<
   ir_version: 10,
   opset_import: ["" : 23],
   producer_name: "pytorch"
>
main_graph (uint8[batch,384,384,3] image) => (float[batch,768] image_embedding) 
   <
      float[batch,768,48,48] add_1072
      float[batch,768,48,48] add_1148
      float[batch,1536,24,24] add_1244
      float[batch,1536,24,24] add_1320
      float[batch,1536,24,24] add_1396
      float[batch,384,96,96] add_140
      float[batch,1536,24,24] add_1472
      float[batch,1536,24,24] add_1548
      float[batch,1536,24,24] add_1624
      float[batch,1536,24,24] add_1700
      float[batch,1536,24,24] add_1776
      float[batch,1536,24,24] add_1852
      float[batch,1536,24,24] add_1928
      float[batch,1536,24,24] add_2004
      float[batch,1536,24,24] add_2080
      float[batch,1536,24,24] add_2156
      float[batch,384,96,96] add_216
      float[batch,1536,24,24] add_2232
      float[batch,1536,24,24] add_2308
      float[batch,1536,24,24] add_2384
      float[batch,1536,24,24] add_2460
      float[batch,1536,24,24] add_2536
      float[batch,3072,12,12] add_2632
      float[batch,3072,12,12] add_2708
      float[batch,3072,12,12] add_2784
      float[batch,3072,12,12] add_2860
      float[batch,384,96,96] add_292
      float[batch,3072,12,12] add_2936
      float[batch,3072,12,12] add_3012
      float[batch,3072,12,12] add_3088
      float[batch,3072,12,12] add_3164
      float[145,batch,3072] add_3194
      float[batch,384,96,96] add_368
      float[batch,384,96,96] add_444
      float[batch,384,96,96] add_520
      float[batch,768,48,48] add_616
      float[batch,768,48,48] add_692
      float[batch,768,48,48] add_768
      float[batch,768,48,48] add_844
      float[batch,768,48,48] add_920
      float[batch,768,48,48] add_996
      float[batch,96,96,96] avg_pool2d
      float[batch,192,48,48] avg_pool2d_2
      float[batch,384,48,48] avg_pool2d_3
      float[batch,384,24,24] avg_pool2d_4
      float[batch,768,24,24] avg_pool2d_5
      float[batch,768,12,12] avg_pool2d_6
      float[batch,1536,12,12] avg_pool2d_7
      float[145,batch,3072] cat
      float[batch,1] clamp_min
      float[batch,48,192,192] getitem
      float[batch,768,48,48] getitem_102
      float[batch,192,48,48] getitem_105
      float[batch,192,48,48] getitem_108
      float[batch,768,48,48] getitem_111
      float[batch,192,48,48] getitem_114
      float[batch,192,48,48] getitem_117
      float[batch,96,96,96] getitem_12
      float[batch,768,48,48] getitem_120
      float[batch,192,48,48] getitem_123
      float[batch,192,48,48] getitem_126
      float[batch,768,48,48] getitem_129
      float[batch,192,48,48] getitem_132
      float[batch,192,48,48] getitem_135
      float[batch,768,48,48] getitem_138
      float[batch,384,48,48] getitem_141
      float[batch,384,48,48] getitem_144
      float[batch,1536,24,24] getitem_147
      float[batch,384,96,96] getitem_15
      float[batch,1536,24,24] getitem_150
      float[batch,384,24,24] getitem_153
      float[batch,384,24,24] getitem_156
      float[batch,1536,24,24] getitem_159
      float[batch,384,24,24] getitem_162
      float[batch,384,24,24] getitem_165
      float[batch,1536,24,24] getitem_168
      float[batch,384,24,24] getitem_171
      float[batch,384,24,24] getitem_174
      float[batch,1536,24,24] getitem_177
      float[batch,384,96,96] getitem_18
      float[batch,384,24,24] getitem_180
      float[batch,384,24,24] getitem_183
      float[batch,1536,24,24] getitem_186
      float[batch,384,24,24] getitem_189
      float[batch,384,24,24] getitem_192
      float[batch,1536,24,24] getitem_195
      float[batch,384,24,24] getitem_198
      float[batch,384,24,24] getitem_201
      float[batch,1536,24,24] getitem_204
      float[batch,384,24,24] getitem_207
      float[batch,96,96,96] getitem_21
      float[batch,384,24,24] getitem_210
      float[batch,1536,24,24] getitem_213
      float[batch,384,24,24] getitem_216
      float[batch,384,24,24] getitem_219
      float[batch,1536,24,24] getitem_222
      float[batch,384,24,24] getitem_225
      float[batch,384,24,24] getitem_228
      float[batch,1536,24,24] getitem_231
      float[batch,384,24,24] getitem_234
      float[batch,384,24,24] getitem_237
      float[batch,96,96,96] getitem_24
      float[batch,1536,24,24] getitem_240
      float[batch,384,24,24] getitem_243
      float[batch,384,24,24] getitem_246
      float[batch,1536,24,24] getitem_249
      float[batch,384,24,24] getitem_252
      float[batch,384,24,24] getitem_255
      float[batch,1536,24,24] getitem_258
      float[batch,384,24,24] getitem_261
      float[batch,384,24,24] getitem_264
      float[batch,1536,24,24] getitem_267
      float[batch,384,96,96] getitem_27
      float[batch,384,24,24] getitem_270
      float[batch,384,24,24] getitem_273
      float[batch,1536,24,24] getitem_276
      float[batch,384,24,24] getitem_279
      float[batch,384,24,24] getitem_282
      float[batch,1536,24,24] getitem_285
      float[batch,384,24,24] getitem_288
      float[batch,384,24,24] getitem_291
      float[batch,1536,24,24] getitem_294
      float[batch,384,24,24] getitem_297
      float[batch,48,192,192] getitem_3
      float[batch,96,96,96] getitem_30
      float[batch,384,24,24] getitem_300
      float[batch,1536,24,24] getitem_303
      float[batch,768,24,24] getitem_306
      float[batch,768,24,24] getitem_309
      float[batch,3072,12,12] getitem_312
      float[batch,3072,12,12] getitem_315
      float[batch,768,12,12] getitem_318
      float[batch,768,12,12] getitem_321
      float[batch,3072,12,12] getitem_324
      float[batch,768,12,12] getitem_327
      float[batch,96,96,96] getitem_33
      float[batch,768,12,12] getitem_330
      float[batch,3072,12,12] getitem_333
      float[batch,768,12,12] getitem_336
      float[batch,768,12,12] getitem_339
      float[batch,3072,12,12] getitem_342
      float[batch,768,12,12] getitem_345
      float[batch,768,12,12] getitem_348
      float[batch,3072,12,12] getitem_351
      float[batch,768,12,12] getitem_354
      float[batch,768,12,12] getitem_357
      float[batch,384,96,96] getitem_36
      float[batch,3072,12,12] getitem_360
      float[batch,768,12,12] getitem_363
      float[batch,768,12,12] getitem_366
      float[batch,3072,12,12] getitem_369
      float[batch,768,12,12] getitem_372
      float[batch,768,12,12] getitem_375
      float[batch,3072,12,12] getitem_378
      float[batch,96,96,96] getitem_39
      float[batch,96,96,96] getitem_42
      float[batch,384,96,96] getitem_45
      float[batch,96,96,96] getitem_48
      float[batch,96,96,96] getitem_51
      float[batch,384,96,96] getitem_54
      float[batch,96,96,96] getitem_57
      float[batch,96,192,192] getitem_6
      float[batch,96,96,96] getitem_60
      float[batch,384,96,96] getitem_63
      float[batch,192,96,96] getitem_66
      float[batch,192,96,96] getitem_69
      float[batch,768,48,48] getitem_72
      float[batch,768,48,48] getitem_75
      float[batch,192,48,48] getitem_78
      float[batch,192,48,48] getitem_81
      float[batch,768,48,48] getitem_84
