = Sigmoid (mul_1171)
   mul_1176 = Mul (linear_42, sigmoid_10)
   val_97 = MatMul (mul_1176, val_35)
   linear_43 = Add (val_97, "transformer.resblocks.10.mlp.c_proj.bias")
   add_1588 = Add (add_1559, linear_43)
   layer_norm_22 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1588, "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_36)
   [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_1703 = Add (add_1588, node_scaled_dot_product_attention_11_out)
   layer_norm_23 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1703, "transformer.resblocks.11.ln_2.weight", "transformer.resblocks.11.ln_2.bias")
   val_98 = MatMul (layer_norm_23, val_37)
   linear_46 = Add (val_98, "transformer.resblocks.11.mlp.c_fc.bias")
   mul_1278 = Mul (linear_46, val_2)
   sigmoid_11 = Sigmoid (mul_1278)
   mul_1283 = Mul (linear_46, sigmoid_11)
   val_99 = MatMul (mul_1283, val_38)
   linear_47 = Add (val_99, "transformer.resblocks.11.mlp.c_proj.bias")
   add_1732 = Add (add_1703, linear_47)
   layer_norm_24 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1732, "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_39)
   [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_1847 = Add (add_1732, node_scaled_dot_product_attention_12_out)
   layer_norm_25 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1847, "transformer.resblocks.12.ln_2.weight", "transformer.resblocks.12.ln_2.bias")
   val_100 = MatMul (layer_norm_25, val_40)
   linear_50 = Add (val_100, "transformer.resblocks.12.mlp.c_fc.bias")
   mul_1385 = Mul (linear_50, val_2)
   sigmoid_12 = Sigmoid (mul_1385)
   mul_1390 = Mul (linear_50, sigmoid_12)
   val_101 = MatMul (mul_1390, val_41)
   linear_51 = Add (val_101, "transformer.resblocks.12.mlp.c_proj.bias")
   add_1876 = Add (add_1847, linear_51)
   layer_norm_26 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1876, "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_42)
   [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_1991 = Add (add_1876, node_scaled_dot_product_attention_13_out)
   layer_norm_27 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1991, "transformer.resblocks.13.ln_2.weight", "transformer.resblocks.13.ln_2.bias")
   val_102 = MatMul (layer_norm_27, val_43)
   linear_54 = Add (val_102, "transformer.resblocks.13.mlp.c_fc.bias")
   mul_1492 = Mul (linear_54, val_2)
   sigmoid_13 = Sigmoid (mul_1492)
   mul_1497 = Mul (linear_54, sigmoid_13)
   val_103 = MatMul (mul_1497, val_44)
   linear_55 = Add (val_103, "transformer.resblocks.13.mlp.c_proj.bias")
   add_2020 = Add (add_1991, linear_55)
   layer_norm_28 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2020, "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_45)
   [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_2135 = Add (add_2020, node_scaled_dot_product_attention_14_out)
   layer_norm_29 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2135, "transformer.resblocks.14.ln_2.weight", "transformer.resblocks.14.ln_2.bias")
   val_104 = MatMul (layer_norm_29, val_46)
   linear_58 = Add (val_104, "transformer.resblocks.14.mlp.c_fc.bias")
   mul_1599 = Mul (linear_58, val_2)
   sigmoid_14 = Sigmoid (mul_1599)
   mul_1604 = Mul (linear_58, sigmoid_14)
   val_105 = MatMul (mul_1604, val_47)
   linear_59 = Add (val_105, "transformer.resblocks.14.mlp.c_proj.bias")
   add_2164 = Add (add_2135, linear_59)
   layer_norm_30 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2164, "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_48)
