<
   ir_version: 10,
   opset_import: ["" : 23],
   producer_name: "pytorch"
>
main_graph (int32[batch,77] input_ids, int32[batch,77] attention_mask) => (float[batch,512] text_embedding) 
   <
      float[batch,77,1024] add_1044
      float[batch,77,1024] add_1069
      float[batch,77,1024] add_1142
      float[batch,77,1024] add_1167
      float[batch,77,1024] add_1240
      float[batch,77,1024] add_1265
      float[batch,77,1024] add_1338
      float[batch,77,1024] add_1363
      float[batch,77,1024] add_1436
      float[batch,77,1024] add_1461
      float[batch,77,1024] add_1534
      float[batch,77,1024] add_1559
      float[batch,77,1024] add_162
      float[batch,77,1024] add_1632
      float[batch,77,1024] add_1657
      float[batch,77,1024] add_1730
      float[batch,77,1024] add_1755
      float[batch,77,1024] add_1828
      float[batch,77,1024] add_1853
      float[batch,77,1024] add_187
      float[batch,77,1024] add_1926
      float[batch,77,1024] add_1951
      float[batch,77,1024] add_2024
      float[batch,77,1024] add_2049
      float[batch,77,1024] add_2122
      float[batch,77,1024] add_2147
      float[batch,77,1024] add_2220
      float[batch,77,1024] add_2245
      float[batch,77,1024] add_2318
      float[batch,77,1024] add_2343
      float[batch,77,1024] add_2416
      float[batch,77,1024] add_2441
      float[batch,77,1024] add_260
      float[batch,77,1024] add_285
      float[batch,77,1024] add_358
      float[batch,77,1024] add_383
      float[batch,77,1024] add_40
      float[batch,77,1024] add_456
      float[batch,77,1024] add_481
      float[batch,77,1024] add_49
      float[batch,77,1024] add_554
      float[batch,77,1024] add_579
      float[batch,77,1024] add_652
      float[batch,77,1024] add_677
      float[batch,77,1024] add_750
      float[batch,77,1024] add_775
      float[batch,77,1024] add_848
      float[batch,77,1024] add_873
      float[batch,77,1024] add_946
      float[batch,77,1024] add_971
      float[batch,1,77,77] bitwise_and_1
      float[batch,1] clamp_min
      float[batch,77,1] convert_element_type_default_1
      float[batch,1] convert_element_type_default_3
      float[batch,1024] div
      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,77,4096] 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,77,1024] layer_norm_47
      float[batch,77,1024] layer_norm_48
      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,77,1024] linear
      float[batch,77,1024] linear_1
      float[batch,77,4096] linear_10
      float[batch,77,4096] linear_100
      float[batch,77,1024] linear_101
      float[batch,77,1024] linear_102
      float[batch,77,1024] linear_103
      float[batch,77,1024] linear_105
      float[batch,77,4096] linear_106
      float[batch,77,1024] linear_107
      float[batch,77,1024] linear_108
      float[batch,77,1024] linear_109
      float[batch,77,1024] linear_11
      float[batch,77,1024] linear_111
      float[batch,77,4096] linear_112
      float[batch,77,1024] linear_113
      float[batch,77,1024] linear_114
      float[batch,77,1024] linear_115
      float[batch,77,1024] linear_117
      float[batch,77,4096] linear_118
      float[batch,77,1024] linear_119
      float[batch,77,1024] linear_12
      float[batch,77,1024] linear_120
      float[batch,77,1024] linear_121
      float[batch,77,1024] linear_123
      float[batch,77,4096] linear_124
      float[batch,77,1024] linear_125
      float[batch,77,1024] linear_126
      float[batch,77,1024] linear_127
      float[batch,77,1024] linear_129
      float[batch,77,1024] linear_13
      float[batch,77,4096] linear_130
      float[batch,77,1024] linear_131
      float[batch,77,1024] linear_132
      float[batch,77,1024] linear_133
      float[batch,77,1024] linear_135
      float[batch,77,4096] linear_136
      float[batch,77,1024] linear_137
      float[batch,77,1024] linear_138
      float[batch,77,1024] linear_139
      float[batch,77,1024] linear_141
      float[batch,77,4096] linear_142
      float[batch,77,1024] linear_143
      float[batch,512] linear_145
      float[batch,77,1024] linear_15
      float[batch,77,4096] linear_16
      float[batch,77,1024] linear_17
      float[batch,77,1024] linear_18
      float[batch,77,1024] linear_19
      float[batch,77,1024] linear_21
      float[batch,77,4096] linear_22
      float[batch,77,1024] linear_23
      float[batch,77,1024] linear_24
      float[batch,77,1024] linear_25
      float[batch,77,1024] linear_27
      float[batch,77,4096] linear_28
      float[batch,77,1024] linear_29
      float[batch,77,1024] linear_3
      float[batch,77,1024] linear_30
      float[batch,77,1024] linear_31
      float[batch,77,1024] linear_33
      float[batch,77,4096] linear_34
      float[batch,77,1024] linear_35
      float[batch,77,1024] linear_36
      float[batch,77,1024] linear_37
      float[batch,77,1024] linear_39
      float[batch,77,4096] linear_4
      float[batch,77,4096] linear_40
      float[batch,77,1024] linear_41
      float[batch,77,1024] linear_42
      float[batch,77,1024] linear_43
      float[batch,77,1024] linear_45
      float[batch,77,4096] linear_46
      float[batch,77,1024] linear_47
      float[batch,77,1024] linear_48
      float[batch,77,1024] linear_49
      float[batch,77,1024] linear_5
      float[batch,77,1024] linear_51
      float[batch,77,4096] linear_52
      float[batch,77,1024] linear_53
      float[batch,77,1024] linear_54
      float[batch,77,1024] linear_55
      float[batch,77,1024] linear_57
      float[batch,77,4096] linear_58
      float[batch,77,1024] linear_59
      float[batch,77,1024] linear_6
      float[batch,77,1024] linear_60
      float[batch,77,1024] linear_61
      float[batch,77,1024] linear_63
      float[batch,77,4096] linear_64
      float[batch,77,1024] linear_65
      float[batch,77,1024] linear_66
      float[batch,77,1024] linear_67
      float[batch,77,1024] linear_69
      float[batch,77,1024] linear_7
      float[batch,77,4096] linear_70
      float[batch,77,1024] linear_71
      float[batch,77,1024] linear_72
      float[batch,77,1024] linear_73
      float[batch,77,1024] linear_75
      float[batch,77,4096] linear_76
      float[batch,77,1024] linear_77
      float[batch,77,1024] linear_78
      float[batch,77,1024] linear_79
      float[batch,77,1024] linear_81
      float[batch,77,4096] linear_82
      float[batch,77,1024] linear_83
      float[batch,77,1024] linear_84
      float[batch,77,1024] linear_85
      float[batch,77,1024] linear_87
      float[batch,77,4096] linear_88
      float[batch,77,1024] linear_89
      float[batch,77,1024] linear_9
      float[batch,77,1024] linear_90
      float[batch,77,1024] linear_91
      float[batch,77,1024] linear_93
      float[batch,77,4096] linear_94
      float[batch,77,1024] linear_95
      float[batch,77,1024] linear_96
      float[batch,77,1024] linear_97
      float[batch,77,1024] linear_99
      float[batch,77,1024] mul_1230
      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,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,1024] sum_1
      int32[batch,77,1] unsqueeze_12
      float[batch,77] val_153
      float[batch,77] val_154
      float[batch,77] val_155
      float[batch,1,1,77] val_156
      float[batch,77,1024] val_157
      float[batch,77,1024] val_158
      float[batch,77,1024] val_159
      float[batch,77,1024] val_160
      float[batch,77,4096] val_161
      float[batch,77,1024] val_162
      float[batch,77,1024] val_163
      float[batch,77,1024] val_164
      float[batch,77,1024] val_165
      float[batch,77,1024] val_166
      float[batch,77,4096] val_167
      float[batch,77,1024] val_168
      float[batch,77,1024] val_169
      float[batch,77,1024] val_170
      float[batch,77,1024] val_171
      float[batch,77,1024] val_172
      float[batch,77,4096] val_173
      float[batch,77,1024] val_174
      float[batch,77,1024] val_175
      float[batch,77,1024] val_176
      float[batch,77,1024] val_177
      float[batch,77,1024] val_178
      float[batch,77,4096] val_179
      float[batch,77,1024] val_180
      float[batch,77,1024] val_181
      float[batch,77,1024] val_182
      float[batch,77,1024] val_183
      float[batch,77,1024] val_184
      float[batch,77,4096] val_185
      float[batch,77,1024] val_186
      float[batch,77,1024] val_187
      float[batch,77,1024] val_188
      float[batch,77,1024] val_189
      float[batch,77,1024] val_190
      float[batch,77,4096] val_191
      float[batch,77,1024] val_192
      float[batch,77,1024] val_193
      float[batch,77,1024] val_194
      float[batch,77,1024] val_195
      float[batch,77,1024] val_196
      float[batch,77,4096] val_197
      float[batch,77,1024] val_198
      float[batch,77,1024] val_199
      float[batch,77,1024] val_200
      float[batch,77,1024] val_201
      float[batch,77,1024] val_202
      float[batch,77,4096] val_203
      float[batch,77,1024] val_204