      float[batch,192,48,48] getitem_87
      float[batch,96,96,96] getitem_9
      float[batch,192,48,48] getitem_90
      float[batch,768,48,48] getitem_93
      float[batch,192,48,48] getitem_96
      float[batch,192,48,48] getitem_99
      float[batch,3,384,384] image_chw
      float[batch,384,384,3] image_f32
      float[batch,384,384,3] image_shifted
      float[batch,1] linalg_vector_norm
      float[1,batch,3072] linear
      float[145,batch,3072] linear_1
      float[145,batch,3072] linear_2
      float[batch,768] linear_3
      float[1,batch,3072] mean
      float[1,batch,3072] node_scaled_dot_product_attention_q_row
      float[144,batch,3072] permute_1
      float[1,batch,48,64] permute_2
      float[batch,48,192,192] relu
      float[batch,48,192,192] relu_1
      float[batch,96,96,96] relu_10
      float[batch,768,24,24] relu_100
      float[batch,3072,12,12] relu_101
      float[batch,768,12,12] relu_102
      float[batch,768,12,12] relu_103
      float[batch,3072,12,12] relu_104
      float[batch,768,12,12] relu_105
      float[batch,768,12,12] relu_106
      float[batch,3072,12,12] relu_107
      float[batch,768,12,12] relu_108
      float[batch,768,12,12] relu_109
      float[batch,384,96,96] relu_11
      float[batch,3072,12,12] relu_110
      float[batch,768,12,12] relu_111
      float[batch,768,12,12] relu_112
      float[batch,3072,12,12] relu_113
      float[batch,768,12,12] relu_114
      float[batch,768,12,12] relu_115
      float[batch,3072,12,12] relu_116
      float[batch,768,12,12] relu_117
      float[batch,768,12,12] relu_118
      float[batch,3072,12,12] relu_119
      float[batch,96,96,96] relu_12
      float[batch,768,12,12] relu_120
      float[batch,768,12,12] relu_121
      float[batch,3072,12,12] relu_122
      float[batch,96,96,96] relu_13
      float[batch,384,96,96] relu_14
      float[batch,96,96,96] relu_15
      float[batch,96,96,96] relu_16
      float[batch,384,96,96] relu_17
      float[batch,96,96,96] relu_18
      float[batch,96,96,96] relu_19
      float[batch,96,192,192] relu_2
      float[batch,384,96,96] relu_20
      float[batch,192,96,96] relu_21
      float[batch,192,96,96] relu_22
      float[batch,768,48,48] relu_23
      float[batch,192,48,48] relu_24
      float[batch,192,48,48] relu_25
      float[batch,768,48,48] relu_26
      float[batch,192,48,48] relu_27
      float[batch,192,48,48] relu_28
      float[batch,768,48,48] relu_29
      float[batch,96,96,96] relu_3
      float[batch,192,48,48] relu_30
      float[batch,192,48,48] relu_31
      float[batch,768,48,48] relu_32
      float[batch,192,48,48] relu_33
      float[batch,192,48,48] relu_34
      float[batch,768,48,48] relu_35
      float[batch,192,48,48] relu_36
      float[batch,192,48,48] relu_37
      float[batch,768,48,48] relu_38
      float[batch,192,48,48] relu_39
      float[batch,96,96,96] relu_4
      float[batch,192,48,48] relu_40
      float[batch,768,48,48] relu_41
      float[batch,192,48,48] relu_42
      float[batch,192,48,48] relu_43
      float[batch,768,48,48] relu_44
      float[batch,384,48,48] relu_45
      float[batch,384,48,48] relu_46
      float[batch,1536,24,24] relu_47
      float[batch,384,24,24] relu_48
      float[batch,384,24,24] relu_49
      float[batch,384,96,96] relu_5
      float[batch,1536,24,24] relu_50
      float[batch,384,24,24] relu_51
      float[batch,384,24,24] relu_52
      float[batch,1536,24,24] relu_53
      float[batch,384,24,24] relu_54
      float[batch,384,24,24] relu_55
      float[batch,1536,24,24] relu_56
      float[batch,384,24,24] relu_57
      float[batch,384,24,24] relu_58
      float[batch,1536,24,24] relu_59
      float[batch,96,96,96] relu_6
      float[batch,384,24,24] relu_60
      float[batch,384,24,24] relu_61
      float[batch,1536,24,24] relu_62
      float[batch,384,24,24] relu_63
      float[batch,384,24,24] relu_64
      float[batch,1536,24,24] relu_65
      float[batch,384,24,24] relu_66
      float[batch,384,24,24] relu_67
      float[batch,1536,24,24] relu_68
      float[batch,384,24,24] relu_69
      float[batch,96,96,96] relu_7
      float[batch,384,24,24] relu_70
      float[batch,1536,24,24] relu_71
      float[batch,384,24,24] relu_72
      float[batch,384,24,24] relu_73
      float[batch,1536,24,24] relu_74
      float[batch,384,24,24] relu_75
      float[batch,384,24,24] relu_76
      float[batch,1536,24,24] relu_77
      float[batch,384,24,24] relu_78
      float[batch,384,24,24] relu_79
      float[batch,384,96,96] relu_8
      float[batch,1536,24,24] relu_80
      float[batch,384,24,24] relu_81
      float[batch,384,24,24] relu_82
      float[batch,1536,24,24] relu_83
      float[batch,384,24,24] relu_84
      float[batch,384,24,24] relu_85
      float[batch,1536,24,24] relu_86
      float[batch,384,24,24] relu_87
      float[batch,384,24,24] relu_88
      float[batch,1536,24,24] relu_89
      float[batch,96,96,96] relu_9
      float[batch,384,24,24] relu_90
      float[batch,384,24,24] relu_91
      float[batch,1536,24,24] relu_92
      float[batch,384,24,24] relu_93
      float[batch,384,24,24] relu_94
      float[batch,1536,24,24] relu_95
      float[batch,384,24,24] relu_96
      float[batch,384,24,24] relu_97
      float[batch,1536,24,24] relu_98
      float[batch,768,24,24] relu_99
      float[batch,48,1,64] scaled_dot_product_attention
      float[batch,768] select
      float[3072] split_split_0
      float[3072] split_split_1
      float[3072] split_split_2
      float[unk__1,1,64] transpose
      float[unk__1,145,64] transpose_1
      float[unk__1,145,64] transpose_2
      float[145,1,3072] unsqueeze
      float[1,batch,3072] val_7
      float[145,batch,3072] val_8
      float[145,batch,3072] val_9
      float[1,batch,768] view_10
      float[batch,3072,144] view_2
      float[1,unk__1,64] view_3
      float[145,unk__1,64] view_4
      float[145,unk__1,64] view_5
      float[batch,48,1,64] view_6
      float[batch,48,145,64] view_7
      float[batch,48,145,64] view_8
      float[batch,3072] view_9
   >
{
   [pre_cast] image_f32 = Cast <to: int = 1> (image)
   [pre_shift] image_shifted = Sub (image_f32, image_shift)
   [pre_nhwc_to_nchw] image_chw = Transpose <perm: ints = [0, 3, 1, 2]> (image_shifted)
   getitem = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [2, 2]> (image_chw, "visual.conv1.weight", "visual.conv1.weight_bias")
   [node_relu] relu = Relu (getitem)
   getitem_3 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu, "visual.conv2.weight", "visual.conv2.weight_bias")
   relu_1 = Relu (getitem_3)
   getitem_6 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_1, "visual.conv3.weight", "visual.conv3.weight_bias")
   relu_2 = Relu (getitem_6)
   [node_avg_pool2d] avg_pool2d = AveragePool <auto_pad: string = "NOTSET", ceil_mode: int = 0, count_include_pad: int = 1, kernel_shape: ints = [2, 2], pads: ints = [0, 0, 0, 0], strides: ints = [2, 2]> (relu_2)
   getitem_9 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (avg_pool2d, "visual.layer1.0.conv1.weight", "visual.layer1.0.conv1.weight_bias")