   [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_2279 = Add (add_2164, node_scaled_dot_product_attention_15_out)
   layer_norm_31 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2279, "transformer.resblocks.15.ln_2.weight", "transformer.resblocks.15.ln_2.bias")
   val_106 = MatMul (layer_norm_31, val_49)
   linear_62 = Add (val_106, "transformer.resblocks.15.mlp.c_fc.bias")
   mul_1706 = Mul (linear_62, val_2)
   sigmoid_15 = Sigmoid (mul_1706)
   mul_1711 = Mul (linear_62, sigmoid_15)
   val_107 = MatMul (mul_1711, val_50)
   linear_63 = Add (val_107, "transformer.resblocks.15.mlp.c_proj.bias")
   add_2308 = Add (add_2279, linear_63)
   layer_norm_32 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2308, "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_51)
   [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_2423 = Add (add_2308, node_scaled_dot_product_attention_16_out)
   layer_norm_33 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2423, "transformer.resblocks.16.ln_2.weight", "transformer.resblocks.16.ln_2.bias")
   val_108 = MatMul (layer_norm_33, val_52)
   linear_66 = Add (val_108, "transformer.resblocks.16.mlp.c_fc.bias")
   mul_1813 = Mul (linear_66, val_2)
   sigmoid_16 = Sigmoid (mul_1813)
   mul_1818 = Mul (linear_66, sigmoid_16)
   val_109 = MatMul (mul_1818, val_53)
   linear_67 = Add (val_109, "transformer.resblocks.16.mlp.c_proj.bias")
   add_2452 = Add (add_2423, linear_67)
   layer_norm_34 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2452, "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_54)
   [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_2567 = Add (add_2452, node_scaled_dot_product_attention_17_out)
   layer_norm_35 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2567, "transformer.resblocks.17.ln_2.weight", "transformer.resblocks.17.ln_2.bias")
   val_110 = MatMul (layer_norm_35, val_55)
   linear_70 = Add (val_110, "transformer.resblocks.17.mlp.c_fc.bias")
   mul_1920 = Mul (linear_70, val_2)
   sigmoid_17 = Sigmoid (mul_1920)
   mul_1925 = Mul (linear_70, sigmoid_17)
   val_111 = MatMul (mul_1925, val_56)
   linear_71 = Add (val_111, "transformer.resblocks.17.mlp.c_proj.bias")
   add_2596 = Add (add_2567, linear_71)
   layer_norm_36 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2596, "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_57)
   [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_2711 = Add (add_2596, node_scaled_dot_product_attention_18_out)
   layer_norm_37 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2711, "transformer.resblocks.18.ln_2.weight", "transformer.resblocks.18.ln_2.bias")
   val_112 = MatMul (layer_norm_37, val_58)
   linear_74 = Add (val_112, "transformer.resblocks.18.mlp.c_fc.bias")
   mul_2027 = Mul (linear_74, val_2)
   sigmoid_18 = Sigmoid (mul_2027)
   mul_2032 = Mul (linear_74, sigmoid_18)
   val_113 = MatMul (mul_2032, val_59)
   linear_75 = Add (val_113, "transformer.resblocks.18.mlp.c_proj.bias")
   add_2740 = Add (add_2711, linear_75)
   layer_norm_38 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2740, "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_60)
   [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_2855 = Add (add_2740, node_scaled_dot_product_attention_19_out)
   layer_norm_39 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2855, "transformer.resblocks.19.ln_2.weight", "transformer.resblocks.19.ln_2.bias")
   val_114 = MatMul (layer_norm_39, val_61)