      float[batch,77,1024] val_205
      float[batch,77,1024] val_206
      float[batch,77,1024] val_207
      float[batch,77,1024] val_208
      float[batch,77,4096] val_209
      float[batch,77,1024] val_210
      float[batch,77,1024] val_211
      float[batch,77,1024] val_212
      float[batch,77,1024] val_213
      float[batch,77,1024] val_214
      float[batch,77,4096] val_215
      float[batch,77,1024] val_216
      float[batch,77,1024] val_217
      float[batch,77,1024] val_218
      float[batch,77,1024] val_219
      float[batch,77,1024] val_220
      float[batch,77,4096] val_221
      float[batch,77,1024] val_222
      float[batch,77,1024] val_223
      float[batch,77,1024] val_224
      float[batch,77,1024] val_225
      float[batch,77,1024] val_226
      float[batch,77,4096] val_227
      float[batch,77,1024] val_228
      float[batch,77,1024] val_229
      float[batch,77,1024] val_230
      float[batch,77,1024] val_231
      float[batch,77,1024] val_232
      float[batch,77,4096] val_233
      float[batch,77,1024] val_234
      float[batch,77,1024] val_235
      float[batch,77,1024] val_236
      float[batch,77,1024] val_237
      float[batch,77,1024] val_238
      float[batch,77,4096] val_239
      float[batch,77,1024] val_240
      float[batch,77,1024] val_241
      float[batch,77,1024] val_242
      float[batch,77,1024] val_243
      float[batch,77,1024] val_244
      float[batch,77,4096] val_245
      float[batch,77,1024] val_246
      float[batch,77,1024] val_247
      float[batch,77,1024] val_248
      float[batch,77,1024] val_249
      float[batch,77,1024] val_250
      float[batch,77,4096] val_251
      float[batch,77,1024] val_252
      float[batch,77,1024] val_253
      float[batch,77,1024] val_254
      float[batch,77,1024] val_255
      float[batch,77,1024] val_256
      float[batch,77,4096] val_257
      float[batch,77,1024] val_258
      float[batch,77,1024] val_259
      float[batch,77,1024] val_260
      float[batch,77,1024] val_261
      float[batch,77,1024] val_262
      float[batch,77,4096] val_263
      float[batch,77,1024] val_264
      float[batch,77,1024] val_265
      float[batch,77,1024] val_266
      float[batch,77,1024] val_267
      float[batch,77,1024] val_268
      float[batch,77,4096] val_269
      float[batch,77,1024] val_270
      float[batch,77,1024] val_271
      float[batch,77,1024] val_272
      float[batch,77,1024] val_273
      float[batch,77,1024] val_274
      float[batch,77,4096] val_275
      float[batch,77,1024] val_276
      float[batch,77,1024] val_277
      float[batch,77,1024] val_278
      float[batch,77,1024] val_279
      float[batch,77,1024] val_280
      float[batch,77,4096] val_281
      float[batch,77,1024] val_282
      float[batch,77,1024] val_283
      float[batch,77,1024] val_284
      float[batch,77,1024] val_285
      float[batch,77,1024] val_286
      float[batch,77,4096] val_287
      float[batch,77,1024] val_288
      float[batch,77,1024] val_289
      float[batch,77,1024] val_290
      float[batch,77,1024] val_291
      float[batch,77,1024] val_292
      float[batch,77,4096] val_293
      float[batch,77,1024] val_294
      float[batch,77,1024] val_295
      float[batch,77,1024] val_296
      float[batch,77,1024] val_297
      float[batch,77,1024] val_298
      float[batch,77,4096] val_299
      float[batch,77,1024] val_300
      float[batch,77] val_301
      float[batch] val_302
   >
{
   val_152 = Gather <axis: int = 0> ("transformer.embeddings.word_embeddings.weight_fp16", input_ids)
   embedding = Cast <to: int = 1> (val_152)
   add_40 = Add (embedding, embedding_1)
   add_49 = Add (add_40, embedding_2)
   [node_layer_norm] layer_norm = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_49, "transformer.embeddings.LayerNorm.weight", "transformer.embeddings.LayerNorm.bias")
   val_153 = Cast <to: int = 1> (attention_mask)
   val_154 = Sub (val_153, val_4)
   val_155 = Mul (val_154, val_5)
   val_156 = Unsqueeze (val_155, val_6)
   bitwise_and_1 = Add (val_156, val_7)
   val_157 = MatMul (layer_norm, val_8)
   [node_linear] linear = Add (val_157, "transformer.encoder.layer.0.attention.self.query.bias")
   val_158 = MatMul (layer_norm, val_9)
   linear_1 = Add (val_158, "transformer.encoder.layer.0.attention.self.key.bias")
   val_159 = MatMul (layer_norm, val_10)
   [node_scaled_dot_product_attention] scaled_dot_product_attention = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, scale: float = 0.125, softcap: float = 0> (linear, linear_1, val_159, bitwise_and_1)
   val_160 = MatMul (scaled_dot_product_attention, val_11)
   linear_3 = Add (val_160, "transformer.encoder.layer.0.attention.output.dense.bias")
   add_162 = Add (linear_3, layer_norm)
   layer_norm_1 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_162, "transformer.encoder.layer.0.attention.output.LayerNorm.weight", "transformer.encoder.layer.0.attention.output.LayerNorm.bias")
   val_161 = MatMul (layer_norm_1, val_12)
   linear_4 = Add (val_161, "transformer.encoder.layer.0.intermediate.dense.bias")
   [node_gelu] gelu = Gelu <approximate: string = "none"> (linear_4)
   val_162 = MatMul (gelu, val_13)
   linear_5 = Add (val_162, "transformer.encoder.layer.0.output.dense.bias")
   add_187 = Add (linear_5, layer_norm_1)
   layer_norm_2 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_187, "transformer.encoder.layer.0.output.LayerNorm.weight", "transformer.encoder.layer.0.output.LayerNorm.bias")
   val_163 = MatMul (layer_norm_2, val_14)
   linear_6 = Add (val_163, "transformer.encoder.layer.1.attention.self.query.bias")
   val_164 = MatMul (layer_norm_2, val_15)
   linear_7 = Add (val_164, "transformer.encoder.layer.1.attention.self.key.bias")
   val_165 = MatMul (layer_norm_2, val_16)
   scaled_dot_product_attention_1 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, scale: float = 0.125, softcap: float = 0> (linear_6, linear_7, val_165, bitwise_and_1)
   val_166 = MatMul (scaled_dot_product_attention_1, val_17)
   linear_9 = Add (val_166, "transformer.encoder.layer.1.attention.output.dense.bias")
   add_260 = Add (linear_9, layer_norm_2)
   layer_norm_3 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_260, "transformer.encoder.layer.1.attention.output.LayerNorm.weight", "transformer.encoder.layer.1.attention.output.LayerNorm.bias")
   val_167 = MatMul (layer_norm_3, val_18)
   linear_10 = Add (val_167, "transformer.encoder.layer.1.intermediate.dense.bias")
   gelu_1 = Gelu <approximate: string = "none"> (linear_10)
   val_168 = MatMul (gelu_1, val_19)
   linear_11 = Add (val_168, "transformer.encoder.layer.1.output.dense.bias")
   add_285 = Add (linear_11, layer_norm_3)
   layer_norm_4 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_285, "transformer.encoder.layer.1.output.LayerNorm.weight", "transformer.encoder.layer.1.output.LayerNorm.bias")
   val_169 = MatMul (layer_norm_4, val_20)
   linear_12 = Add (val_169, "transformer.encoder.layer.2.attention.self.query.bias")
   val_170 = MatMul (layer_norm_4, val_21)
   linear_13 = Add (val_170, "transformer.encoder.layer.2.attention.self.key.bias")
   val_171 = MatMul (layer_norm_4, val_22)
   scaled_dot_product_attention_2 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, scale: float = 0.125, softcap: float = 0> (linear_12, linear_13, val_171, bitwise_and_1)
   val_172 = MatMul (scaled_dot_product_attention_2, val_23)
   linear_15 = Add (val_172, "transformer.encoder.layer.2.attention.output.dense.bias")
   add_358 = Add (linear_15, layer_norm_4)
   layer_norm_5 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_358, "transformer.encoder.layer.2.attention.output.LayerNorm.weight", "transformer.encoder.layer.2.attention.output.LayerNorm.bias")
   val_173 = MatMul (layer_norm_5, val_24)
   linear_16 = Add (val_173, "transformer.encoder.layer.2.intermediate.dense.bias")
   gelu_2 = Gelu <approximate: string = "none"> (linear_16)
   val_174 = MatMul (gelu_2, val_25)
   linear_17 = Add (val_174, "transformer.encoder.layer.2.output.dense.bias")
   add_383 = Add (linear_17, layer_norm_5)
   layer_norm_6 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_383, "transformer.encoder.layer.2.output.LayerNorm.weight", "transformer.encoder.layer.2.output.LayerNorm.bias")
   val_175 = MatMul (layer_norm_6, val_26)
   linear_18 = Add (val_175, "transformer.encoder.layer.3.attention.self.query.bias")
   val_176 = MatMul (layer_norm_6, val_27)
   linear_19 = Add (val_176, "transformer.encoder.layer.3.attention.self.key.bias")
   val_177 = MatMul (layer_norm_6, val_28)
   scaled_dot_product_attention_3 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, scale: float = 0.125, softcap: float = 0> (linear_18, linear_19, val_177, bitwise_and_1)
   val_178 = MatMul (scaled_dot_product_attention_3, val_29)
   linear_21 = Add (val_178, "transformer.encoder.layer.3.attention.output.dense.bias")