   relu_3 = Relu (getitem_9)
   getitem_12 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_3, "visual.layer1.0.conv2.weight", "visual.layer1.0.conv2.weight_bias")
   relu_4 = Relu (getitem_12)
   getitem_15 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_4, "visual.layer1.0.conv3.weight", "visual.layer1.0.conv3.weight_bias")
   getitem_18 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (avg_pool2d, "visual.layer1.0.downsample.0.weight", "visual.layer1.0.downsample.0.weight_bias")
   add_140 = Add (getitem_15, getitem_18)
   relu_5 = Relu (add_140)
   getitem_21 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_5, "visual.layer1.1.conv1.weight", "visual.layer1.1.conv1.weight_bias")
   relu_6 = Relu (getitem_21)
   getitem_24 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_6, "visual.layer1.1.conv2.weight", "visual.layer1.1.conv2.weight_bias")
   relu_7 = Relu (getitem_24)
   getitem_27 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_7, "visual.layer1.1.conv3.weight", "visual.layer1.1.conv3.weight_bias")
   add_216 = Add (getitem_27, relu_5)
   relu_8 = Relu (add_216)
   getitem_30 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_8, "visual.layer1.2.conv1.weight", "visual.layer1.2.conv1.weight_bias")
   relu_9 = Relu (getitem_30)
   getitem_33 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_9, "visual.layer1.2.conv2.weight", "visual.layer1.2.conv2.weight_bias")
   relu_10 = Relu (getitem_33)
   getitem_36 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_10, "visual.layer1.2.conv3.weight", "visual.layer1.2.conv3.weight_bias")
   add_292 = Add (getitem_36, relu_8)
   relu_11 = Relu (add_292)
   getitem_39 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_11, "visual.layer1.3.conv1.weight", "visual.layer1.3.conv1.weight_bias")
   relu_12 = Relu (getitem_39)
   getitem_42 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_12, "visual.layer1.3.conv2.weight", "visual.layer1.3.conv2.weight_bias")
   relu_13 = Relu (getitem_42)
   getitem_45 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_13, "visual.layer1.3.conv3.weight", "visual.layer1.3.conv3.weight_bias")
   add_368 = Add (getitem_45, relu_11)
   relu_14 = Relu (add_368)
   getitem_48 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_14, "visual.layer1.4.conv1.weight", "visual.layer1.4.conv1.weight_bias")
   relu_15 = Relu (getitem_48)
   getitem_51 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_15, "visual.layer1.4.conv2.weight", "visual.layer1.4.conv2.weight_bias")
   relu_16 = Relu (getitem_51)
   getitem_54 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_16, "visual.layer1.4.conv3.weight", "visual.layer1.4.conv3.weight_bias")
   add_444 = Add (getitem_54, relu_14)
   relu_17 = Relu (add_444)
   getitem_57 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_17, "visual.layer1.5.conv1.weight", "visual.layer1.5.conv1.weight_bias")
   relu_18 = Relu (getitem_57)
   getitem_60 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_18, "visual.layer1.5.conv2.weight", "visual.layer1.5.conv2.weight_bias")
   relu_19 = Relu (getitem_60)
   getitem_63 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_19, "visual.layer1.5.conv3.weight", "visual.layer1.5.conv3.weight_bias")
   add_520 = Add (getitem_63, relu_17)
   relu_20 = Relu (add_520)
   getitem_66 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_20, "visual.layer2.0.conv1.weight", "visual.layer2.0.conv1.weight_bias")
   relu_21 = Relu (getitem_66)
   getitem_69 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_21, "visual.layer2.0.conv2.weight", "visual.layer2.0.conv2.weight_bias")
   relu_22 = Relu (getitem_69)
   avg_pool2d_2 = AveragePool <auto_pad: string = "NOTSET", ceil_mode: int = 0, count_include_pad: int = 1, kernel_shape: ints = [2, 2], pads: ints = [0, 0, 0, 0], strides: ints = [2, 2]> (relu_22)
   getitem_72 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (avg_pool2d_2, "visual.layer2.0.conv3.weight", "visual.layer2.0.conv3.weight_bias")
   avg_pool2d_3 = AveragePool <auto_pad: string = "NOTSET", ceil_mode: int = 0, count_include_pad: int = 1, kernel_shape: ints = [2, 2], pads: ints = [0, 0, 0, 0], strides: ints = [2, 2]> (relu_20)
   getitem_75 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (avg_pool2d_3, "visual.layer2.0.downsample.0.weight", "visual.layer2.0.downsample.0.weight_bias")
   add_616 = Add (getitem_72, getitem_75)
   relu_23 = Relu (add_616)
   getitem_78 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_23, "visual.layer2.1.conv1.weight", "visual.layer2.1.conv1.weight_bias")
   relu_24 = Relu (getitem_78)
   getitem_81 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_24, "visual.layer2.1.conv2.weight", "visual.layer2.1.conv2.weight_bias")
   relu_25 = Relu (getitem_81)
   getitem_84 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_25, "visual.layer2.1.conv3.weight", "visual.layer2.1.conv3.weight_bias")
   add_692 = Add (getitem_84, relu_23)
   relu_26 = Relu (add_692)
   getitem_87 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_26, "visual.layer2.2.conv1.weight", "visual.layer2.2.conv1.weight_bias")
   relu_27 = Relu (getitem_87)
   getitem_90 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_27, "visual.layer2.2.conv2.weight", "visual.layer2.2.conv2.weight_bias")
   relu_28 = Relu (getitem_90)
   getitem_93 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_28, "visual.layer2.2.conv3.weight", "visual.layer2.2.conv3.weight_bias")
   add_768 = Add (getitem_93, relu_26)
   relu_29 = Relu (add_768)
   getitem_96 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_29, "visual.layer2.3.conv1.weight", "visual.layer2.3.conv1.weight_bias")
   relu_30 = Relu (getitem_96)
   getitem_99 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_30, "visual.layer2.3.conv2.weight", "visual.layer2.3.conv2.weight_bias")
   relu_31 = Relu (getitem_99)
   getitem_102 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_31, "visual.layer2.3.conv3.weight", "visual.layer2.3.conv3.weight_bias")
   add_844 = Add (getitem_102, relu_29)
   relu_32 = Relu (add_844)
   getitem_105 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_32, "visual.layer2.4.conv1.weight", "visual.layer2.4.conv1.weight_bias")
   relu_33 = Relu (getitem_105)
   getitem_108 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_33, "visual.layer2.4.conv2.weight", "visual.layer2.4.conv2.weight_bias")
   relu_34 = Relu (getitem_108)
   getitem_111 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_34, "visual.layer2.4.conv3.weight", "visual.layer2.4.conv3.weight_bias")
   add_920 = Add (getitem_111, relu_32)
   relu_35 = Relu (add_920)
   getitem_114 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_35, "visual.layer2.5.conv1.weight", "visual.layer2.5.conv1.weight_bias")
   relu_36 = Relu (getitem_114)
   getitem_117 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_36, "visual.layer2.5.conv2.weight", "visual.layer2.5.conv2.weight_bias")
   relu_37 = Relu (getitem_117)