   linear_78 = Add (val_114, "transformer.resblocks.19.mlp.c_fc.bias")
   mul_2134 = Mul (linear_78, val_2)
   sigmoid_19 = Sigmoid (mul_2134)
   mul_2139 = Mul (linear_78, sigmoid_19)
   val_115 = MatMul (mul_2139, val_62)
   linear_79 = Add (val_115, "transformer.resblocks.19.mlp.c_proj.bias")
   add_2884 = Add (add_2855, linear_79)
   layer_norm_40 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2884, "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_63)
   [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_2999 = Add (add_2884, node_scaled_dot_product_attention_20_out)
   layer_norm_41 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2999, "transformer.resblocks.20.ln_2.weight", "transformer.resblocks.20.ln_2.bias")
   val_116 = MatMul (layer_norm_41, val_64)
   linear_82 = Add (val_116, "transformer.resblocks.20.mlp.c_fc.bias")
   mul_2241 = Mul (linear_82, val_2)
   sigmoid_20 = Sigmoid (mul_2241)
   mul_2246 = Mul (linear_82, sigmoid_20)
   val_117 = MatMul (mul_2246, val_65)
   linear_83 = Add (val_117, "transformer.resblocks.20.mlp.c_proj.bias")
   add_3028 = Add (add_2999, linear_83)
   layer_norm_42 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_3028, "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_66)
   [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_3143 = Add (add_3028, node_scaled_dot_product_attention_21_out)
   layer_norm_43 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_3143, "transformer.resblocks.21.ln_2.weight", "transformer.resblocks.21.ln_2.bias")
   val_118 = MatMul (layer_norm_43, val_67)
   linear_86 = Add (val_118, "transformer.resblocks.21.mlp.c_fc.bias")
   mul_2348 = Mul (linear_86, val_2)
   sigmoid_21 = Sigmoid (mul_2348)
   mul_2353 = Mul (linear_86, sigmoid_21)
   val_119 = MatMul (mul_2353, val_68)
   linear_87 = Add (val_119, "transformer.resblocks.21.mlp.c_proj.bias")
   add_3172 = Add (add_3143, linear_87)
   layer_norm_44 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_3172, "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_69)
   [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_3287 = Add (add_3172, node_scaled_dot_product_attention_22_out)
   layer_norm_45 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_3287, "transformer.resblocks.22.ln_2.weight", "transformer.resblocks.22.ln_2.bias")
   val_120 = MatMul (layer_norm_45, val_70)
   linear_90 = Add (val_120, "transformer.resblocks.22.mlp.c_fc.bias")
   mul_2455 = Mul (linear_90, val_2)
   sigmoid_22 = Sigmoid (mul_2455)
   mul_2460 = Mul (linear_90, sigmoid_22)
   val_121 = MatMul (mul_2460, val_71)
   linear_91 = Add (val_121, "transformer.resblocks.22.mlp.c_proj.bias")
   add_3316 = Add (add_3287, linear_91)
   layer_norm_46 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_3316, "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_72)
   [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_122 = Gather <axis: int = 0> (val_1488_eye, argmax)
   val_123 = Unsqueeze (val_122, val_1488_axes1)
   [pool_hoist_scaled_dot_product_attention_23] scaled_dot_product_attention_23_pooled = MatMul (val_123, 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_3316] add_3316_pooled = MatMul (val_123, add_3316)
   add_3431 = Add (add_3316_pooled, node_scaled_dot_product_attention_23_out)
   layer_norm_47 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_3431, "transformer.resblocks.23.ln_2.weight", "transformer.resblocks.23.ln_2.bias")
   val_124 = MatMul (layer_norm_47, val_73)