   add_456 = Add (linear_21, layer_norm_6)
   layer_norm_7 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_456, "transformer.encoder.layer.3.attention.output.LayerNorm.weight", "transformer.encoder.layer.3.attention.output.LayerNorm.bias")
   val_179 = MatMul (layer_norm_7, val_30)
   linear_22 = Add (val_179, "transformer.encoder.layer.3.intermediate.dense.bias")
   gelu_3 = Gelu <approximate: string = "none"> (linear_22)
   val_180 = MatMul (gelu_3, val_31)
   linear_23 = Add (val_180, "transformer.encoder.layer.3.output.dense.bias")
   add_481 = Add (linear_23, layer_norm_7)
   layer_norm_8 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_481, "transformer.encoder.layer.3.output.LayerNorm.weight", "transformer.encoder.layer.3.output.LayerNorm.bias")
   val_181 = MatMul (layer_norm_8, val_32)
   linear_24 = Add (val_181, "transformer.encoder.layer.4.attention.self.query.bias")
   val_182 = MatMul (layer_norm_8, val_33)
   linear_25 = Add (val_182, "transformer.encoder.layer.4.attention.self.key.bias")
   val_183 = MatMul (layer_norm_8, val_34)
   scaled_dot_product_attention_4 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, scale: float = 0.125, softcap: float = 0> (linear_24, linear_25, val_183, bitwise_and_1)
   val_184 = MatMul (scaled_dot_product_attention_4, val_35)
   linear_27 = Add (val_184, "transformer.encoder.layer.4.attention.output.dense.bias")
   add_554 = Add (linear_27, layer_norm_8)
   layer_norm_9 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_554, "transformer.encoder.layer.4.attention.output.LayerNorm.weight", "transformer.encoder.layer.4.attention.output.LayerNorm.bias")
   val_185 = MatMul (layer_norm_9, val_36)
   linear_28 = Add (val_185, "transformer.encoder.layer.4.intermediate.dense.bias")
   gelu_4 = Gelu <approximate: string = "none"> (linear_28)
   val_186 = MatMul (gelu_4, val_37)
   linear_29 = Add (val_186, "transformer.encoder.layer.4.output.dense.bias")
   add_579 = Add (linear_29, layer_norm_9)
   layer_norm_10 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_579, "transformer.encoder.layer.4.output.LayerNorm.weight", "transformer.encoder.layer.4.output.LayerNorm.bias")
   val_187 = MatMul (layer_norm_10, val_38)
   linear_30 = Add (val_187, "transformer.encoder.layer.5.attention.self.query.bias")
   val_188 = MatMul (layer_norm_10, val_39)
   linear_31 = Add (val_188, "transformer.encoder.layer.5.attention.self.key.bias")
   val_189 = MatMul (layer_norm_10, val_40)
   scaled_dot_product_attention_5 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, scale: float = 0.125, softcap: float = 0> (linear_30, linear_31, val_189, bitwise_and_1)
   val_190 = MatMul (scaled_dot_product_attention_5, val_41)
   linear_33 = Add (val_190, "transformer.encoder.layer.5.attention.output.dense.bias")
   add_652 = Add (linear_33, layer_norm_10)
   layer_norm_11 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_652, "transformer.encoder.layer.5.attention.output.LayerNorm.weight", "transformer.encoder.layer.5.attention.output.LayerNorm.bias")
   val_191 = MatMul (layer_norm_11, val_42)
   linear_34 = Add (val_191, "transformer.encoder.layer.5.intermediate.dense.bias")
   gelu_5 = Gelu <approximate: string = "none"> (linear_34)
   val_192 = MatMul (gelu_5, val_43)
   linear_35 = Add (val_192, "transformer.encoder.layer.5.output.dense.bias")
   add_677 = Add (linear_35, layer_norm_11)
   layer_norm_12 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_677, "transformer.encoder.layer.5.output.LayerNorm.weight", "transformer.encoder.layer.5.output.LayerNorm.bias")
   val_193 = MatMul (layer_norm_12, val_44)
   linear_36 = Add (val_193, "transformer.encoder.layer.6.attention.self.query.bias")
   val_194 = MatMul (layer_norm_12, val_45)
   linear_37 = Add (val_194, "transformer.encoder.layer.6.attention.self.key.bias")
   val_195 = MatMul (layer_norm_12, val_46)
   scaled_dot_product_attention_6 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, scale: float = 0.125, softcap: float = 0> (linear_36, linear_37, val_195, bitwise_and_1)
   val_196 = MatMul (scaled_dot_product_attention_6, val_47)
   linear_39 = Add (val_196, "transformer.encoder.layer.6.attention.output.dense.bias")
   add_750 = Add (linear_39, layer_norm_12)
   layer_norm_13 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_750, "transformer.encoder.layer.6.attention.output.LayerNorm.weight", "transformer.encoder.layer.6.attention.output.LayerNorm.bias")
   val_197 = MatMul (layer_norm_13, val_48)
   linear_40 = Add (val_197, "transformer.encoder.layer.6.intermediate.dense.bias")
   gelu_6 = Gelu <approximate: string = "none"> (linear_40)
   val_198 = MatMul (gelu_6, val_49)
   linear_41 = Add (val_198, "transformer.encoder.layer.6.output.dense.bias")
   add_775 = Add (linear_41, layer_norm_13)
   layer_norm_14 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_775, "transformer.encoder.layer.6.output.LayerNorm.weight", "transformer.encoder.layer.6.output.LayerNorm.bias")
   val_199 = MatMul (layer_norm_14, val_50)
   linear_42 = Add (val_199, "transformer.encoder.layer.7.attention.self.query.bias")
   val_200 = MatMul (layer_norm_14, val_51)
   linear_43 = Add (val_200, "transformer.encoder.layer.7.attention.self.key.bias")
   val_201 = MatMul (layer_norm_14, val_52)
   scaled_dot_product_attention_7 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, scale: float = 0.125, softcap: float = 0> (linear_42, linear_43, val_201, bitwise_and_1)
   val_202 = MatMul (scaled_dot_product_attention_7, val_53)
   linear_45 = Add (val_202, "transformer.encoder.layer.7.attention.output.dense.bias")
   add_848 = Add (linear_45, layer_norm_14)
   layer_norm_15 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_848, "transformer.encoder.layer.7.attention.output.LayerNorm.weight", "transformer.encoder.layer.7.attention.output.LayerNorm.bias")
   val_203 = MatMul (layer_norm_15, val_54)
   linear_46 = Add (val_203, "transformer.encoder.layer.7.intermediate.dense.bias")
   gelu_7 = Gelu <approximate: string = "none"> (linear_46)
   val_204 = MatMul (gelu_7, val_55)
   linear_47 = Add (val_204, "transformer.encoder.layer.7.output.dense.bias")
   add_873 = Add (linear_47, layer_norm_15)
   layer_norm_16 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_873, "transformer.encoder.layer.7.output.LayerNorm.weight", "transformer.encoder.layer.7.output.LayerNorm.bias")
   val_205 = MatMul (layer_norm_16, val_56)
   linear_48 = Add (val_205, "transformer.encoder.layer.8.attention.self.query.bias")
   val_206 = MatMul (layer_norm_16, val_57)
   linear_49 = Add (val_206, "transformer.encoder.layer.8.attention.self.key.bias")
   val_207 = MatMul (layer_norm_16, val_58)
   scaled_dot_product_attention_8 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, scale: float = 0.125, softcap: float = 0> (linear_48, linear_49, val_207, bitwise_and_1)
   val_208 = MatMul (scaled_dot_product_attention_8, val_59)
   linear_51 = Add (val_208, "transformer.encoder.layer.8.attention.output.dense.bias")
   add_946 = Add (linear_51, layer_norm_16)
   layer_norm_17 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_946, "transformer.encoder.layer.8.attention.output.LayerNorm.weight", "transformer.encoder.layer.8.attention.output.LayerNorm.bias")
   val_209 = MatMul (layer_norm_17, val_60)
   linear_52 = Add (val_209, "transformer.encoder.layer.8.intermediate.dense.bias")
   gelu_8 = Gelu <approximate: string = "none"> (linear_52)
   val_210 = MatMul (gelu_8, val_61)
   linear_53 = Add (val_210, "transformer.encoder.layer.8.output.dense.bias")
   add_971 = Add (linear_53, layer_norm_17)
   layer_norm_18 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_971, "transformer.encoder.layer.8.output.LayerNorm.weight", "transformer.encoder.layer.8.output.LayerNorm.bias")
   val_211 = MatMul (layer_norm_18, val_62)
   linear_54 = Add (val_211, "transformer.encoder.layer.9.attention.self.query.bias")
   val_212 = MatMul (layer_norm_18, val_63)
   linear_55 = Add (val_212, "transformer.encoder.layer.9.attention.self.key.bias")
   val_213 = MatMul (layer_norm_18, val_64)
   scaled_dot_product_attention_9 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, scale: float = 0.125, softcap: float = 0> (linear_54, linear_55, val_213, bitwise_and_1)
   val_214 = MatMul (scaled_dot_product_attention_9, val_65)
   linear_57 = Add (val_214, "transformer.encoder.layer.9.attention.output.dense.bias")
   add_1044 = Add (linear_57, layer_norm_18)
   layer_norm_19 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1044, "transformer.encoder.layer.9.attention.output.LayerNorm.weight", "transformer.encoder.layer.9.attention.output.LayerNorm.bias")
   val_215 = MatMul (layer_norm_19, val_66)
   linear_58 = Add (val_215, "transformer.encoder.layer.9.intermediate.dense.bias")
   gelu_9 = Gelu <approximate: string = "none"> (linear_58)
   val_216 = MatMul (gelu_9, val_67)