   getitem_120 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_37, "visual.layer2.5.conv3.weight", "visual.layer2.5.conv3.weight_bias")
   add_996 = Add (getitem_120, relu_35)
   relu_38 = Relu (add_996)
   getitem_123 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_38, "visual.layer2.6.conv1.weight", "visual.layer2.6.conv1.weight_bias")
   relu_39 = Relu (getitem_123)
   getitem_126 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_39, "visual.layer2.6.conv2.weight", "visual.layer2.6.conv2.weight_bias")
   relu_40 = Relu (getitem_126)
   getitem_129 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_40, "visual.layer2.6.conv3.weight", "visual.layer2.6.conv3.weight_bias")
   add_1072 = Add (getitem_129, relu_38)
   relu_41 = Relu (add_1072)
   getitem_132 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_41, "visual.layer2.7.conv1.weight", "visual.layer2.7.conv1.weight_bias")
   relu_42 = Relu (getitem_132)
   getitem_135 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_42, "visual.layer2.7.conv2.weight", "visual.layer2.7.conv2.weight_bias")
   relu_43 = Relu (getitem_135)
   getitem_138 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_43, "visual.layer2.7.conv3.weight", "visual.layer2.7.conv3.weight_bias")
   add_1148 = Add (getitem_138, relu_41)
   relu_44 = Relu (add_1148)
   getitem_141 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_44, "visual.layer3.0.conv1.weight", "visual.layer3.0.conv1.weight_bias")
   relu_45 = Relu (getitem_141)
   getitem_144 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_45, "visual.layer3.0.conv2.weight", "visual.layer3.0.conv2.weight_bias")
   relu_46 = Relu (getitem_144)
   avg_pool2d_4 = AveragePool <auto_pad: string = "NOTSET", ceil_mode: int = 0, count_include_pad: int = 1, kernel_shape: ints = [2, 2], pads: ints = [0, 0, 0, 0], strides: ints = [2, 2]> (relu_46)
   getitem_147 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (avg_pool2d_4, "visual.layer3.0.conv3.weight", "visual.layer3.0.conv3.weight_bias")
   avg_pool2d_5 = AveragePool <auto_pad: string = "NOTSET", ceil_mode: int = 0, count_include_pad: int = 1, kernel_shape: ints = [2, 2], pads: ints = [0, 0, 0, 0], strides: ints = [2, 2]> (relu_44)
   getitem_150 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (avg_pool2d_5, "visual.layer3.0.downsample.0.weight", "visual.layer3.0.downsample.0.weight_bias")
   add_1244 = Add (getitem_147, getitem_150)
   relu_47 = Relu (add_1244)
   getitem_153 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_47, "visual.layer3.1.conv1.weight", "visual.layer3.1.conv1.weight_bias")
   relu_48 = Relu (getitem_153)
   getitem_156 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_48, "visual.layer3.1.conv2.weight", "visual.layer3.1.conv2.weight_bias")
   relu_49 = Relu (getitem_156)
   getitem_159 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_49, "visual.layer3.1.conv3.weight", "visual.layer3.1.conv3.weight_bias")
   add_1320 = Add (getitem_159, relu_47)
   relu_50 = Relu (add_1320)
   getitem_162 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_50, "visual.layer3.2.conv1.weight", "visual.layer3.2.conv1.weight_bias")
   relu_51 = Relu (getitem_162)
   getitem_165 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_51, "visual.layer3.2.conv2.weight", "visual.layer3.2.conv2.weight_bias")
   relu_52 = Relu (getitem_165)
   getitem_168 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_52, "visual.layer3.2.conv3.weight", "visual.layer3.2.conv3.weight_bias")
   add_1396 = Add (getitem_168, relu_50)
   relu_53 = Relu (add_1396)
   getitem_171 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_53, "visual.layer3.3.conv1.weight", "visual.layer3.3.conv1.weight_bias")
   relu_54 = Relu (getitem_171)
   getitem_174 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_54, "visual.layer3.3.conv2.weight", "visual.layer3.3.conv2.weight_bias")
   relu_55 = Relu (getitem_174)
   getitem_177 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_55, "visual.layer3.3.conv3.weight", "visual.layer3.3.conv3.weight_bias")
   add_1472 = Add (getitem_177, relu_53)
   relu_56 = Relu (add_1472)
   getitem_180 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_56, "visual.layer3.4.conv1.weight", "visual.layer3.4.conv1.weight_bias")
   relu_57 = Relu (getitem_180)
   getitem_183 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_57, "visual.layer3.4.conv2.weight", "visual.layer3.4.conv2.weight_bias")
   relu_58 = Relu (getitem_183)
   getitem_186 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_58, "visual.layer3.4.conv3.weight", "visual.layer3.4.conv3.weight_bias")
   add_1548 = Add (getitem_186, relu_56)
   relu_59 = Relu (add_1548)
   getitem_189 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_59, "visual.layer3.5.conv1.weight", "visual.layer3.5.conv1.weight_bias")
   relu_60 = Relu (getitem_189)
   getitem_192 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_60, "visual.layer3.5.conv2.weight", "visual.layer3.5.conv2.weight_bias")
   relu_61 = Relu (getitem_192)
   getitem_195 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_61, "visual.layer3.5.conv3.weight", "visual.layer3.5.conv3.weight_bias")
   add_1624 = Add (getitem_195, relu_59)
   relu_62 = Relu (add_1624)
   getitem_198 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_62, "visual.layer3.6.conv1.weight", "visual.layer3.6.conv1.weight_bias")
   relu_63 = Relu (getitem_198)
   getitem_201 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_63, "visual.layer3.6.conv2.weight", "visual.layer3.6.conv2.weight_bias")
   relu_64 = Relu (getitem_201)
   getitem_204 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_64, "visual.layer3.6.conv3.weight", "visual.layer3.6.conv3.weight_bias")
   add_1700 = Add (getitem_204, relu_62)
   relu_65 = Relu (add_1700)
   getitem_207 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_65, "visual.layer3.7.conv1.weight", "visual.layer3.7.conv1.weight_bias")
   relu_66 = Relu (getitem_207)
   getitem_210 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_66, "visual.layer3.7.conv2.weight", "visual.layer3.7.conv2.weight_bias")
   relu_67 = Relu (getitem_210)
   getitem_213 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_67, "visual.layer3.7.conv3.weight", "visual.layer3.7.conv3.weight_bias")
   add_1776 = Add (getitem_213, relu_65)
   relu_68 = Relu (add_1776)
   getitem_216 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_68, "visual.layer3.8.conv1.weight", "visual.layer3.8.conv1.weight_bias")
   relu_69 = Relu (getitem_216)
   getitem_219 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_69, "visual.layer3.8.conv2.weight", "visual.layer3.8.conv2.weight_bias")
   relu_70 = Relu (getitem_219)
   getitem_222 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_70, "visual.layer3.8.conv3.weight", "visual.layer3.8.conv3.weight_bias")
   add_1852 = Add (getitem_222, relu_68)
   relu_71 = Relu (add_1852)
   getitem_225 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_71, "visual.layer3.9.conv1.weight", "visual.layer3.9.conv1.weight_bias")