   linear_94 = Add (val_124, "transformer.resblocks.23.mlp.c_fc.bias")
   mul_2562 = Mul (linear_94, val_2)
   sigmoid_23 = Sigmoid (mul_2562)
   mul_2567 = Mul (linear_94, sigmoid_23)
   val_125 = MatMul (mul_2567, val_74)
   linear_95 = Add (val_125, "transformer.resblocks.23.mlp.c_proj.bias")
   add_3460 = Add (add_3431, linear_95)
   val_126 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_3460, "ln_final.weight", "ln_final.bias")
   val_127 = Squeeze (val_126, val_1488_axes1)
   [node_matmul] matmul = MatMul (val_127, 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] d21e115a39f4
ln_final.weight FLOAT[1024] ba6636dafb40
node_scaled_dot_product_attention_10_wo_t FLOAT[1024,1024] 51ad6e2a5cd8
node_scaled_dot_product_attention_11_wo_t FLOAT[1024,1024] d976b59cbd05
node_scaled_dot_product_attention_12_wo_t FLOAT[1024,1024] 5d0cc3a82a64
node_scaled_dot_product_attention_13_wo_t FLOAT[1024,1024] 377bb504aba0
node_scaled_dot_product_attention_14_wo_t FLOAT[1024,1024] 99a35d01b88d
node_scaled_dot_product_attention_15_wo_t FLOAT[1024,1024] fd245c7c5c6b
node_scaled_dot_product_attention_16_wo_t FLOAT[1024,1024] bdd283f70184
node_scaled_dot_product_attention_17_wo_t FLOAT[1024,1024] 867f38f438d3
node_scaled_dot_product_attention_18_wo_t FLOAT[1024,1024] e1df18d51eec
node_scaled_dot_product_attention_19_wo_t FLOAT[1024,1024] 820193b243ea
node_scaled_dot_product_attention_1_wo_t FLOAT[1024,1024] 6e257e7aca54
node_scaled_dot_product_attention_20_wo_t FLOAT[1024,1024] f9b56bd8340e
node_scaled_dot_product_attention_21_wo_t FLOAT[1024,1024] 686f3fc8c777
node_scaled_dot_product_attention_22_wo_t FLOAT[1024,1024] 8eafc9bc942d
node_scaled_dot_product_attention_23_wo_t FLOAT[1024,1024] eaca287d99d0
node_scaled_dot_product_attention_2_wo_t FLOAT[1024,1024] 9bd7669820fb
node_scaled_dot_product_attention_3_wo_t FLOAT[1024,1024] a34434fa02b1
node_scaled_dot_product_attention_4_wo_t FLOAT[1024,1024] b1e70f3f27f9
node_scaled_dot_product_attention_5_wo_t FLOAT[1024,1024] c739a92bf191
node_scaled_dot_product_attention_6_wo_t FLOAT[1024,1024] 1e5f6220488b
node_scaled_dot_product_attention_7_wo_t FLOAT[1024,1024] 6391a5ce157a
node_scaled_dot_product_attention_8_wo_t FLOAT[1024,1024] eed104011781
node_scaled_dot_product_attention_9_wo_t FLOAT[1024,1024] fcc6b7ce89f7
node_scaled_dot_product_attention_wo_t FLOAT[1024,1024] 6ed32f4fde24
positional_embedding FLOAT[77,1024] 810ac30c837f
text_projection FLOAT[1024,1024] e4645fc3606d
token_embedding.weight_fp16 FLOAT16[49408,1024] 761e5798bd90
transformer.resblocks.0.attn.in_proj_bias FLOAT[3072] 8e697e60368b
transformer.resblocks.0.attn.out_proj.bias FLOAT[1024] 7fda9c08164c
transformer.resblocks.0.ln_1.bias FLOAT[1024] 3068d085c936
transformer.resblocks.0.ln_1.weight FLOAT[1024] a61c2e1808ae
transformer.resblocks.0.ln_2.bias FLOAT[1024] 8642b266f60e
transformer.resblocks.0.ln_2.weight FLOAT[1024] c71f690c8231
transformer.resblocks.0.mlp.c_fc.bias FLOAT[4096] 790a8c6a02f3
transformer.resblocks.0.mlp.c_proj.bias FLOAT[1024] efc25ebeaded
transformer.resblocks.1.attn.in_proj_bias FLOAT[3072] bf3c80bdfb88
transformer.resblocks.1.attn.out_proj.bias FLOAT[1024] 35bb163e2a23
transformer.resblocks.1.ln_1.bias FLOAT[1024] e9a23516abd8
transformer.resblocks.1.ln_1.weight FLOAT[1024] 7a31938ab848
transformer.resblocks.1.ln_2.bias FLOAT[1024] 2c0f11f93c8c
transformer.resblocks.1.ln_2.weight FLOAT[1024] 4ef1c7d5de42
transformer.resblocks.1.mlp.c_fc.bias FLOAT[4096] 8c713ff07180
transformer.resblocks.1.mlp.c_proj.bias FLOAT[1024] 35fa81435228