   linear_59 = Add (val_216, "transformer.encoder.layer.9.output.dense.bias")
   add_1069 = Add (linear_59, layer_norm_19)
   layer_norm_20 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1069, "transformer.encoder.layer.9.output.LayerNorm.weight", "transformer.encoder.layer.9.output.LayerNorm.bias")
   val_217 = MatMul (layer_norm_20, val_68)
   linear_60 = Add (val_217, "transformer.encoder.layer.10.attention.self.query.bias")
   val_218 = MatMul (layer_norm_20, val_69)
   linear_61 = Add (val_218, "transformer.encoder.layer.10.attention.self.key.bias")
   val_219 = MatMul (layer_norm_20, val_70)
   scaled_dot_product_attention_10 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, scale: float = 0.125, softcap: float = 0> (linear_60, linear_61, val_219, bitwise_and_1)
   val_220 = MatMul (scaled_dot_product_attention_10, val_71)
   linear_63 = Add (val_220, "transformer.encoder.layer.10.attention.output.dense.bias")
   add_1142 = Add (linear_63, layer_norm_20)
   layer_norm_21 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1142, "transformer.encoder.layer.10.attention.output.LayerNorm.weight", "transformer.encoder.layer.10.attention.output.LayerNorm.bias")
   val_221 = MatMul (layer_norm_21, val_72)
   linear_64 = Add (val_221, "transformer.encoder.layer.10.intermediate.dense.bias")
   gelu_10 = Gelu <approximate: string = "none"> (linear_64)
   val_222 = MatMul (gelu_10, val_73)
   linear_65 = Add (val_222, "transformer.encoder.layer.10.output.dense.bias")
   add_1167 = Add (linear_65, layer_norm_21)
   layer_norm_22 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1167, "transformer.encoder.layer.10.output.LayerNorm.weight", "transformer.encoder.layer.10.output.LayerNorm.bias")
   val_223 = MatMul (layer_norm_22, val_74)
   linear_66 = Add (val_223, "transformer.encoder.layer.11.attention.self.query.bias")
   val_224 = MatMul (layer_norm_22, val_75)
   linear_67 = Add (val_224, "transformer.encoder.layer.11.attention.self.key.bias")
   val_225 = MatMul (layer_norm_22, val_76)
   scaled_dot_product_attention_11 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, scale: float = 0.125, softcap: float = 0> (linear_66, linear_67, val_225, bitwise_and_1)
   val_226 = MatMul (scaled_dot_product_attention_11, val_77)
   linear_69 = Add (val_226, "transformer.encoder.layer.11.attention.output.dense.bias")
   add_1240 = Add (linear_69, layer_norm_22)
   layer_norm_23 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1240, "transformer.encoder.layer.11.attention.output.LayerNorm.weight", "transformer.encoder.layer.11.attention.output.LayerNorm.bias")
   val_227 = MatMul (layer_norm_23, val_78)
   linear_70 = Add (val_227, "transformer.encoder.layer.11.intermediate.dense.bias")
   gelu_11 = Gelu <approximate: string = "none"> (linear_70)
   val_228 = MatMul (gelu_11, val_79)
   linear_71 = Add (val_228, "transformer.encoder.layer.11.output.dense.bias")
   add_1265 = Add (linear_71, layer_norm_23)
   layer_norm_24 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1265, "transformer.encoder.layer.11.output.LayerNorm.weight", "transformer.encoder.layer.11.output.LayerNorm.bias")
   val_229 = MatMul (layer_norm_24, val_80)
   linear_72 = Add (val_229, "transformer.encoder.layer.12.attention.self.query.bias")
   val_230 = MatMul (layer_norm_24, val_81)
   linear_73 = Add (val_230, "transformer.encoder.layer.12.attention.self.key.bias")
   val_231 = MatMul (layer_norm_24, val_82)
   scaled_dot_product_attention_12 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, scale: float = 0.125, softcap: float = 0> (linear_72, linear_73, val_231, bitwise_and_1)
   val_232 = MatMul (scaled_dot_product_attention_12, val_83)
   linear_75 = Add (val_232, "transformer.encoder.layer.12.attention.output.dense.bias")
   add_1338 = Add (linear_75, layer_norm_24)
   layer_norm_25 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1338, "transformer.encoder.layer.12.attention.output.LayerNorm.weight", "transformer.encoder.layer.12.attention.output.LayerNorm.bias")
   val_233 = MatMul (layer_norm_25, val_84)
   linear_76 = Add (val_233, "transformer.encoder.layer.12.intermediate.dense.bias")
   gelu_12 = Gelu <approximate: string = "none"> (linear_76)
   val_234 = MatMul (gelu_12, val_85)
   linear_77 = Add (val_234, "transformer.encoder.layer.12.output.dense.bias")
   add_1363 = Add (linear_77, layer_norm_25)
   layer_norm_26 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1363, "transformer.encoder.layer.12.output.LayerNorm.weight", "transformer.encoder.layer.12.output.LayerNorm.bias")
   val_235 = MatMul (layer_norm_26, val_86)
   linear_78 = Add (val_235, "transformer.encoder.layer.13.attention.self.query.bias")
   val_236 = MatMul (layer_norm_26, val_87)
   linear_79 = Add (val_236, "transformer.encoder.layer.13.attention.self.key.bias")
   val_237 = MatMul (layer_norm_26, val_88)
   scaled_dot_product_attention_13 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, scale: float = 0.125, softcap: float = 0> (linear_78, linear_79, val_237, bitwise_and_1)
   val_238 = MatMul (scaled_dot_product_attention_13, val_89)
   linear_81 = Add (val_238, "transformer.encoder.layer.13.attention.output.dense.bias")
   add_1436 = Add (linear_81, layer_norm_26)
   layer_norm_27 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1436, "transformer.encoder.layer.13.attention.output.LayerNorm.weight", "transformer.encoder.layer.13.attention.output.LayerNorm.bias")
   val_239 = MatMul (layer_norm_27, val_90)
   linear_82 = Add (val_239, "transformer.encoder.layer.13.intermediate.dense.bias")
   gelu_13 = Gelu <approximate: string = "none"> (linear_82)
   val_240 = MatMul (gelu_13, val_91)
   linear_83 = Add (val_240, "transformer.encoder.layer.13.output.dense.bias")
   add_1461 = Add (linear_83, layer_norm_27)
   layer_norm_28 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1461, "transformer.encoder.layer.13.output.LayerNorm.weight", "transformer.encoder.layer.13.output.LayerNorm.bias")
   val_241 = MatMul (layer_norm_28, val_92)
   linear_84 = Add (val_241, "transformer.encoder.layer.14.attention.self.query.bias")
   val_242 = MatMul (layer_norm_28, val_93)
   linear_85 = Add (val_242, "transformer.encoder.layer.14.attention.self.key.bias")
   val_243 = MatMul (layer_norm_28, val_94)
   scaled_dot_product_attention_14 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, scale: float = 0.125, softcap: float = 0> (linear_84, linear_85, val_243, bitwise_and_1)
   val_244 = MatMul (scaled_dot_product_attention_14, val_95)
   linear_87 = Add (val_244, "transformer.encoder.layer.14.attention.output.dense.bias")
   add_1534 = Add (linear_87, layer_norm_28)
   layer_norm_29 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1534, "transformer.encoder.layer.14.attention.output.LayerNorm.weight", "transformer.encoder.layer.14.attention.output.LayerNorm.bias")
   val_245 = MatMul (layer_norm_29, val_96)
   linear_88 = Add (val_245, "transformer.encoder.layer.14.intermediate.dense.bias")
   gelu_14 = Gelu <approximate: string = "none"> (linear_88)
   val_246 = MatMul (gelu_14, val_97)
   linear_89 = Add (val_246, "transformer.encoder.layer.14.output.dense.bias")
   add_1559 = Add (linear_89, layer_norm_29)
   layer_norm_30 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1559, "transformer.encoder.layer.14.output.LayerNorm.weight", "transformer.encoder.layer.14.output.LayerNorm.bias")
   val_247 = MatMul (layer_norm_30, val_98)
   linear_90 = Add (val_247, "transformer.encoder.layer.15.attention.self.query.bias")
   val_248 = MatMul (layer_norm_30, val_99)
   linear_91 = Add (val_248, "transformer.encoder.layer.15.attention.self.key.bias")
   val_249 = MatMul (layer_norm_30, val_100)
   scaled_dot_product_attention_15 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, scale: float = 0.125, softcap: float = 0> (linear_90, linear_91, val_249, bitwise_and_1)
   val_250 = MatMul (scaled_dot_product_attention_15, val_101)
   linear_93 = Add (val_250, "transformer.encoder.layer.15.attention.output.dense.bias")
   add_1632 = Add (linear_93, layer_norm_30)
   layer_norm_31 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1632, "transformer.encoder.layer.15.attention.output.LayerNorm.weight", "transformer.encoder.layer.15.attention.output.LayerNorm.bias")
   val_251 = MatMul (layer_norm_31, val_102)
   linear_94 = Add (val_251, "transformer.encoder.layer.15.intermediate.dense.bias")
   gelu_15 = Gelu <approximate: string = "none"> (linear_94)
   val_252 = MatMul (gelu_15, val_103)
   linear_95 = Add (val_252, "transformer.encoder.layer.15.output.dense.bias")
   add_1657 = Add (linear_95, layer_norm_31)
   layer_norm_32 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1657, "transformer.encoder.layer.15.output.LayerNorm.weight", "transformer.encoder.layer.15.output.LayerNorm.bias")
   val_253 = MatMul (layer_norm_32, val_104)