   relu_72 = Relu (getitem_225)
   getitem_228 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_72, "visual.layer3.9.conv2.weight", "visual.layer3.9.conv2.weight_bias")
   relu_73 = Relu (getitem_228)
   getitem_231 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_73, "visual.layer3.9.conv3.weight", "visual.layer3.9.conv3.weight_bias")
   add_1928 = Add (getitem_231, relu_71)
   relu_74 = Relu (add_1928)
   getitem_234 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_74, "visual.layer3.10.conv1.weight", "visual.layer3.10.conv1.weight_bias")
   relu_75 = Relu (getitem_234)
   getitem_237 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_75, "visual.layer3.10.conv2.weight", "visual.layer3.10.conv2.weight_bias")
   relu_76 = Relu (getitem_237)
   getitem_240 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_76, "visual.layer3.10.conv3.weight", "visual.layer3.10.conv3.weight_bias")
   add_2004 = Add (getitem_240, relu_74)
   relu_77 = Relu (add_2004)
   getitem_243 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_77, "visual.layer3.11.conv1.weight", "visual.layer3.11.conv1.weight_bias")
   relu_78 = Relu (getitem_243)
   getitem_246 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_78, "visual.layer3.11.conv2.weight", "visual.layer3.11.conv2.weight_bias")
   relu_79 = Relu (getitem_246)
   getitem_249 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_79, "visual.layer3.11.conv3.weight", "visual.layer3.11.conv3.weight_bias")
   add_2080 = Add (getitem_249, relu_77)
   relu_80 = Relu (add_2080)
   getitem_252 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_80, "visual.layer3.12.conv1.weight", "visual.layer3.12.conv1.weight_bias")
   relu_81 = Relu (getitem_252)
   getitem_255 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_81, "visual.layer3.12.conv2.weight", "visual.layer3.12.conv2.weight_bias")
   relu_82 = Relu (getitem_255)
   getitem_258 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_82, "visual.layer3.12.conv3.weight", "visual.layer3.12.conv3.weight_bias")
   add_2156 = Add (getitem_258, relu_80)
   relu_83 = Relu (add_2156)
   getitem_261 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_83, "visual.layer3.13.conv1.weight", "visual.layer3.13.conv1.weight_bias")
   relu_84 = Relu (getitem_261)
   getitem_264 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_84, "visual.layer3.13.conv2.weight", "visual.layer3.13.conv2.weight_bias")
   relu_85 = Relu (getitem_264)
   getitem_267 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_85, "visual.layer3.13.conv3.weight", "visual.layer3.13.conv3.weight_bias")
   add_2232 = Add (getitem_267, relu_83)
   relu_86 = Relu (add_2232)
   getitem_270 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_86, "visual.layer3.14.conv1.weight", "visual.layer3.14.conv1.weight_bias")
   relu_87 = Relu (getitem_270)
   getitem_273 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_87, "visual.layer3.14.conv2.weight", "visual.layer3.14.conv2.weight_bias")
   relu_88 = Relu (getitem_273)
   getitem_276 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_88, "visual.layer3.14.conv3.weight", "visual.layer3.14.conv3.weight_bias")
   add_2308 = Add (getitem_276, relu_86)
   relu_89 = Relu (add_2308)
   getitem_279 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_89, "visual.layer3.15.conv1.weight", "visual.layer3.15.conv1.weight_bias")
   relu_90 = Relu (getitem_279)
   getitem_282 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_90, "visual.layer3.15.conv2.weight", "visual.layer3.15.conv2.weight_bias")
   relu_91 = Relu (getitem_282)
   getitem_285 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_91, "visual.layer3.15.conv3.weight", "visual.layer3.15.conv3.weight_bias")
   add_2384 = Add (getitem_285, relu_89)
   relu_92 = Relu (add_2384)
   getitem_288 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_92, "visual.layer3.16.conv1.weight", "visual.layer3.16.conv1.weight_bias")
   relu_93 = Relu (getitem_288)
   getitem_291 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_93, "visual.layer3.16.conv2.weight", "visual.layer3.16.conv2.weight_bias")
   relu_94 = Relu (getitem_291)
   getitem_294 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_94, "visual.layer3.16.conv3.weight", "visual.layer3.16.conv3.weight_bias")
   add_2460 = Add (getitem_294, relu_92)
   relu_95 = Relu (add_2460)
   getitem_297 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_95, "visual.layer3.17.conv1.weight", "visual.layer3.17.conv1.weight_bias")
   relu_96 = Relu (getitem_297)
   getitem_300 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_96, "visual.layer3.17.conv2.weight", "visual.layer3.17.conv2.weight_bias")
   relu_97 = Relu (getitem_300)
   getitem_303 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_97, "visual.layer3.17.conv3.weight", "visual.layer3.17.conv3.weight_bias")
   add_2536 = Add (getitem_303, relu_95)
   relu_98 = Relu (add_2536)
   getitem_306 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_98, "visual.layer4.0.conv1.weight", "visual.layer4.0.conv1.weight_bias")
   relu_99 = Relu (getitem_306)
   getitem_309 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_99, "visual.layer4.0.conv2.weight", "visual.layer4.0.conv2.weight_bias")
   relu_100 = Relu (getitem_309)
   avg_pool2d_6 = AveragePool <auto_pad: string = "NOTSET", ceil_mode: int = 0, count_include_pad: int = 1, kernel_shape: ints = [2, 2], pads: ints = [0, 0, 0, 0], strides: ints = [2, 2]> (relu_100)
   getitem_312 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (avg_pool2d_6, "visual.layer4.0.conv3.weight", "visual.layer4.0.conv3.weight_bias")
   avg_pool2d_7 = AveragePool <auto_pad: string = "NOTSET", ceil_mode: int = 0, count_include_pad: int = 1, kernel_shape: ints = [2, 2], pads: ints = [0, 0, 0, 0], strides: ints = [2, 2]> (relu_98)
   getitem_315 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (avg_pool2d_7, "visual.layer4.0.downsample.0.weight", "visual.layer4.0.downsample.0.weight_bias")
   add_2632 = Add (getitem_312, getitem_315)
   relu_101 = Relu (add_2632)
   getitem_318 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_101, "visual.layer4.1.conv1.weight", "visual.layer4.1.conv1.weight_bias")
   relu_102 = Relu (getitem_318)
   getitem_321 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_102, "visual.layer4.1.conv2.weight", "visual.layer4.1.conv2.weight_bias")
   relu_103 = Relu (getitem_321)
   getitem_324 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_103, "visual.layer4.1.conv3.weight", "visual.layer4.1.conv3.weight_bias")
   add_2708 = Add (getitem_324, relu_101)
   relu_104 = Relu (add_2708)
   getitem_327 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_104, "visual.layer4.2.conv1.weight", "visual.layer4.2.conv1.weight_bias")
   relu_105 = Relu (getitem_327)
   getitem_330 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_105, "visual.layer4.2.conv2.weight", "visual.layer4.2.conv2.weight_bias")