transformer.resblocks.10.attn.in_proj_bias FLOAT[3072] ef5b7c42bd9f
transformer.resblocks.10.attn.out_proj.bias FLOAT[1024] 34c21b49a1c3
transformer.resblocks.10.ln_1.bias FLOAT[1024] fe7b822a23fe
transformer.resblocks.10.ln_1.weight FLOAT[1024] b54afd8a423a
transformer.resblocks.10.ln_2.bias FLOAT[1024] 6f0aba43711f
transformer.resblocks.10.ln_2.weight FLOAT[1024] 07c9710f05e3
transformer.resblocks.10.mlp.c_fc.bias FLOAT[4096] 3395cf7b25df
transformer.resblocks.10.mlp.c_proj.bias FLOAT[1024] 113dafe024d4
transformer.resblocks.11.attn.in_proj_bias FLOAT[3072] 81bb97f42a55
transformer.resblocks.11.attn.out_proj.bias FLOAT[1024] 2608f98b8549
transformer.resblocks.11.ln_1.bias FLOAT[1024] e12ab98a7f40
transformer.resblocks.11.ln_1.weight FLOAT[1024] 88dffb3da2fb
transformer.resblocks.11.ln_2.bias FLOAT[1024] d97dcc4a14c7
transformer.resblocks.11.ln_2.weight FLOAT[1024] bb4bba9a153f
transformer.resblocks.11.mlp.c_fc.bias FLOAT[4096] 53800d6d5654
transformer.resblocks.11.mlp.c_proj.bias FLOAT[1024] a496bb4397ab
transformer.resblocks.12.attn.in_proj_bias FLOAT[3072] 49509eb83101
transformer.resblocks.12.attn.out_proj.bias FLOAT[1024] adcdfd1d8367
transformer.resblocks.12.ln_1.bias FLOAT[1024] c855d03f5c2b
transformer.resblocks.12.ln_1.weight FLOAT[1024] 95f7e81939d4
transformer.resblocks.12.ln_2.bias FLOAT[1024] 8e0c9827d996
transformer.resblocks.12.ln_2.weight FLOAT[1024] 25a15596cc15
transformer.resblocks.12.mlp.c_fc.bias FLOAT[4096] b39220a046a9
transformer.resblocks.12.mlp.c_proj.bias FLOAT[1024] d710aa323a86
transformer.resblocks.13.attn.in_proj_bias FLOAT[3072] ee9a25adc238
transformer.resblocks.13.attn.out_proj.bias FLOAT[1024] b61d098f702d
transformer.resblocks.13.ln_1.bias FLOAT[1024] 46ba691b49f9
transformer.resblocks.13.ln_1.weight FLOAT[1024] ab1336e09fa6
transformer.resblocks.13.ln_2.bias FLOAT[1024] 90a17a57f3f9
transformer.resblocks.13.ln_2.weight FLOAT[1024] 28b6b24bd168
transformer.resblocks.13.mlp.c_fc.bias FLOAT[4096] 5737d34e74aa
transformer.resblocks.13.mlp.c_proj.bias FLOAT[1024] 86bd4d782884
transformer.resblocks.14.attn.in_proj_bias FLOAT[3072] 0b0965ce28bb
transformer.resblocks.14.attn.out_proj.bias FLOAT[1024] ec016211ae4e
transformer.resblocks.14.ln_1.bias FLOAT[1024] 9235567e217d
transformer.resblocks.14.ln_1.weight FLOAT[1024] 15b2cf4fed3e
transformer.resblocks.14.ln_2.bias FLOAT[1024] a2b7ab2b00fe
transformer.resblocks.14.ln_2.weight FLOAT[1024] 5e5ffaa23406
transformer.resblocks.14.mlp.c_fc.bias FLOAT[4096] c5cf920b536c
transformer.resblocks.14.mlp.c_proj.bias FLOAT[1024] 6848248793dc
transformer.resblocks.15.attn.in_proj_bias FLOAT[3072] 5c474ca46184
transformer.resblocks.15.attn.out_proj.bias FLOAT[1024] 1bec708da725
transformer.resblocks.15.ln_1.bias FLOAT[1024] 989ac344aa90
transformer.resblocks.15.ln_1.weight FLOAT[1024] 353fe321d141
transformer.resblocks.15.ln_2.bias FLOAT[1024] ac85633c89f8
transformer.resblocks.15.ln_2.weight FLOAT[1024] 7607c1545cb6
transformer.resblocks.15.mlp.c_fc.bias FLOAT[4096] 0bf7de16f84f
transformer.resblocks.15.mlp.c_proj.bias FLOAT[1024] 6702a37bfd19
transformer.resblocks.16.attn.in_proj_bias FLOAT[3072] a39ad1b7ff59
transformer.resblocks.16.attn.out_proj.bias FLOAT[1024] 5d551cdae004
transformer.resblocks.16.ln_1.bias FLOAT[1024] e604200cc926
transformer.resblocks.16.ln_1.weight FLOAT[1024] 05924a4cf46d
transformer.resblocks.16.ln_2.bias FLOAT[1024] 1246489940fb
transformer.resblocks.16.ln_2.weight FLOAT[1024] 34473835ac9d
transformer.resblocks.16.mlp.c_fc.bias FLOAT[4096] 29daed9b2201