   linear_96 = Add (val_253, "transformer.encoder.layer.16.attention.self.query.bias")
   val_254 = MatMul (layer_norm_32, val_105)
   linear_97 = Add (val_254, "transformer.encoder.layer.16.attention.self.key.bias")
   val_255 = MatMul (layer_norm_32, val_106)
   scaled_dot_product_attention_16 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, scale: float = 0.125, softcap: float = 0> (linear_96, linear_97, val_255, bitwise_and_1)
   val_256 = MatMul (scaled_dot_product_attention_16, val_107)
   linear_99 = Add (val_256, "transformer.encoder.layer.16.attention.output.dense.bias")
   add_1730 = Add (linear_99, layer_norm_32)
   layer_norm_33 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1730, "transformer.encoder.layer.16.attention.output.LayerNorm.weight", "transformer.encoder.layer.16.attention.output.LayerNorm.bias")
   val_257 = MatMul (layer_norm_33, val_108)
   linear_100 = Add (val_257, "transformer.encoder.layer.16.intermediate.dense.bias")
   gelu_16 = Gelu <approximate: string = "none"> (linear_100)
   val_258 = MatMul (gelu_16, val_109)
   linear_101 = Add (val_258, "transformer.encoder.layer.16.output.dense.bias")
   add_1755 = Add (linear_101, layer_norm_33)
   layer_norm_34 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1755, "transformer.encoder.layer.16.output.LayerNorm.weight", "transformer.encoder.layer.16.output.LayerNorm.bias")
   val_259 = MatMul (layer_norm_34, val_110)
   linear_102 = Add (val_259, "transformer.encoder.layer.17.attention.self.query.bias")
   val_260 = MatMul (layer_norm_34, val_111)
   linear_103 = Add (val_260, "transformer.encoder.layer.17.attention.self.key.bias")
   val_261 = MatMul (layer_norm_34, val_112)
   scaled_dot_product_attention_17 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, scale: float = 0.125, softcap: float = 0> (linear_102, linear_103, val_261, bitwise_and_1)
   val_262 = MatMul (scaled_dot_product_attention_17, val_113)
   linear_105 = Add (val_262, "transformer.encoder.layer.17.attention.output.dense.bias")
   add_1828 = Add (linear_105, layer_norm_34)
   layer_norm_35 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1828, "transformer.encoder.layer.17.attention.output.LayerNorm.weight", "transformer.encoder.layer.17.attention.output.LayerNorm.bias")
   val_263 = MatMul (layer_norm_35, val_114)
   linear_106 = Add (val_263, "transformer.encoder.layer.17.intermediate.dense.bias")
   gelu_17 = Gelu <approximate: string = "none"> (linear_106)
   val_264 = MatMul (gelu_17, val_115)
   linear_107 = Add (val_264, "transformer.encoder.layer.17.output.dense.bias")
   add_1853 = Add (linear_107, layer_norm_35)
   layer_norm_36 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1853, "transformer.encoder.layer.17.output.LayerNorm.weight", "transformer.encoder.layer.17.output.LayerNorm.bias")
   val_265 = MatMul (layer_norm_36, val_116)
   linear_108 = Add (val_265, "transformer.encoder.layer.18.attention.self.query.bias")
   val_266 = MatMul (layer_norm_36, val_117)
   linear_109 = Add (val_266, "transformer.encoder.layer.18.attention.self.key.bias")
   val_267 = MatMul (layer_norm_36, val_118)
   scaled_dot_product_attention_18 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, scale: float = 0.125, softcap: float = 0> (linear_108, linear_109, val_267, bitwise_and_1)
   val_268 = MatMul (scaled_dot_product_attention_18, val_119)
   linear_111 = Add (val_268, "transformer.encoder.layer.18.attention.output.dense.bias")
   add_1926 = Add (linear_111, layer_norm_36)
   layer_norm_37 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1926, "transformer.encoder.layer.18.attention.output.LayerNorm.weight", "transformer.encoder.layer.18.attention.output.LayerNorm.bias")
   val_269 = MatMul (layer_norm_37, val_120)
   linear_112 = Add (val_269, "transformer.encoder.layer.18.intermediate.dense.bias")
   gelu_18 = Gelu <approximate: string = "none"> (linear_112)
   val_270 = MatMul (gelu_18, val_121)
   linear_113 = Add (val_270, "transformer.encoder.layer.18.output.dense.bias")
   add_1951 = Add (linear_113, layer_norm_37)
   layer_norm_38 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1951, "transformer.encoder.layer.18.output.LayerNorm.weight", "transformer.encoder.layer.18.output.LayerNorm.bias")
   val_271 = MatMul (layer_norm_38, val_122)
   linear_114 = Add (val_271, "transformer.encoder.layer.19.attention.self.query.bias")
   val_272 = MatMul (layer_norm_38, val_123)
   linear_115 = Add (val_272, "transformer.encoder.layer.19.attention.self.key.bias")
   val_273 = MatMul (layer_norm_38, val_124)
   scaled_dot_product_attention_19 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, scale: float = 0.125, softcap: float = 0> (linear_114, linear_115, val_273, bitwise_and_1)
   val_274 = MatMul (scaled_dot_product_attention_19, val_125)
   linear_117 = Add (val_274, "transformer.encoder.layer.19.attention.output.dense.bias")
   add_2024 = Add (linear_117, layer_norm_38)
   layer_norm_39 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2024, "transformer.encoder.layer.19.attention.output.LayerNorm.weight", "transformer.encoder.layer.19.attention.output.LayerNorm.bias")
   val_275 = MatMul (layer_norm_39, val_126)
   linear_118 = Add (val_275, "transformer.encoder.layer.19.intermediate.dense.bias")
   gelu_19 = Gelu <approximate: string = "none"> (linear_118)
   val_276 = MatMul (gelu_19, val_127)
   linear_119 = Add (val_276, "transformer.encoder.layer.19.output.dense.bias")
   add_2049 = Add (linear_119, layer_norm_39)
   layer_norm_40 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2049, "transformer.encoder.layer.19.output.LayerNorm.weight", "transformer.encoder.layer.19.output.LayerNorm.bias")
   val_277 = MatMul (layer_norm_40, val_128)
   linear_120 = Add (val_277, "transformer.encoder.layer.20.attention.self.query.bias")
   val_278 = MatMul (layer_norm_40, val_129)
   linear_121 = Add (val_278, "transformer.encoder.layer.20.attention.self.key.bias")
   val_279 = MatMul (layer_norm_40, val_130)
   scaled_dot_product_attention_20 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, scale: float = 0.125, softcap: float = 0> (linear_120, linear_121, val_279, bitwise_and_1)
   val_280 = MatMul (scaled_dot_product_attention_20, val_131)
   linear_123 = Add (val_280, "transformer.encoder.layer.20.attention.output.dense.bias")
   add_2122 = Add (linear_123, layer_norm_40)
   layer_norm_41 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2122, "transformer.encoder.layer.20.attention.output.LayerNorm.weight", "transformer.encoder.layer.20.attention.output.LayerNorm.bias")
   val_281 = MatMul (layer_norm_41, val_132)
   linear_124 = Add (val_281, "transformer.encoder.layer.20.intermediate.dense.bias")
   gelu_20 = Gelu <approximate: string = "none"> (linear_124)
   val_282 = MatMul (gelu_20, val_133)
   linear_125 = Add (val_282, "transformer.encoder.layer.20.output.dense.bias")
   add_2147 = Add (linear_125, layer_norm_41)
   layer_norm_42 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2147, "transformer.encoder.layer.20.output.LayerNorm.weight", "transformer.encoder.layer.20.output.LayerNorm.bias")
   val_283 = MatMul (layer_norm_42, val_134)
   linear_126 = Add (val_283, "transformer.encoder.layer.21.attention.self.query.bias")
   val_284 = MatMul (layer_norm_42, val_135)
   linear_127 = Add (val_284, "transformer.encoder.layer.21.attention.self.key.bias")
   val_285 = MatMul (layer_norm_42, val_136)
   scaled_dot_product_attention_21 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, scale: float = 0.125, softcap: float = 0> (linear_126, linear_127, val_285, bitwise_and_1)
   val_286 = MatMul (scaled_dot_product_attention_21, val_137)
   linear_129 = Add (val_286, "transformer.encoder.layer.21.attention.output.dense.bias")
   add_2220 = Add (linear_129, layer_norm_42)
   layer_norm_43 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2220, "transformer.encoder.layer.21.attention.output.LayerNorm.weight", "transformer.encoder.layer.21.attention.output.LayerNorm.bias")
   val_287 = MatMul (layer_norm_43, val_138)
   linear_130 = Add (val_287, "transformer.encoder.layer.21.intermediate.dense.bias")
   gelu_21 = Gelu <approximate: string = "none"> (linear_130)
   val_288 = MatMul (gelu_21, val_139)
   linear_131 = Add (val_288, "transformer.encoder.layer.21.output.dense.bias")
   add_2245 = Add (linear_131, layer_norm_43)
   layer_norm_44 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2245, "transformer.encoder.layer.21.output.LayerNorm.weight", "transformer.encoder.layer.21.output.LayerNorm.bias")
   val_289 = MatMul (layer_norm_44, val_140)
   linear_132 = Add (val_289, "transformer.encoder.layer.22.attention.self.query.bias")
   val_290 = MatMul (layer_norm_44, val_141)
   linear_133 = Add (val_290, "transformer.encoder.layer.22.attention.self.key.bias")
   val_291 = MatMul (layer_norm_44, val_142)