   relu_106 = Relu (getitem_330)
   getitem_333 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_106, "visual.layer4.2.conv3.weight", "visual.layer4.2.conv3.weight_bias")
   add_2784 = Add (getitem_333, relu_104)
   relu_107 = Relu (add_2784)
   getitem_336 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_107, "visual.layer4.3.conv1.weight", "visual.layer4.3.conv1.weight_bias")
   relu_108 = Relu (getitem_336)
   getitem_339 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_108, "visual.layer4.3.conv2.weight", "visual.layer4.3.conv2.weight_bias")
   relu_109 = Relu (getitem_339)
   getitem_342 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_109, "visual.layer4.3.conv3.weight", "visual.layer4.3.conv3.weight_bias")
   add_2860 = Add (getitem_342, relu_107)
   relu_110 = Relu (add_2860)
   getitem_345 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_110, "visual.layer4.4.conv1.weight", "visual.layer4.4.conv1.weight_bias")
   relu_111 = Relu (getitem_345)
   getitem_348 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_111, "visual.layer4.4.conv2.weight", "visual.layer4.4.conv2.weight_bias")
   relu_112 = Relu (getitem_348)
   getitem_351 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_112, "visual.layer4.4.conv3.weight", "visual.layer4.4.conv3.weight_bias")
   add_2936 = Add (getitem_351, relu_110)
   relu_113 = Relu (add_2936)
   getitem_354 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_113, "visual.layer4.5.conv1.weight", "visual.layer4.5.conv1.weight_bias")
   relu_114 = Relu (getitem_354)
   getitem_357 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_114, "visual.layer4.5.conv2.weight", "visual.layer4.5.conv2.weight_bias")
   relu_115 = Relu (getitem_357)
   getitem_360 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_115, "visual.layer4.5.conv3.weight", "visual.layer4.5.conv3.weight_bias")
   add_3012 = Add (getitem_360, relu_113)
   relu_116 = Relu (add_3012)
   getitem_363 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_116, "visual.layer4.6.conv1.weight", "visual.layer4.6.conv1.weight_bias")
   relu_117 = Relu (getitem_363)
   getitem_366 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_117, "visual.layer4.6.conv2.weight", "visual.layer4.6.conv2.weight_bias")
   relu_118 = Relu (getitem_366)
   getitem_369 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_118, "visual.layer4.6.conv3.weight", "visual.layer4.6.conv3.weight_bias")
   add_3088 = Add (getitem_369, relu_116)
   relu_119 = Relu (add_3088)
   getitem_372 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_119, "visual.layer4.7.conv1.weight", "visual.layer4.7.conv1.weight_bias")
   relu_120 = Relu (getitem_372)
   getitem_375 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (relu_120, "visual.layer4.7.conv2.weight", "visual.layer4.7.conv2.weight_bias")
   relu_121 = Relu (getitem_375)
   getitem_378 = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> (relu_121, "visual.layer4.7.conv3.weight", "visual.layer4.7.conv3.weight_bias")
   add_3164 = Add (getitem_378, relu_119)
   relu_122 = Relu (add_3164)
   view_2 = Reshape <allowzero: int = 1> (relu_122, view_2_target)
   permute_1 = Transpose <perm: ints = [2, 0, 1]> (view_2)
   [node_mean] mean = ReduceMean <keepdims: int = 1, noop_with_empty_axes: int = 0> (permute_1, val_0)
   [node_cat] cat = Concat <axis: int = 0> (mean, permute_1)
   [node_unsqueeze] unsqueeze = Unsqueeze ("visual.attnpool.positional_embedding", val_3)
   add_3194 = Add (cat, unsqueeze)
   split_split_0, split_split_1, split_split_2 = Split <axis: int = 0, num_outputs: int = 3> (cat_1)
   node_scaled_dot_product_attention_q_row = Slice (add_3194, val_0, val_3, val_0)
   val_7 = MatMul (node_scaled_dot_product_attention_q_row, val_4)
   [node_linear] linear = Add (val_7, split_split_0)
   val_8 = MatMul (add_3194, val_5)
   linear_1 = Add (val_8, split_split_1)
   val_9 = MatMul (add_3194, val_6)
   linear_2 = Add (val_9, split_split_2)
   view_3 = Reshape <allowzero: int = 1> (linear, node_scaled_dot_product_attention_q_unpack_1)
   [node_transpose] transpose = Transpose <perm: ints = [1, 0, 2]> (view_3)
   view_4 = Reshape <allowzero: int = 1> (linear_1, view_4_target)
   transpose_1 = Transpose <perm: ints = [1, 0, 2]> (view_4)
   view_5 = Reshape <allowzero: int = 1> (linear_2, view_4_target)
   transpose_2 = Transpose <perm: ints = [1, 0, 2]> (view_5)
   view_6 = Reshape <allowzero: int = 1> (transpose, node_scaled_dot_product_attention_q_pack_1)
   view_7 = Reshape <allowzero: int = 1> (transpose_1, view_7_target)
   view_8 = Reshape <allowzero: int = 1> (transpose_2, view_7_target)
   [node_scaled_dot_product_attention] scaled_dot_product_attention = Attention <is_causal: int = 0, qk_matmul_output_mode: int = 0, softcap: float = 0> (view_6, view_7, view_8)
   permute_2 = Transpose <perm: ints = [2, 0, 1, 3]> (scaled_dot_product_attention)
   view_9 = Reshape <allowzero: int = 1> (permute_2, view_9_target)
   linear_3 = Gemm <alpha: float = 1, beta: float = 1, transA: int = 0, transB: int = 1> (view_9, "visual.attnpool.c_proj.weight", "visual.attnpool.c_proj.bias")
   view_10 = Reshape <allowzero: int = 1> (linear_3, node_scaled_dot_product_attention_out_1)
   select = Squeeze (view_10, val_0)
   [node_linalg_vector_norm] linalg_vector_norm = ReduceL2 <keepdims: int = 1, noop_with_empty_axes: int = 0> (select, val_1)
   [node_clamp_min] clamp_min = Clip (linalg_vector_norm, val_2)
   [node_div] image_embedding = Div (select, clamp_min)
}

weights:
cat_1 FLOAT[9216] 5f298c3e502a
image_shift FLOAT[3] 2f7a50e604ad
node_scaled_dot_product_attention_out_1 INT64[3] fa8d086288f9
node_scaled_dot_product_attention_q_pack_1 INT64[4] 2564bf8beb9e
node_scaled_dot_product_attention_q_unpack_1 INT64[3] cfe34a386daf
val_0 INT64[1] af5570f5a181
val_1 INT64[1] 12a3ae445661
val_2 FLOAT[] 6708d9be4956
val_3 INT64[1] 7c9fa136d441
val_4 FLOAT[3072,3072] e41d511859c1
val_5 FLOAT[3072,3072] f5d2fce4950e
val_6 FLOAT[3072,3072] dd21d99158e4
view_2_target INT64[3] 78f76ad494cc
view_4_target INT64[3] fb31726bf5aa
view_7_target INT64[4] cb83b36b0f4b
view_9_target INT64[2] 7a66e878ce45
visual.attnpool.c_proj.bias FLOAT[768] 11751e48178a
visual.attnpool.c_proj.weight FLOAT[768,3072] e84b5146e971
visual.attnpool.positional_embedding FLOAT[145,3072] 391340b02a4c
visual.conv1.weight FLOAT[48,3,3,3] 94baac763782
visual.conv1.weight_bias FLOAT[48] f16e2f8ab2b9
visual.conv2.weight FLOAT[48,48,3,3] 7b12a6a125bf
visual.conv2.weight_bias FLOAT[48] 7603d13d1e7d
visual.conv3.weight FLOAT[96,48,3,3] eb5f6ccb217c
visual.conv3.weight_bias FLOAT[96] fa9e4ae719cd
visual.layer1.0.conv1.weight FLOAT[96,96,1,1] 5c51f34176d3
visual.layer1.0.conv1.weight_bias FLOAT[96] ef07c9705365
visual.layer1.0.conv2.weight FLOAT[96,96,3,3] e217e4607643
visual.layer1.0.conv2.weight_bias FLOAT[96] 929988a1764a
visual.layer1.0.conv3.weight FLOAT[384,96,1,1] f548690c165e
visual.layer1.0.conv3.weight_bias FLOAT[384] c43fa7cd692d
visual.layer1.0.downsample.0.weight FLOAT[384,96,1,1] c27abd7f140a