transformer.resblocks.16.mlp.c_proj.bias FLOAT[1024] 3ced2380654f
transformer.resblocks.17.attn.in_proj_bias FLOAT[3072] 83b28ec197bc
transformer.resblocks.17.attn.out_proj.bias FLOAT[1024] 0a8c66d571fc
transformer.resblocks.17.ln_1.bias FLOAT[1024] bf80f8ba550c
transformer.resblocks.17.ln_1.weight FLOAT[1024] 905819b797de
transformer.resblocks.17.ln_2.bias FLOAT[1024] 7ef072f0965c
transformer.resblocks.17.ln_2.weight FLOAT[1024] 98f7a6caf8f2
transformer.resblocks.17.mlp.c_fc.bias FLOAT[4096] 2865f1ceda46
transformer.resblocks.17.mlp.c_proj.bias FLOAT[1024] 36b534ea8124
transformer.resblocks.18.attn.in_proj_bias FLOAT[3072] f784c1ec4e21
transformer.resblocks.18.attn.out_proj.bias FLOAT[1024] 20cab89ddf08
transformer.resblocks.18.ln_1.bias FLOAT[1024] eea219d2ae41
transformer.resblocks.18.ln_1.weight FLOAT[1024] 1d4438cce5a1
transformer.resblocks.18.ln_2.bias FLOAT[1024] 591a7efcbfb8
transformer.resblocks.18.ln_2.weight FLOAT[1024] 2c6cf985f463
transformer.resblocks.18.mlp.c_fc.bias FLOAT[4096] 1d051f81397f
transformer.resblocks.18.mlp.c_proj.bias FLOAT[1024] a927f9a7aff3
transformer.resblocks.19.attn.in_proj_bias FLOAT[3072] 65ef600d5043
transformer.resblocks.19.attn.out_proj.bias FLOAT[1024] 68be8468d140
transformer.resblocks.19.ln_1.bias FLOAT[1024] c840a111007e
transformer.resblocks.19.ln_1.weight FLOAT[1024] 00a421c76196
transformer.resblocks.19.ln_2.bias FLOAT[1024] ba9768d6bc08
transformer.resblocks.19.ln_2.weight FLOAT[1024] f0f0c7a5ab74
transformer.resblocks.19.mlp.c_fc.bias FLOAT[4096] aaa8ff7bf3a5
transformer.resblocks.19.mlp.c_proj.bias FLOAT[1024] 93ea55e0e937
transformer.resblocks.2.attn.in_proj_bias FLOAT[3072] aaa19532475e
transformer.resblocks.2.attn.out_proj.bias FLOAT[1024] 013d34dac126
transformer.resblocks.2.ln_1.bias FLOAT[1024] 0a31b0ac274e
transformer.resblocks.2.ln_1.weight FLOAT[1024] dd2d6b727af4
transformer.resblocks.2.ln_2.bias FLOAT[1024] b56d41d4148a
transformer.resblocks.2.ln_2.weight FLOAT[1024] 97c44049905a
transformer.resblocks.2.mlp.c_fc.bias FLOAT[4096] 03f722f70b80
transformer.resblocks.2.mlp.c_proj.bias FLOAT[1024] 1cd59696a3d5
transformer.resblocks.20.attn.in_proj_bias FLOAT[3072] b2302468aec6
transformer.resblocks.20.attn.out_proj.bias FLOAT[1024] bb8a769cd373
transformer.resblocks.20.ln_1.bias FLOAT[1024] e0f76d4a5e9b
transformer.resblocks.20.ln_1.weight FLOAT[1024] e3b7fa688e1e
transformer.resblocks.20.ln_2.bias FLOAT[1024] 626df08b18f2
transformer.resblocks.20.ln_2.weight FLOAT[1024] 950d279806fd
transformer.resblocks.20.mlp.c_fc.bias FLOAT[4096] ff6960129f34
transformer.resblocks.20.mlp.c_proj.bias FLOAT[1024] 4ea81cfb4cec
transformer.resblocks.21.attn.in_proj_bias FLOAT[3072] 754e209f95a6
transformer.resblocks.21.attn.out_proj.bias FLOAT[1024] bbfd67b977b9
transformer.resblocks.21.ln_1.bias FLOAT[1024] 07fec175b2b5
transformer.resblocks.21.ln_1.weight FLOAT[1024] 5efe7ccb9708
transformer.resblocks.21.ln_2.bias FLOAT[1024] 153478314d24
transformer.resblocks.21.ln_2.weight FLOAT[1024] 91349c31ce0b
transformer.resblocks.21.mlp.c_fc.bias FLOAT[4096] c89e06752007
transformer.resblocks.21.mlp.c_proj.bias FLOAT[1024] 6b2eb2746ee7
transformer.resblocks.22.attn.in_proj_bias FLOAT[3072] a9dc3cfdc947
transformer.resblocks.22.attn.out_proj.bias FLOAT[1024] ec2025b096ea
transformer.resblocks.22.ln_1.bias FLOAT[1024] a775dc0c6d27
transformer.resblocks.22.ln_1.weight FLOAT[1024] cc83311aed2a
transformer.resblocks.22.ln_2.bias FLOAT[1024] ccf84ac39267
transformer.resblocks.22.ln_2.weight FLOAT[1024] 47a30a135424