   scaled_dot_product_attention_22 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, scale: float = 0.125, softcap: float = 0> (linear_132, linear_133, val_291, bitwise_and_1)
   val_292 = MatMul (scaled_dot_product_attention_22, val_143)
   linear_135 = Add (val_292, "transformer.encoder.layer.22.attention.output.dense.bias")
   add_2318 = Add (linear_135, layer_norm_44)
   layer_norm_45 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2318, "transformer.encoder.layer.22.attention.output.LayerNorm.weight", "transformer.encoder.layer.22.attention.output.LayerNorm.bias")
   val_293 = MatMul (layer_norm_45, val_144)
   linear_136 = Add (val_293, "transformer.encoder.layer.22.intermediate.dense.bias")
   gelu_22 = Gelu <approximate: string = "none"> (linear_136)
   val_294 = MatMul (gelu_22, val_145)
   linear_137 = Add (val_294, "transformer.encoder.layer.22.output.dense.bias")
   add_2343 = Add (linear_137, layer_norm_45)
   layer_norm_46 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2343, "transformer.encoder.layer.22.output.LayerNorm.weight", "transformer.encoder.layer.22.output.LayerNorm.bias")
   val_295 = MatMul (layer_norm_46, val_146)
   linear_138 = Add (val_295, "transformer.encoder.layer.23.attention.self.query.bias")
   val_296 = MatMul (layer_norm_46, val_147)
   linear_139 = Add (val_296, "transformer.encoder.layer.23.attention.self.key.bias")
   val_297 = MatMul (layer_norm_46, val_148)
   scaled_dot_product_attention_23 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, scale: float = 0.125, softcap: float = 0> (linear_138, linear_139, val_297, bitwise_and_1)
   val_298 = MatMul (scaled_dot_product_attention_23, val_149)
   linear_141 = Add (val_298, "transformer.encoder.layer.23.attention.output.dense.bias")
   add_2416 = Add (linear_141, layer_norm_46)
   layer_norm_47 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2416, "transformer.encoder.layer.23.attention.output.LayerNorm.weight", "transformer.encoder.layer.23.attention.output.LayerNorm.bias")
   val_299 = MatMul (layer_norm_47, val_150)
   linear_142 = Add (val_299, "transformer.encoder.layer.23.intermediate.dense.bias")
   gelu_23 = Gelu <approximate: string = "none"> (linear_142)
   val_300 = MatMul (gelu_23, val_151)
   linear_143 = Add (val_300, "transformer.encoder.layer.23.output.dense.bias")
   add_2441 = Add (linear_143, layer_norm_47)
   layer_norm_48 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_2441, "transformer.encoder.layer.23.output.LayerNorm.weight", "transformer.encoder.layer.23.output.LayerNorm.bias")
   unsqueeze_12 = Unsqueeze (attention_mask, val_3)
   convert_element_type_default_1 = Cast <to: int = 1> (unsqueeze_12)
   mul_1230 = Mul (layer_norm_48, convert_element_type_default_1)
   sum_1 = ReduceSum <keepdims: int = 0, noop_with_empty_axes: int = 0> (mul_1230, val_2)
   val_301 = Cast <to: int = 1> (attention_mask)
   val_302 = ReduceSum <keepdims: int = 0, noop_with_empty_axes: int = 0> (val_301, val_2)
   convert_element_type_default_3 = Unsqueeze (val_302, val_2)
   [node_div] div = Div (sum_1, convert_element_type_default_3)
   linear_145 = Gemm <alpha: float = 1, beta: float = 1, transA: int = 0, transB: int = 1> (div, "LinearTransformation.weight", "LinearTransformation.bias")
   [node_linalg_vector_norm] linalg_vector_norm = ReduceL2 <keepdims: int = 1, noop_with_empty_axes: int = 0> (linear_145, val_0)
   [node_clamp_min] clamp_min = Clip (linalg_vector_norm, val_1)
   text_embedding = Div (linear_145, clamp_min)
}

weights:
LinearTransformation.bias FLOAT[512] 871c612bd584
LinearTransformation.weight FLOAT[512,1024] 043bd54b176c
embedding_1 FLOAT[1,1,1024] eb31ef762247
embedding_2 FLOAT[1,77,1024] 59b4cf5f82c8
transformer.embeddings.LayerNorm.bias FLOAT[1024] b63ed293de64
transformer.embeddings.LayerNorm.weight FLOAT[1024] ce3d4fe5824e
transformer.embeddings.word_embeddings.weight_fp16 FLOAT16[250002,1024] 7a523d337f34
transformer.encoder.layer.0.attention.output.LayerNorm.bias FLOAT[1024] c77366f595d3
transformer.encoder.layer.0.attention.output.LayerNorm.weight FLOAT[1024] f7be07e6d784
transformer.encoder.layer.0.attention.output.dense.bias FLOAT[1024] ff743c7b94f3
transformer.encoder.layer.0.attention.self.key.bias FLOAT[1024] 72cce75c24ea
transformer.encoder.layer.0.attention.self.query.bias FLOAT[1024] b1dd9a446169
transformer.encoder.layer.0.intermediate.dense.bias FLOAT[4096] 83f2f253fffd
transformer.encoder.layer.0.output.LayerNorm.bias FLOAT[1024] 4d09d0ce3907
transformer.encoder.layer.0.output.LayerNorm.weight FLOAT[1024] a5bc0f2eabb6
transformer.encoder.layer.0.output.dense.bias FLOAT[1024] c3c26bc0923f
transformer.encoder.layer.1.attention.output.LayerNorm.bias FLOAT[1024] 2d0d6b084b5d
transformer.encoder.layer.1.attention.output.LayerNorm.weight FLOAT[1024] aa3eaa3c7fd7
transformer.encoder.layer.1.attention.output.dense.bias FLOAT[1024] bf5fc776c53c
transformer.encoder.layer.1.attention.self.key.bias FLOAT[1024] 1dd2619f4a5c
transformer.encoder.layer.1.attention.self.query.bias FLOAT[1024] 67e3d4fe1dd5
transformer.encoder.layer.1.intermediate.dense.bias FLOAT[4096] 23735f3adde6
transformer.encoder.layer.1.output.LayerNorm.bias FLOAT[1024] 508772a1f9ad
transformer.encoder.layer.1.output.LayerNorm.weight FLOAT[1024] 4308d943537c
transformer.encoder.layer.1.output.dense.bias FLOAT[1024] 783ea320c30b
transformer.encoder.layer.10.attention.output.LayerNorm.bias FLOAT[1024] 0b20ee8e4c4c
transformer.encoder.layer.10.attention.output.LayerNorm.weight FLOAT[1024] 1594146f5492
transformer.encoder.layer.10.attention.output.dense.bias FLOAT[1024] f1badad6cb96
transformer.encoder.layer.10.attention.self.key.bias FLOAT[1024] 629444e0165b
transformer.encoder.layer.10.attention.self.query.bias FLOAT[1024] c8e8800aaf27
transformer.encoder.layer.10.intermediate.dense.bias FLOAT[4096] 7f9436273f8e
transformer.encoder.layer.10.output.LayerNorm.bias FLOAT[1024] 39a711df072f
transformer.encoder.layer.10.output.LayerNorm.weight FLOAT[1024] 242c4349bd50
transformer.encoder.layer.10.output.dense.bias FLOAT[1024] 4c25fd82b9e5
transformer.encoder.layer.11.attention.output.LayerNorm.bias FLOAT[1024] bd7aec54fe54
transformer.encoder.layer.11.attention.output.LayerNorm.weight FLOAT[1024] d24f098194e0
transformer.encoder.layer.11.attention.output.dense.bias FLOAT[1024] 1040798ba493
transformer.encoder.layer.11.attention.self.key.bias FLOAT[1024] abcf53626f8f
transformer.encoder.layer.11.attention.self.query.bias FLOAT[1024] fe8b604cef0e
transformer.encoder.layer.11.intermediate.dense.bias FLOAT[4096] d8b0b89c4ea2
transformer.encoder.layer.11.output.LayerNorm.bias FLOAT[1024] b6866c67fe3a
transformer.encoder.layer.11.output.LayerNorm.weight FLOAT[1024] a591760cdefd
transformer.encoder.layer.11.output.dense.bias FLOAT[1024] 0c4a5fb61555
transformer.encoder.layer.12.attention.output.LayerNorm.bias FLOAT[1024] 4b2ff19889a1
transformer.encoder.layer.12.attention.output.LayerNorm.weight FLOAT[1024] b81260926bf6
transformer.encoder.layer.12.attention.output.dense.bias FLOAT[1024] 1ae94971f42b
transformer.encoder.layer.12.attention.self.key.bias FLOAT[1024] 8e0f2464f7e3
transformer.encoder.layer.12.attention.self.query.bias FLOAT[1024] bdb2796ae248
transformer.encoder.layer.12.intermediate.dense.bias FLOAT[4096] 39c6d4097353
transformer.encoder.layer.12.output.LayerNorm.bias FLOAT[1024] 0c60d32469b5
transformer.encoder.layer.12.output.LayerNorm.weight FLOAT[1024] 371159ad64d9
transformer.encoder.layer.12.output.dense.bias FLOAT[1024] fedecbc072c0
transformer.encoder.layer.13.attention.output.LayerNorm.bias FLOAT[1024] 514b788fe560
transformer.encoder.layer.13.attention.output.LayerNorm.weight FLOAT[1024] 47c3453a9af5
transformer.encoder.layer.13.attention.output.dense.bias FLOAT[1024] b62bccf65cc4
transformer.encoder.layer.13.attention.self.key.bias FLOAT[1024] 4d5614884cd6
transformer.encoder.layer.13.attention.self.query.bias FLOAT[1024] 01f760257fc0
transformer.encoder.layer.13.intermediate.dense.bias FLOAT[4096] 71ba742b6236
transformer.encoder.layer.13.output.LayerNorm.bias FLOAT[1024] c8267f8a1441
transformer.encoder.layer.13.output.LayerNorm.weight FLOAT[1024] 5fe9b8121c09
transformer.encoder.layer.13.output.dense.bias FLOAT[1024] 71a732aafc70
transformer.encoder.layer.14.attention.output.LayerNorm.bias FLOAT[1024] 4c7777e68b67
transformer.encoder.layer.14.attention.output.LayerNorm.weight FLOAT[1024] bc09a0660845
transformer.encoder.layer.14.attention.output.dense.bias FLOAT[1024] 5b16f0895357
transformer.encoder.layer.14.attention.self.key.bias FLOAT[1024] b9ba8c70debe
transformer.encoder.layer.14.attention.self.query.bias FLOAT[1024] 7d7db7020d08
transformer.encoder.layer.14.intermediate.dense.bias FLOAT[4096] 89b86a392fe3
transformer.encoder.layer.14.output.LayerNorm.bias FLOAT[1024] b6bb93213362
transformer.encoder.layer.14.output.LayerNorm.weight FLOAT[1024] 8c313ae62b28