visual.layer1.0.downsample.0.weight_bias FLOAT[384] 2e60f8fd53ec
visual.layer1.1.conv1.weight FLOAT[96,384,1,1] f2faf43f174e
visual.layer1.1.conv1.weight_bias FLOAT[96] 42e8cbdef849
visual.layer1.1.conv2.weight FLOAT[96,96,3,3] 45a2ac5680af
visual.layer1.1.conv2.weight_bias FLOAT[96] 3fa6c2e74a63
visual.layer1.1.conv3.weight FLOAT[384,96,1,1] 90977330f018
visual.layer1.1.conv3.weight_bias FLOAT[384] 33c6ab7fa125
visual.layer1.2.conv1.weight FLOAT[96,384,1,1] b236d91bc6ec
visual.layer1.2.conv1.weight_bias FLOAT[96] 79d91f16952f
visual.layer1.2.conv2.weight FLOAT[96,96,3,3] 63dfac313c4d
visual.layer1.2.conv2.weight_bias FLOAT[96] 764914249890
visual.layer1.2.conv3.weight FLOAT[384,96,1,1] 0467713e7ff9
visual.layer1.2.conv3.weight_bias FLOAT[384] 27f3acd3c15c
visual.layer1.3.conv1.weight FLOAT[96,384,1,1] db0b6e3e7edc
visual.layer1.3.conv1.weight_bias FLOAT[96] bc5354f9a1ed
visual.layer1.3.conv2.weight FLOAT[96,96,3,3] a853e59b9dd9
visual.layer1.3.conv2.weight_bias FLOAT[96] 6c27e3a329fe
visual.layer1.3.conv3.weight FLOAT[384,96,1,1] b6823e3f9fc8
visual.layer1.3.conv3.weight_bias FLOAT[384] 896457ceca49
visual.layer1.4.conv1.weight FLOAT[96,384,1,1] 3462530111cd
visual.layer1.4.conv1.weight_bias FLOAT[96] f7343658a869
visual.layer1.4.conv2.weight FLOAT[96,96,3,3] 7fb27742743b
visual.layer1.4.conv2.weight_bias FLOAT[96] 34404d5cc103
visual.layer1.4.conv3.weight FLOAT[384,96,1,1] ba4fae0e290e
visual.layer1.4.conv3.weight_bias FLOAT[384] a7b62859c32c
visual.layer1.5.conv1.weight FLOAT[96,384,1,1] 08ff71a767b8
visual.layer1.5.conv1.weight_bias FLOAT[96] 5cbe11322b0d
visual.layer1.5.conv2.weight FLOAT[96,96,3,3] 69987159265a
visual.layer1.5.conv2.weight_bias FLOAT[96] 05bcc6914bb8
visual.layer1.5.conv3.weight FLOAT[384,96,1,1] dc2d1c91c8ba
visual.layer1.5.conv3.weight_bias FLOAT[384] 56245f32f25e
visual.layer2.0.conv1.weight FLOAT[192,384,1,1] a510be647379
visual.layer2.0.conv1.weight_bias FLOAT[192] 009180dd7ee3
visual.layer2.0.conv2.weight FLOAT[192,192,3,3] 5911a3abc49a
visual.layer2.0.conv2.weight_bias FLOAT[192] 2cd5e8fbdae8
visual.layer2.0.conv3.weight FLOAT[768,192,1,1] ba61994eb098
visual.layer2.0.conv3.weight_bias FLOAT[768] f44059758d13
visual.layer2.0.downsample.0.weight FLOAT[768,384,1,1] 213de7d55207
visual.layer2.0.downsample.0.weight_bias FLOAT[768] 4df5cf325874
visual.layer2.1.conv1.weight FLOAT[192,768,1,1] 22ca5be53c1b
visual.layer2.1.conv1.weight_bias FLOAT[192] 2de713331def
visual.layer2.1.conv2.weight FLOAT[192,192,3,3] ae6353cb8963
visual.layer2.1.conv2.weight_bias FLOAT[192] b592f047261a
visual.layer2.1.conv3.weight FLOAT[768,192,1,1] 5d0a468b6620
visual.layer2.1.conv3.weight_bias FLOAT[768] bece70d27add
visual.layer2.2.conv1.weight FLOAT[192,768,1,1] 8ec57c67fdc4
visual.layer2.2.conv1.weight_bias FLOAT[192] bad771e598fd
visual.layer2.2.conv2.weight FLOAT[192,192,3,3] 862a9a1b4c8a
visual.layer2.2.conv2.weight_bias FLOAT[192] d2f871f39da6
visual.layer2.2.conv3.weight FLOAT[768,192,1,1] dc327b8983f9
visual.layer2.2.conv3.weight_bias FLOAT[768] 7d89ad3d7b12
visual.layer2.3.conv1.weight FLOAT[192,768,1,1] c01b158a566d
visual.layer2.3.conv1.weight_bias FLOAT[192] fa008f02e930
visual.layer2.3.conv2.weight FLOAT[192,192,3,3] b74a749292fa
visual.layer2.3.conv2.weight_bias FLOAT[192] 33d7a61fd2c5
visual.layer2.3.conv3.weight FLOAT[768,192,1,1] e6318f561497
visual.layer2.3.conv3.weight_bias FLOAT[768] 0be1e9fbcc17
visual.layer2.4.conv1.weight FLOAT[192,768,1,1] 31ebfeb6f2ec
visual.layer2.4.conv1.weight_bias FLOAT[192] e1c11c285387
visual.layer2.4.conv2.weight FLOAT[192,192,3,3] 6f9216ef7d7e
visual.layer2.4.conv2.weight_bias FLOAT[192] c75b0ecc3b62
visual.layer2.4.conv3.weight FLOAT[768,192,1,1] 6aaf2c9fa64e
visual.layer2.4.conv3.weight_bias FLOAT[768] 6cf00a2f5f23
visual.layer2.5.conv1.weight FLOAT[192,768,1,1] 1fb9f4d2fe94
visual.layer2.5.conv1.weight_bias FLOAT[192] a621e08512dd
visual.layer2.5.conv2.weight FLOAT[192,192,3,3] bde3e0b4ea83
visual.layer2.5.conv2.weight_bias FLOAT[192] 7905f91d9907
visual.layer2.5.conv3.weight FLOAT[768,192,1,1] d9771e37956a
visual.layer2.5.conv3.weight_bias FLOAT[768] bdbe0eba8714
visual.layer2.6.conv1.weight FLOAT[192,768,1,1] fcd41b467588
visual.layer2.6.conv1.weight_bias FLOAT[192] 86fd3640d3cb
visual.layer2.6.conv2.weight FLOAT[192,192,3,3] 79c3beee8fb2
visual.layer2.6.conv2.weight_bias FLOAT[192] b603fd76eb0b
visual.layer2.6.conv3.weight FLOAT[768,192,1,1] 83577e4084e3
visual.layer2.6.conv3.weight_bias FLOAT[768] c9a84d8eff5f
visual.layer2.7.conv1.weight FLOAT[192,768,1,1] a967445e4898
visual.layer2.7.conv1.weight_bias FLOAT[192] 89712714a40a
visual.layer2.7.conv2.weight FLOAT[192,192,3,3] 162011389981
visual.layer2.7.conv2.weight_bias FLOAT[192] cb8ac0f4008e
visual.layer2.7.conv3.weight FLOAT[768,192,1,1] d239c56f2194
visual.layer2.7.conv3.weight_bias FLOAT[768] 1ed2926ada17
visual.layer3.0.conv1.weight FLOAT[384,768,1,1] efbdc65e8341
visual.layer3.0.conv1.weight_bias FLOAT[384] 3d7a7d2f144e
visual.layer3.0.conv2.weight FLOAT[384,384,3,3] 7c499b01152f
visual.layer3.0.conv2.weight_bias FLOAT[384] 0b8792d39d97
visual.layer3.0.conv3.weight FLOAT[1536,384,1,1] 3aafe0b0c4ab
visual.layer3.0.conv3.weight_bias FLOAT[1536] 09c1ab66649e
visual.layer3.0.downsample.0.weight FLOAT[1536,768,1,1] d34c9de36609
visual.layer3.0.downsample.0.weight_bias FLOAT[1536] 36326f840377
visual.layer3.1.conv1.weight FLOAT[384,1536,1,1] 0f13b98faeb0
visual.layer3.1.conv1.weight_bias FLOAT[384] 903fd2c9cefe
visual.layer3.1.conv2.weight FLOAT[384,384,3,3] d30bc2a6ad6a
visual.layer3.1.conv2.weight_bias FLOAT[384] 78548969b53b
visual.layer3.1.conv3.weight FLOAT[1536,384,1,1] c5abd5931d65
visual.layer3.1.conv3.weight_bias FLOAT[1536] ff7f9bab5c25
visual.layer3.10.conv1.weight FLOAT[384,1536,1,1] 18ee9e95c2bb
visual.layer3.10.conv1.weight_bias FLOAT[384] d60538f8bb82
visual.layer3.10.conv2.weight FLOAT[384,384,3,3] 1ec781912725
visual.layer3.10.conv2.weight_bias FLOAT[384] 585b7e010cb5
visual.layer3.10.conv3.weight FLOAT[1536,384,1,1] 8f4b1d8ae33c
visual.layer3.10.conv3.weight_bias FLOAT[1536] 41ea52623c2e
visual.layer3.11.conv1.weight FLOAT[384,1536,1,1] db41912c1e4b
visual.layer3.11.conv1.weight_bias FLOAT[384] 7a7e8329fb90
visual.layer3.11.conv2.weight FLOAT[384,384,3,3] af327e2a8c11
visual.layer3.11.conv2.weight_bias FLOAT[384] 1c010e9a0cbf
visual.layer3.11.conv3.weight FLOAT[1536,384,1,1] 9b8ef8381cac
visual.layer3.11.conv3.weight_bias FLOAT[1536] 59a6d206491f
visual.layer3.12.conv1.weight FLOAT[384,1536,1,1] 55d172c59cc1
visual.layer3.12.conv1.weight_bias FLOAT[384] cac9ad220116
visual.layer3.12.conv2.weight FLOAT[384,384,3,3] e3d672380955
visual.layer3.12.conv2.weight_bias FLOAT[384] 9098b988bce5
visual.layer3.12.conv3.weight FLOAT[1536,384,1,1] fae8e7d159bd
visual.layer3.12.conv3.weight_bias FLOAT[1536] 94c7bb63d4b2
visual.layer3.13.conv1.weight FLOAT[384,1536,1,1] 4ec0699dc4c6
visual.layer3.13.conv1.weight_bias FLOAT[384] f148cc9349cd
visual.layer3.13.conv2.weight FLOAT[384,384,3,3] 9315a995d483
visual.layer3.13.conv2.weight_bias FLOAT[384] 2aa5c4dd432f