transformer.resblocks.22.mlp.c_fc.bias FLOAT[4096] b93629ba82b3
transformer.resblocks.22.mlp.c_proj.bias FLOAT[1024] 9a61c2d8d83a
transformer.resblocks.23.attn.in_proj_bias FLOAT[3072] 2c352c3af8db
transformer.resblocks.23.attn.out_proj.bias FLOAT[1024] a87a3884e8cf
transformer.resblocks.23.ln_1.bias FLOAT[1024] 553f1c229375
transformer.resblocks.23.ln_1.weight FLOAT[1024] 3d811d678c9c
transformer.resblocks.23.ln_2.bias FLOAT[1024] 78417832ff17
transformer.resblocks.23.ln_2.weight FLOAT[1024] 4aaad8b1422b
transformer.resblocks.23.mlp.c_fc.bias FLOAT[4096] 87fc17e2b460
transformer.resblocks.23.mlp.c_proj.bias FLOAT[1024] 1423b4da77db
transformer.resblocks.3.attn.in_proj_bias FLOAT[3072] caa3d39bc764
transformer.resblocks.3.attn.out_proj.bias FLOAT[1024] 2d4f41c12fc6
transformer.resblocks.3.ln_1.bias FLOAT[1024] 9eb07ed84b19
transformer.resblocks.3.ln_1.weight FLOAT[1024] 52b12d7bf441
transformer.resblocks.3.ln_2.bias FLOAT[1024] 8d5a89b09d7d
transformer.resblocks.3.ln_2.weight FLOAT[1024] b3769d540e31
transformer.resblocks.3.mlp.c_fc.bias FLOAT[4096] 3cb4212accd2
transformer.resblocks.3.mlp.c_proj.bias FLOAT[1024] 6ee1d299af82
transformer.resblocks.4.attn.in_proj_bias FLOAT[3072] d0d9607c30b1
transformer.resblocks.4.attn.out_proj.bias FLOAT[1024] b021a9cbcb03
transformer.resblocks.4.ln_1.bias FLOAT[1024] 25abea59c671
transformer.resblocks.4.ln_1.weight FLOAT[1024] 3f824786f768
transformer.resblocks.4.ln_2.bias FLOAT[1024] 26d65ed76c79
transformer.resblocks.4.ln_2.weight FLOAT[1024] c5dfc1613b57
transformer.resblocks.4.mlp.c_fc.bias FLOAT[4096] 4312f34c40ad
transformer.resblocks.4.mlp.c_proj.bias FLOAT[1024] 3d78336440fd
transformer.resblocks.5.attn.in_proj_bias FLOAT[3072] dbc12dfe9985
transformer.resblocks.5.attn.out_proj.bias FLOAT[1024] 66d9ee75701e
transformer.resblocks.5.ln_1.bias FLOAT[1024] 17c37a3a1701
transformer.resblocks.5.ln_1.weight FLOAT[1024] ab61a47529c5
transformer.resblocks.5.ln_2.bias FLOAT[1024] 49bec58f123c
transformer.resblocks.5.ln_2.weight FLOAT[1024] 2d1bc345ff83
transformer.resblocks.5.mlp.c_fc.bias FLOAT[4096] 4c24051fc63a
transformer.resblocks.5.mlp.c_proj.bias FLOAT[1024] f969fbcf9940
transformer.resblocks.6.attn.in_proj_bias FLOAT[3072] 40daeb7dd402
transformer.resblocks.6.attn.out_proj.bias FLOAT[1024] 55d4fa7d2853
transformer.resblocks.6.ln_1.bias FLOAT[1024] 2f653ab683f7
transformer.resblocks.6.ln_1.weight FLOAT[1024] 42c41908ec66
transformer.resblocks.6.ln_2.bias FLOAT[1024] 5f6db17d627c
transformer.resblocks.6.ln_2.weight FLOAT[1024] 0233aba03f81
transformer.resblocks.6.mlp.c_fc.bias FLOAT[4096] 97088bd40bed
transformer.resblocks.6.mlp.c_proj.bias FLOAT[1024] 8f13686b4d0e
transformer.resblocks.7.attn.in_proj_bias FLOAT[3072] ef1f06739b57
transformer.resblocks.7.attn.out_proj.bias FLOAT[1024] 2556501c3f60
transformer.resblocks.7.ln_1.bias FLOAT[1024] 3446e3243707
transformer.resblocks.7.ln_1.weight FLOAT[1024] 245958849fff
transformer.resblocks.7.ln_2.bias FLOAT[1024] fc0d9d121d3c
transformer.resblocks.7.ln_2.weight FLOAT[1024] 89e8afb98deb
transformer.resblocks.7.mlp.c_fc.bias FLOAT[4096] c007360827b1
transformer.resblocks.7.mlp.c_proj.bias FLOAT[1024] 3396b833a740
transformer.resblocks.8.attn.in_proj_bias FLOAT[3072] 6d0421c16391
transformer.resblocks.8.attn.out_proj.bias FLOAT[1024] 54a5dd209f16
transformer.resblocks.8.ln_1.bias FLOAT[1024] 4733f9ba4638
transformer.resblocks.8.ln_1.weight FLOAT[1024] 97521cf5ff7d
transformer.resblocks.8.ln_2.bias FLOAT[1024] e903e97c3e9d