transformer.encoder.layer.14.output.dense.bias FLOAT[1024] 682d655bf2de
transformer.encoder.layer.15.attention.output.LayerNorm.bias FLOAT[1024] 6b564b3a0c02
transformer.encoder.layer.15.attention.output.LayerNorm.weight FLOAT[1024] 4de6d80b2cff
transformer.encoder.layer.15.attention.output.dense.bias FLOAT[1024] f7c616a768ed
transformer.encoder.layer.15.attention.self.key.bias FLOAT[1024] 71c949ce9e79
transformer.encoder.layer.15.attention.self.query.bias FLOAT[1024] e767be4ccecf
transformer.encoder.layer.15.intermediate.dense.bias FLOAT[4096] 3012ac99bee5
transformer.encoder.layer.15.output.LayerNorm.bias FLOAT[1024] 30343869cc41
transformer.encoder.layer.15.output.LayerNorm.weight FLOAT[1024] 31b2bfd411d1
transformer.encoder.layer.15.output.dense.bias FLOAT[1024] e01cc30c2bcd
transformer.encoder.layer.16.attention.output.LayerNorm.bias FLOAT[1024] 59857ae76d5f
transformer.encoder.layer.16.attention.output.LayerNorm.weight FLOAT[1024] 641f5c009ad2
transformer.encoder.layer.16.attention.output.dense.bias FLOAT[1024] 9d4956d03790
transformer.encoder.layer.16.attention.self.key.bias FLOAT[1024] 77948247192c
transformer.encoder.layer.16.attention.self.query.bias FLOAT[1024] d9165d56f175
transformer.encoder.layer.16.intermediate.dense.bias FLOAT[4096] aa59e69e2cdf
transformer.encoder.layer.16.output.LayerNorm.bias FLOAT[1024] 8f651c91f169
transformer.encoder.layer.16.output.LayerNorm.weight FLOAT[1024] b66e2e0cf7b2
transformer.encoder.layer.16.output.dense.bias FLOAT[1024] 638361a88c74
transformer.encoder.layer.17.attention.output.LayerNorm.bias FLOAT[1024] ad97ecadfe14
transformer.encoder.layer.17.attention.output.LayerNorm.weight FLOAT[1024] 548540b97363
transformer.encoder.layer.17.attention.output.dense.bias FLOAT[1024] 8bb7af910743
transformer.encoder.layer.17.attention.self.key.bias FLOAT[1024] d68f6bbaadd2
transformer.encoder.layer.17.attention.self.query.bias FLOAT[1024] 5da7c129449e
transformer.encoder.layer.17.intermediate.dense.bias FLOAT[4096] 09dae2b89b36
transformer.encoder.layer.17.output.LayerNorm.bias FLOAT[1024] a8efad333fd5
transformer.encoder.layer.17.output.LayerNorm.weight FLOAT[1024] 1b4d743d12ca
transformer.encoder.layer.17.output.dense.bias FLOAT[1024] c9a62f14d6c1
transformer.encoder.layer.18.attention.output.LayerNorm.bias FLOAT[1024] 49346f1e4f1a
transformer.encoder.layer.18.attention.output.LayerNorm.weight FLOAT[1024] ec566123f707
transformer.encoder.layer.18.attention.output.dense.bias FLOAT[1024] 2d9829e378fc
transformer.encoder.layer.18.attention.self.key.bias FLOAT[1024] 2d1f3ddfe264
transformer.encoder.layer.18.attention.self.query.bias FLOAT[1024] 0eb2e9bf153c
transformer.encoder.layer.18.intermediate.dense.bias FLOAT[4096] 8d7cb07e4e48
transformer.encoder.layer.18.output.LayerNorm.bias FLOAT[1024] 16caaef60a18
transformer.encoder.layer.18.output.LayerNorm.weight FLOAT[1024] 4cb6973a2ca1
transformer.encoder.layer.18.output.dense.bias FLOAT[1024] d3a09b681a19
transformer.encoder.layer.19.attention.output.LayerNorm.bias FLOAT[1024] 509137fe6502
transformer.encoder.layer.19.attention.output.LayerNorm.weight FLOAT[1024] ce3b21b50080
transformer.encoder.layer.19.attention.output.dense.bias FLOAT[1024] d1202038d83e
transformer.encoder.layer.19.attention.self.key.bias FLOAT[1024] 7755bfd6038f
transformer.encoder.layer.19.attention.self.query.bias FLOAT[1024] bd30be9f4bf1
transformer.encoder.layer.19.intermediate.dense.bias FLOAT[4096] c0a163c6ed53
transformer.encoder.layer.19.output.LayerNorm.bias FLOAT[1024] d65ea273d367
transformer.encoder.layer.19.output.LayerNorm.weight FLOAT[1024] e04ad9285bb8
transformer.encoder.layer.19.output.dense.bias FLOAT[1024] d8176baf44ff
transformer.encoder.layer.2.attention.output.LayerNorm.bias FLOAT[1024] 9721094baa9f
transformer.encoder.layer.2.attention.output.LayerNorm.weight FLOAT[1024] 15c5d80933a1
transformer.encoder.layer.2.attention.output.dense.bias FLOAT[1024] 1127a2ef5b89
transformer.encoder.layer.2.attention.self.key.bias FLOAT[1024] f16c740e23c3
transformer.encoder.layer.2.attention.self.query.bias FLOAT[1024] 1b20a59ff8eb
transformer.encoder.layer.2.intermediate.dense.bias FLOAT[4096] 15bd896b4aa1
transformer.encoder.layer.2.output.LayerNorm.bias FLOAT[1024] c23858ca7949
transformer.encoder.layer.2.output.LayerNorm.weight FLOAT[1024] 650d2b2c33b3
transformer.encoder.layer.2.output.dense.bias FLOAT[1024] 9326a8256094
transformer.encoder.layer.20.attention.output.LayerNorm.bias FLOAT[1024] 15434898851e
transformer.encoder.layer.20.attention.output.LayerNorm.weight FLOAT[1024] 5b454400b96c
transformer.encoder.layer.20.attention.output.dense.bias FLOAT[1024] b360422cd171
transformer.encoder.layer.20.attention.self.key.bias FLOAT[1024] 3fbd6517605f
transformer.encoder.layer.20.attention.self.query.bias FLOAT[1024] e6e44dbbcc08
transformer.encoder.layer.20.intermediate.dense.bias FLOAT[4096] 921e10109b31
transformer.encoder.layer.20.output.LayerNorm.bias FLOAT[1024] 886daf96ca81
transformer.encoder.layer.20.output.LayerNorm.weight FLOAT[1024] c785c668182e
transformer.encoder.layer.20.output.dense.bias FLOAT[1024] 1e8e6a49238a
transformer.encoder.layer.21.attention.output.LayerNorm.bias FLOAT[1024] a1f98ca42a52
transformer.encoder.layer.21.attention.output.LayerNorm.weight FLOAT[1024] 7c1141f9eda2
transformer.encoder.layer.21.attention.output.dense.bias FLOAT[1024] 823c3c243feb
transformer.encoder.layer.21.attention.self.key.bias FLOAT[1024] e5ce1d467eeb
transformer.encoder.layer.21.attention.self.query.bias FLOAT[1024] ee90469f476c
transformer.encoder.layer.21.intermediate.dense.bias FLOAT[4096] 7c2629f6d96a
transformer.encoder.layer.21.output.LayerNorm.bias FLOAT[1024] 883bb5258fb8
transformer.encoder.layer.21.output.LayerNorm.weight FLOAT[1024] e76139f65367
transformer.encoder.layer.21.output.dense.bias FLOAT[1024] 072394c5dc83
transformer.encoder.layer.22.attention.output.LayerNorm.bias FLOAT[1024] ba5d1259372d
transformer.encoder.layer.22.attention.output.LayerNorm.weight FLOAT[1024] 4363731a3ce6
transformer.encoder.layer.22.attention.output.dense.bias FLOAT[1024] ebb631162077
transformer.encoder.layer.22.attention.self.key.bias FLOAT[1024] 77cb897ca4f7
transformer.encoder.layer.22.attention.self.query.bias FLOAT[1024] 29044ed7a8f0
transformer.encoder.layer.22.intermediate.dense.bias FLOAT[4096] 321096330ec8
transformer.encoder.layer.22.output.LayerNorm.bias FLOAT[1024] 7238f0cfe24e
transformer.encoder.layer.22.output.LayerNorm.weight FLOAT[1024] 00d2a33cd46a
transformer.encoder.layer.22.output.dense.bias FLOAT[1024] 0ac6108d978c
transformer.encoder.layer.23.attention.output.LayerNorm.bias FLOAT[1024] 8e37a27801c3
transformer.encoder.layer.23.attention.output.LayerNorm.weight FLOAT[1024] 234097b1a750
transformer.encoder.layer.23.attention.output.dense.bias FLOAT[1024] c3e5f57efc4c
transformer.encoder.layer.23.attention.self.key.bias FLOAT[1024] e9caa2f3f8d6
transformer.encoder.layer.23.attention.self.query.bias FLOAT[1024] 7b4422194537
transformer.encoder.layer.23.intermediate.dense.bias FLOAT[4096] 88630c13ff76
transformer.encoder.layer.23.output.LayerNorm.bias FLOAT[1024] 92ab8b07b570
transformer.encoder.layer.23.output.LayerNorm.weight FLOAT[1024] ed880373c722
transformer.encoder.layer.23.output.dense.bias FLOAT[1024] 8b02af070180
transformer.encoder.layer.3.attention.output.LayerNorm.bias FLOAT[1024] fff45c381b9c
transformer.encoder.layer.3.attention.output.LayerNorm.weight FLOAT[1024] b7ac1c6f0137
transformer.encoder.layer.3.attention.output.dense.bias FLOAT[1024] 8a09f9dcadbd
transformer.encoder.layer.3.attention.self.key.bias FLOAT[1024] 6c627af2a0a0
transformer.encoder.layer.3.attention.self.query.bias FLOAT[1024] ecffb92b3ca0
transformer.encoder.layer.3.intermediate.dense.bias FLOAT[4096] 88ed2eac47f3
transformer.encoder.layer.3.output.LayerNorm.bias FLOAT[1024] 9f40de613725
transformer.encoder.layer.3.output.LayerNorm.weight FLOAT[1024] 965e7ccb8a95
transformer.encoder.layer.3.output.dense.bias FLOAT[1024] 5bffa1e0bbc6
transformer.encoder.layer.4.attention.output.LayerNorm.bias FLOAT[1024] f698c3a69233
transformer.encoder.layer.4.attention.output.LayerNorm.weight FLOAT[1024] b267919edb6a
transformer.encoder.layer.4.attention.output.dense.bias FLOAT[1024] 70ca1c793513
transformer.encoder.layer.4.attention.self.key.bias FLOAT[1024] 3465a3a6c0cd
transformer.encoder.layer.4.attention.self.query.bias FLOAT[1024] 5a89566dd80a
transformer.encoder.layer.4.intermediate.dense.bias FLOAT[4096] 12db4a55201e
transformer.encoder.layer.4.output.LayerNorm.bias FLOAT[1024] 521c587e0612
transformer.encoder.layer.4.output.LayerNorm.weight FLOAT[1024] 3e70d5c6281c
transformer.encoder.layer.4.output.dense.bias FLOAT[1024] 433539f4321c
transformer.encoder.layer.5.attention.output.LayerNorm.bias FLOAT[1024] 63f6f4260f3c
transformer.encoder.layer.5.attention.output.LayerNorm.weight FLOAT[1024] 5a81978dde1d
transformer.encoder.layer.5.attention.output.dense.bias FLOAT[1024] 422de24cf289