transformer.resblocks.8.ln_2.weight FLOAT[1024] 086fe8ecfa80
transformer.resblocks.8.mlp.c_fc.bias FLOAT[4096] db85792548f5
transformer.resblocks.8.mlp.c_proj.bias FLOAT[1024] 4ce56d3c9dfd
transformer.resblocks.9.attn.in_proj_bias FLOAT[3072] fc41b26d2f1e
transformer.resblocks.9.attn.out_proj.bias FLOAT[1024] 4c79079835db
transformer.resblocks.9.ln_1.bias FLOAT[1024] 20060e7f316c
transformer.resblocks.9.ln_1.weight FLOAT[1024] 4ec38540a91f
transformer.resblocks.9.ln_2.bias FLOAT[1024] 68d1ce5224c4
transformer.resblocks.9.ln_2.weight FLOAT[1024] 1252f4342c51
transformer.resblocks.9.mlp.c_fc.bias FLOAT[4096] 0a0e3ab928fd
transformer.resblocks.9.mlp.c_proj.bias FLOAT[1024] de2d6d8858d8
val_0 INT64[1] 12a3ae445661
val_1 FLOAT[] 6708d9be4956
val_10 FLOAT[1024,4096] 7746d5309a02
val_11 FLOAT[4096,1024] 5a3ea7b4f035
val_12 FLOAT[1024,3072] a8f817662908
val_13 FLOAT[1024,4096] 561d489df76e
val_14 FLOAT[4096,1024] a67c7dc61f14
val_1488_axes1 INT64[1] 7c9fa136d441
val_1488_eye FLOAT[77,77] 39ccea14e08c
val_15 FLOAT[1024,3072] 749f522535de
val_16 FLOAT[1024,4096] fa74e9bea1c3
val_17 FLOAT[4096,1024] b4e18155fbbc
val_18 FLOAT[1024,3072] 2bb0799b0a00
val_19 FLOAT[1024,4096] 862378a3f836
val_2 FLOAT[] c2e7ddfe3114
val_20 FLOAT[4096,1024] be799ab28e68
val_21 FLOAT[1024,3072] 7b76bf6d6d8e
val_22 FLOAT[1024,4096] 9a0fe5c689e7
val_23 FLOAT[4096,1024] 86a0fe2f4141
val_24 FLOAT[1024,3072] dd8cc5bee12f
val_25 FLOAT[1024,4096] d87837d6632b
val_26 FLOAT[4096,1024] 83a61bbf1120
val_27 FLOAT[1024,3072] d2900eb2942e
val_28 FLOAT[1024,4096] e7a5c3c8fd37
val_29 FLOAT[4096,1024] 2ac99c7aee1c
val_3 FLOAT[1024,3072] d7c58effa68e
val_30 FLOAT[1024,3072] ac6919df2a73
val_31 FLOAT[1024,4096] 6c1969b18979
val_32 FLOAT[4096,1024] 8ef87563e017
val_33 FLOAT[1024,3072] c8f83852c876
val_34 FLOAT[1024,4096] a425fcf2114f
val_35 FLOAT[4096,1024] 39efb31e2452
val_36 FLOAT[1024,3072] 7652223dd66d
val_37 FLOAT[1024,4096] f7b5433ddb10
val_38 FLOAT[4096,1024] 866d65b2730d
val_39 FLOAT[1024,3072] 72a6d22bdf89
val_4 FLOAT[1024,4096] 35155e31ec66
val_40 FLOAT[1024,4096] f5c1620c0ef8
val_41 FLOAT[4096,1024] 77517c7e7031
val_42 FLOAT[1024,3072] 19ce4b5aeee1
val_43 FLOAT[1024,4096] 44d4b204d987
val_44 FLOAT[4096,1024] a96391b3d22d
val_45 FLOAT[1024,3072] ae9c5f3d4d9e
val_46 FLOAT[1024,4096] 3a877b986dfe
val_47 FLOAT[4096,1024] e52c9397843b
val_48 FLOAT[1024,3072] 2a21e09cccf4
val_49 FLOAT[1024,4096] 3be55cd18aab
val_5 FLOAT[4096,1024] 25879bbbb255
val_50 FLOAT[4096,1024] d2871614d5d0
val_51 FLOAT[1024,3072] 95570f1c96fc
val_52 FLOAT[1024,4096] aed6c66484c1
val_53 FLOAT[4096,1024] a375a8d4eadb
val_54 FLOAT[1024,3072] a069f98dcbe5
val_55 FLOAT[1024,4096] eb85299c0e71
val_56 FLOAT[4096,1024] b023b9810b9b
val_57 FLOAT[1024,3072] fed908dc0db5
val_58 FLOAT[1024,4096] c607ff07e507
val_59 FLOAT[4096,1024] 6ed8c0c72d3d
val_6 FLOAT[1024,3072] 7c2bb3fe84d0
val_60 FLOAT[1024,3072] 920c7633a011
val_61 FLOAT[1024,4096] c2325b63b5a0
val_62 FLOAT[4096,1024] abec52c102f9
val_63 FLOAT[1024,3072] faff83345e2b
val_64 FLOAT[1024,4096] 0b64fbe0664d
val_65 FLOAT[4096,1024] 9072f4f2a9f2
val_66 FLOAT[1024,3072] 89f738c188e4
val_67 FLOAT[1024,4096] d0ab3559aaff
val_68 FLOAT[4096,1024] a1f63864d9e6
val_69 FLOAT[1024,3072] 5f0f3892340b
val_7 FLOAT[1024,4096] 715f97a5b945
val_70 FLOAT[1024,4096] c9fe62e84204
val_71 FLOAT[4096,1024] d9c99d91e93a
val_72 FLOAT[1024,3072] 73acc861d204
val_73 FLOAT[1024,4096] fb20f492ff89
val_74 FLOAT[4096,1024] d42105028e5a
val_8 FLOAT[4096,1024] c22f5809d6b2
val_9 FLOAT[1024,3072] a46af35b9725