transformer.encoder.layer.5.attention.self.key.bias FLOAT[1024] 8263dde85eaf
transformer.encoder.layer.5.attention.self.query.bias FLOAT[1024] fec9b4ea87a8
transformer.encoder.layer.5.intermediate.dense.bias FLOAT[4096] 35ba969c55ea
transformer.encoder.layer.5.output.LayerNorm.bias FLOAT[1024] d62fba7d66a2
transformer.encoder.layer.5.output.LayerNorm.weight FLOAT[1024] 99e7fa1940c2
transformer.encoder.layer.5.output.dense.bias FLOAT[1024] 845e0fb3a4c6
transformer.encoder.layer.6.attention.output.LayerNorm.bias FLOAT[1024] 77b5a13828cf
transformer.encoder.layer.6.attention.output.LayerNorm.weight FLOAT[1024] fd54f430086a
transformer.encoder.layer.6.attention.output.dense.bias FLOAT[1024] 521794e677a2
transformer.encoder.layer.6.attention.self.key.bias FLOAT[1024] 70c7efca9f91
transformer.encoder.layer.6.attention.self.query.bias FLOAT[1024] 2d37831b4d23
transformer.encoder.layer.6.intermediate.dense.bias FLOAT[4096] 89aaa5bb7509
transformer.encoder.layer.6.output.LayerNorm.bias FLOAT[1024] 164ba687796e
transformer.encoder.layer.6.output.LayerNorm.weight FLOAT[1024] e17a4b90abff
transformer.encoder.layer.6.output.dense.bias FLOAT[1024] 7756628b4b63
transformer.encoder.layer.7.attention.output.LayerNorm.bias FLOAT[1024] 1fd95783722c
transformer.encoder.layer.7.attention.output.LayerNorm.weight FLOAT[1024] 437f2a2766bb
transformer.encoder.layer.7.attention.output.dense.bias FLOAT[1024] 0d15b0175b44
transformer.encoder.layer.7.attention.self.key.bias FLOAT[1024] a2efc562efce
transformer.encoder.layer.7.attention.self.query.bias FLOAT[1024] 4ce11b6a3ed6
transformer.encoder.layer.7.intermediate.dense.bias FLOAT[4096] 2b537384cfa9
transformer.encoder.layer.7.output.LayerNorm.bias FLOAT[1024] 6a266842cf90
transformer.encoder.layer.7.output.LayerNorm.weight FLOAT[1024] 94651e6cbe0d
transformer.encoder.layer.7.output.dense.bias FLOAT[1024] 722f4861cbad
transformer.encoder.layer.8.attention.output.LayerNorm.bias FLOAT[1024] 2d319a36580c
transformer.encoder.layer.8.attention.output.LayerNorm.weight FLOAT[1024] bf7ba472fc73
transformer.encoder.layer.8.attention.output.dense.bias FLOAT[1024] 0bb37995187f
transformer.encoder.layer.8.attention.self.key.bias FLOAT[1024] f333cba66cc0
transformer.encoder.layer.8.attention.self.query.bias FLOAT[1024] 29de6e736e1b
transformer.encoder.layer.8.intermediate.dense.bias FLOAT[4096] 44275d10d69e
transformer.encoder.layer.8.output.LayerNorm.bias FLOAT[1024] 2bc2e3ca9b4e
transformer.encoder.layer.8.output.LayerNorm.weight FLOAT[1024] f9ce70e1e93a
transformer.encoder.layer.8.output.dense.bias FLOAT[1024] 0223e79b09ce
transformer.encoder.layer.9.attention.output.LayerNorm.bias FLOAT[1024] 2c7f8dd60555
transformer.encoder.layer.9.attention.output.LayerNorm.weight FLOAT[1024] d27d569833d3
transformer.encoder.layer.9.attention.output.dense.bias FLOAT[1024] acf4bc8efc0d
transformer.encoder.layer.9.attention.self.key.bias FLOAT[1024] 562426d19677
transformer.encoder.layer.9.attention.self.query.bias FLOAT[1024] 24af995385bf
transformer.encoder.layer.9.intermediate.dense.bias FLOAT[4096] 93f0597d0bef
transformer.encoder.layer.9.output.LayerNorm.bias FLOAT[1024] 07c97e8bb390
transformer.encoder.layer.9.output.LayerNorm.weight FLOAT[1024] cd13c6d62f41
transformer.encoder.layer.9.output.dense.bias FLOAT[1024] d06bb566c5d1
val_0 INT64[1] 12a3ae445661
val_1 FLOAT[] 6708d9be4956
val_10 FLOAT[1024,1024] 8c9c947b18a8
val_100 FLOAT[1024,1024] 6524381a33f7
val_101 FLOAT[1024,1024] e1b2470b9c9d
val_102 FLOAT[1024,4096] 43070f98268f
val_103 FLOAT[4096,1024] 7e91c4bdf160
val_104 FLOAT[1024,1024] 61b5345a5fd0
val_105 FLOAT[1024,1024] 57d4b3920257
val_106 FLOAT[1024,1024] 82d8d680de55
val_107 FLOAT[1024,1024] e8a7553a9161
val_108 FLOAT[1024,4096] 6e52ec2c6e1c
val_109 FLOAT[4096,1024] 756a5125a5ac
val_11 FLOAT[1024,1024] a5a146512ae9
val_110 FLOAT[1024,1024] 8c01fba202f0
val_111 FLOAT[1024,1024] e8a7db49cf9f
val_112 FLOAT[1024,1024] 48d54774be4c
val_113 FLOAT[1024,1024] 96880ca1f4b0
val_114 FLOAT[1024,4096] ca3594014518
val_115 FLOAT[4096,1024] 548416093c8d
val_116 FLOAT[1024,1024] 37888c2889c2
val_117 FLOAT[1024,1024] 09f90eae83aa
val_118 FLOAT[1024,1024] bf1c52db4fee
val_119 FLOAT[1024,1024] 259df1b4d93a
val_12 FLOAT[1024,4096] db91fadec725
val_120 FLOAT[1024,4096] 4c1acdd9c8f2
val_121 FLOAT[4096,1024] 03533645535c
val_122 FLOAT[1024,1024] fea0a8aa615a
val_123 FLOAT[1024,1024] 45854c2d2968
val_124 FLOAT[1024,1024] 6e23a3a4a0f2
val_125 FLOAT[1024,1024] 0998d8cd468f
val_126 FLOAT[1024,4096] 0b36a7f8cc9b
val_127 FLOAT[4096,1024] 4c7f487fbae8
val_128 FLOAT[1024,1024] 94d46542b4f7
val_129 FLOAT[1024,1024] 4c65434df16c
val_13 FLOAT[4096,1024] 62b8dab66039
val_130 FLOAT[1024,1024] 41fa6d1e355a
val_131 FLOAT[1024,1024] 462eef2fd901
val_132 FLOAT[1024,4096] bbf44621f8a5
val_133 FLOAT[4096,1024] 56bdd409e0ce
val_134 FLOAT[1024,1024] cc0ae64fc135
val_135 FLOAT[1024,1024] 7b14bbdd9c85
val_136 FLOAT[1024,1024] 7134d7baa6e1
val_137 FLOAT[1024,1024] 29d50c7b539e
val_138 FLOAT[1024,4096] b3dd6e575675
val_139 FLOAT[4096,1024] 3699816f19c6
val_14 FLOAT[1024,1024] 8a6c1045893c
val_140 FLOAT[1024,1024] 12614d08de07
val_141 FLOAT[1024,1024] e347b685fda4
val_142 FLOAT[1024,1024] e73586e30cbd
val_143 FLOAT[1024,1024] 801c3e933275
val_144 FLOAT[1024,4096] 69066ef8f380
val_145 FLOAT[4096,1024] 996320b7052c
val_146 FLOAT[1024,1024] 4f7c978136c7
val_147 FLOAT[1024,1024] cefaae9f650c
val_148 FLOAT[1024,1024] 8cb09755827f
val_149 FLOAT[1024,1024] 370dce804259
val_15 FLOAT[1024,1024] 5814dbeac81d
val_150 FLOAT[1024,4096] dbe1c32e313b
val_151 FLOAT[4096,1024] f336bc4f3df9
val_16 FLOAT[1024,1024] 21aa266946ab
val_17 FLOAT[1024,1024] 1668ae5aa161
val_18 FLOAT[1024,4096] d0e26f64c299
val_19 FLOAT[4096,1024] da749b84b2ae
val_2 INT64[1] 7c9fa136d441
val_20 FLOAT[1024,1024] 50314236553f
val_21 FLOAT[1024,1024] 8456b9ce9ec3
val_22 FLOAT[1024,1024] 689936ada5af
val_23 FLOAT[1024,1024] 9be01bdeb7f8
val_24 FLOAT[1024,4096] 2d06fdb86bbc
val_25 FLOAT[4096,1024] 083734335e61
val_26 FLOAT[1024,1024] e7e1ff3fbd8e
val_27 FLOAT[1024,1024] 76e6a666e602
val_28 FLOAT[1024,1024] ca9a7285977e
val_29 FLOAT[1024,1024] fec1feda913d
val_3 INT64[1] d86e8112f3c4
val_30 FLOAT[1024,4096] f04caed8e403
val_31 FLOAT[4096,1024] 43f55646d6a6
val_32 FLOAT[1024,1024] 3ab5e9fac309
val_33 FLOAT[1024,1024] 0d053950318f
val_34 FLOAT[1024,1024] b54560798079
val_35 FLOAT[1024,1024] 1bea91193f6a
val_36 FLOAT[1024,4096] a73a07cd3255
val_37 FLOAT[4096,1024] 8074d761099d
val_38 FLOAT[1024,1024] 70384388558f
val_39 FLOAT[1024,1024] 151003f3e53b
val_4 FLOAT[] e00e5eb94441
val_40 FLOAT[1024,1024] 9e9cd2b24deb
val_41 FLOAT[1024,1024] 3e4fbfd311d3
val_42 FLOAT[1024,4096] 228c146d8aaa
val_43 FLOAT[4096,1024] 5bea9126b522
val_44 FLOAT[1024,1024] 322cd45dea4d
val_45 FLOAT[1024,1024] a5c1733fd423
val_46 FLOAT[1024,1024] fed67c226c22
val_47 FLOAT[1024,1024] 24569b58657e
val_48 FLOAT[1024,4096] 6d98e8383173
val_49 FLOAT[4096,1024] df8d03d0b314
val_5 FLOAT[] e401200e5808
val_50 FLOAT[1024,1024] 4e5600703c8f
val_51 FLOAT[1024,1024] 8b2636e59229
val_52 FLOAT[1024,1024] 4ad0d2fde264
val_53 FLOAT[1024,1024] 15fd1bf550cf
val_54 FLOAT[1024,4096] 28536f727d1f
val_55 FLOAT[4096,1024] 7c03ba66d2c0
val_56 FLOAT[1024,1024] 3ef936f458cf
val_57 FLOAT[1024,1024] b02322c58d16
val_58 FLOAT[1024,1024] ab86b6398742
val_59 FLOAT[1024,1024] c6d4b5b39048
val_6 INT64[2] 0c730b69905c
val_60 FLOAT[1024,4096] ab4c912c6c75
val_61 FLOAT[4096,1024] 390f74e676c7
val_62 FLOAT[1024,1024] 6ca4c3fa0bd5
val_63 FLOAT[1024,1024] 018db1eda3f9
val_64 FLOAT[1024,1024] bf655802c94c
val_65 FLOAT[1024,1024] b3fc64b27ff9
val_66 FLOAT[1024,4096] d70f0c9c997a
val_67 FLOAT[4096,1024] c6445235042d
val_68 FLOAT[1024,1024] 49bc501d7a0c
val_69 FLOAT[1024,1024] 691b9c0fa7e7
val_7 FLOAT[77,1] 9575b2125169
val_70 FLOAT[1024,1024] fc50d5accdfa
val_71 FLOAT[1024,1024] 81ae54d9d4e3
val_72 FLOAT[1024,4096] 0ed4168e14ab
val_73 FLOAT[4096,1024] 3bc1ac25d836
val_74 FLOAT[1024,1024] c85bfacbb418
val_75 FLOAT[1024,1024] 6d59c85a321c
val_76 FLOAT[1024,1024] 69ee991c168e
val_77 FLOAT[1024,1024] cf33e7fe9a8a
val_78 FLOAT[1024,4096] e40d3b2b2bcf
val_79 FLOAT[4096,1024] 32a5419ddec3
val_8 FLOAT[1024,1024] 9165cc0332c7
val_80 FLOAT[1024,1024] 43182abdf233
val_81 FLOAT[1024,1024] c9de450ad2aa
val_82 FLOAT[1024,1024] 4c7ebc3c2dad
val_83 FLOAT[1024,1024] 32a34b7fd7fe
val_84 FLOAT[1024,4096] fa469f3963b7
val_85 FLOAT[4096,1024] d2e59eea722f
val_86 FLOAT[1024,1024] 50ea0e54c234
val_87 FLOAT[1024,1024] cd45161843c8
val_88 FLOAT[1024,1024] 6ee9ff041843
val_89 FLOAT[1024,1024] 1c86733b3dec
val_9 FLOAT[1024,1024] d99b24a3e12e
val_90 FLOAT[1024,4096] 6525b04140c2
val_91 FLOAT[4096,1024] 64d81c0ed8c4
val_92 FLOAT[1024,1024] 743e95332636
val_93 FLOAT[1024,1024] e3168cecf59a
val_94 FLOAT[1024,1024] 8872caefcc1c
val_95 FLOAT[1024,1024] b504911ea50f
val_96 FLOAT[1024,4096] ed662bbb67cb
val_97 FLOAT[4096,1024] 43c1cbfad2ae
val_98 FLOAT[1024,1024] f420b7789544
val_99 FLOAT[1024,1024] 2277bf3b9639