visual.layer3.13.conv3.weight FLOAT[1536,384,1,1] 3f0c7916cc9d
visual.layer3.13.conv3.weight_bias FLOAT[1536] 28c59af94d11
visual.layer3.14.conv1.weight FLOAT[384,1536,1,1] e49a4124b100
visual.layer3.14.conv1.weight_bias FLOAT[384] fc2508770eb6
visual.layer3.14.conv2.weight FLOAT[384,384,3,3] ad7e91370e9b
visual.layer3.14.conv2.weight_bias FLOAT[384] d1f4c0a83216
visual.layer3.14.conv3.weight FLOAT[1536,384,1,1] 5c01e3bcf39b
visual.layer3.14.conv3.weight_bias FLOAT[1536] aeaa0b16c3ab
visual.layer3.15.conv1.weight FLOAT[384,1536,1,1] 7affe7c62039
visual.layer3.15.conv1.weight_bias FLOAT[384] b991cb772c9e
visual.layer3.15.conv2.weight FLOAT[384,384,3,3] 70a774f3e9ab
visual.layer3.15.conv2.weight_bias FLOAT[384] 2bed00808e48
visual.layer3.15.conv3.weight FLOAT[1536,384,1,1] 60716c3a91ea
visual.layer3.15.conv3.weight_bias FLOAT[1536] 57f6761b7490
visual.layer3.16.conv1.weight FLOAT[384,1536,1,1] 95585caeda31
visual.layer3.16.conv1.weight_bias FLOAT[384] 156340c56819
visual.layer3.16.conv2.weight FLOAT[384,384,3,3] d436e1db88df
visual.layer3.16.conv2.weight_bias FLOAT[384] d0c57d2e5211
visual.layer3.16.conv3.weight FLOAT[1536,384,1,1] b5f42bbf0a56
visual.layer3.16.conv3.weight_bias FLOAT[1536] 7ddb9040e36c
visual.layer3.17.conv1.weight FLOAT[384,1536,1,1] 76d9160aa365
visual.layer3.17.conv1.weight_bias FLOAT[384] 8d6f7505ada2
visual.layer3.17.conv2.weight FLOAT[384,384,3,3] ff8d832e2a33
visual.layer3.17.conv2.weight_bias FLOAT[384] 6daeb51037d4
visual.layer3.17.conv3.weight FLOAT[1536,384,1,1] 0439097498d4
visual.layer3.17.conv3.weight_bias FLOAT[1536] c729abbe4b7e
visual.layer3.2.conv1.weight FLOAT[384,1536,1,1] 011598123945
visual.layer3.2.conv1.weight_bias FLOAT[384] e6d83d566c84
visual.layer3.2.conv2.weight FLOAT[384,384,3,3] 8b62c0ae0b0e
visual.layer3.2.conv2.weight_bias FLOAT[384] a4cbba71cd75
visual.layer3.2.conv3.weight FLOAT[1536,384,1,1] c2cf2ddd5d08
visual.layer3.2.conv3.weight_bias FLOAT[1536] 23b98d48c0ed
visual.layer3.3.conv1.weight FLOAT[384,1536,1,1] 8623b5e6a7d2
visual.layer3.3.conv1.weight_bias FLOAT[384] 4b6dc08ede3c
visual.layer3.3.conv2.weight FLOAT[384,384,3,3] df9ab2e3efd2
visual.layer3.3.conv2.weight_bias FLOAT[384] e70ba06acce5
visual.layer3.3.conv3.weight FLOAT[1536,384,1,1] d85e5b4e0516
visual.layer3.3.conv3.weight_bias FLOAT[1536] 75c56e738c21
visual.layer3.4.conv1.weight FLOAT[384,1536,1,1] dd5607aae16d
visual.layer3.4.conv1.weight_bias FLOAT[384] 4a49785e13ab
visual.layer3.4.conv2.weight FLOAT[384,384,3,3] e80c6572671d
visual.layer3.4.conv2.weight_bias FLOAT[384] 616150ac52e0
visual.layer3.4.conv3.weight FLOAT[1536,384,1,1] b671e0ea1a86
visual.layer3.4.conv3.weight_bias FLOAT[1536] 87bc065cd62c
visual.layer3.5.conv1.weight FLOAT[384,1536,1,1] b531d32a8e78
visual.layer3.5.conv1.weight_bias FLOAT[384] 3c69d354360e
visual.layer3.5.conv2.weight FLOAT[384,384,3,3] 44d1ba1ca38c
visual.layer3.5.conv2.weight_bias FLOAT[384] 5d514ba3ebf9
visual.layer3.5.conv3.weight FLOAT[1536,384,1,1] 260d32792c81
visual.layer3.5.conv3.weight_bias FLOAT[1536] ce0a56bcb1f6
visual.layer3.6.conv1.weight FLOAT[384,1536,1,1] a41e9b85dbef
visual.layer3.6.conv1.weight_bias FLOAT[384] 86a1c13da678
visual.layer3.6.conv2.weight FLOAT[384,384,3,3] 0c0aba7522a8
visual.layer3.6.conv2.weight_bias FLOAT[384] 36ef8af1b88a
visual.layer3.6.conv3.weight FLOAT[1536,384,1,1] cc4c083ec206
visual.layer3.6.conv3.weight_bias FLOAT[1536] 86356c93aad0
visual.layer3.7.conv1.weight FLOAT[384,1536,1,1] 12bacb1d1729
visual.layer3.7.conv1.weight_bias FLOAT[384] cefc320b7e5a
visual.layer3.7.conv2.weight FLOAT[384,384,3,3] 3b51d860fbb9
visual.layer3.7.conv2.weight_bias FLOAT[384] 3aeb23524f41
visual.layer3.7.conv3.weight FLOAT[1536,384,1,1] 71e1ee495b58
visual.layer3.7.conv3.weight_bias FLOAT[1536] b05c443d9638
visual.layer3.8.conv1.weight FLOAT[384,1536,1,1] 836eb37caca6
visual.layer3.8.conv1.weight_bias FLOAT[384] aafd2e4c10b4
visual.layer3.8.conv2.weight FLOAT[384,384,3,3] d4bc3cc3998c
visual.layer3.8.conv2.weight_bias FLOAT[384] c7de0e5c00f2
visual.layer3.8.conv3.weight FLOAT[1536,384,1,1] dd361fa421d8
visual.layer3.8.conv3.weight_bias FLOAT[1536] af5ed08e955d
visual.layer3.9.conv1.weight FLOAT[384,1536,1,1] 128e641e3c51
visual.layer3.9.conv1.weight_bias FLOAT[384] be19d26f1331
visual.layer3.9.conv2.weight FLOAT[384,384,3,3] 921712ebc6cf
visual.layer3.9.conv2.weight_bias FLOAT[384] 35974e689dae
visual.layer3.9.conv3.weight FLOAT[1536,384,1,1] 9583495a1b58
visual.layer3.9.conv3.weight_bias FLOAT[1536] bdd9f656d9d0
visual.layer4.0.conv1.weight FLOAT[768,1536,1,1] d13d6d47962c
visual.layer4.0.conv1.weight_bias FLOAT[768] 39cd48197351
visual.layer4.0.conv2.weight FLOAT[768,768,3,3] f2489e38e919
visual.layer4.0.conv2.weight_bias FLOAT[768] 043d8a89969a
visual.layer4.0.conv3.weight FLOAT[3072,768,1,1] 0ef32922752f
visual.layer4.0.conv3.weight_bias FLOAT[3072] 96d3f24e84e1
visual.layer4.0.downsample.0.weight FLOAT[3072,1536,1,1] 0e3e2c667c03
visual.layer4.0.downsample.0.weight_bias FLOAT[3072] b9bc52b12f0a
visual.layer4.1.conv1.weight FLOAT[768,3072,1,1] 847bf04cc598
visual.layer4.1.conv1.weight_bias FLOAT[768] 2f1d4bddbf80
visual.layer4.1.conv2.weight FLOAT[768,768,3,3] 1c5c3192472a
visual.layer4.1.conv2.weight_bias FLOAT[768] a01da007167e
visual.layer4.1.conv3.weight FLOAT[3072,768,1,1] 4e81d83926fd
visual.layer4.1.conv3.weight_bias FLOAT[3072] 1dc9339df7af
visual.layer4.2.conv1.weight FLOAT[768,3072,1,1] 0e6d87c69cd9
visual.layer4.2.conv1.weight_bias FLOAT[768] 539a1ecb3721
visual.layer4.2.conv2.weight FLOAT[768,768,3,3] e506cbf8a8e8
visual.layer4.2.conv2.weight_bias FLOAT[768] 535f938f25e2
visual.layer4.2.conv3.weight FLOAT[3072,768,1,1] 02cc5eeb69d7
visual.layer4.2.conv3.weight_bias FLOAT[3072] 7308050614bc
visual.layer4.3.conv1.weight FLOAT[768,3072,1,1] a391b121c89f
visual.layer4.3.conv1.weight_bias FLOAT[768] 132090d609cf
visual.layer4.3.conv2.weight FLOAT[768,768,3,3] 8358e75dbda4
visual.layer4.3.conv2.weight_bias FLOAT[768] e12d041b74ce
visual.layer4.3.conv3.weight FLOAT[3072,768,1,1] 062ef697b5fd
visual.layer4.3.conv3.weight_bias FLOAT[3072] 58bebde27743
visual.layer4.4.conv1.weight FLOAT[768,3072,1,1] b03c31db1ed5
visual.layer4.4.conv1.weight_bias FLOAT[768] c500c179a44f
visual.layer4.4.conv2.weight FLOAT[768,768,3,3] 698f09451e89
visual.layer4.4.conv2.weight_bias FLOAT[768] 3387d622da54
visual.layer4.4.conv3.weight FLOAT[3072,768,1,1] 1a3097e10ded
visual.layer4.4.conv3.weight_bias FLOAT[3072] 35c9ef927560
visual.layer4.5.conv1.weight FLOAT[768,3072,1,1] 8361d620d747
visual.layer4.5.conv1.weight_bias FLOAT[768] 46d3190e524b
visual.layer4.5.conv2.weight FLOAT[768,768,3,3] 183b29394732
visual.layer4.5.conv2.weight_bias FLOAT[768] 17296ccf0ea0
visual.layer4.5.conv3.weight FLOAT[3072,768,1,1] fa572600c03c
visual.layer4.5.conv3.weight_bias FLOAT[3072] 1afbf729c53b
visual.layer4.6.conv1.weight FLOAT[768,3072,1,1] bb863ba982ed
visual.layer4.6.conv1.weight_bias FLOAT[768] 333bb7a987b5
visual.layer4.6.conv2.weight FLOAT[768,768,3,3] 017b42b5b5a8
visual.layer4.6.conv2.weight_bias FLOAT[768] 98d7febe37df
visual.layer4.6.conv3.weight FLOAT[3072,768,1,1] 290e30caef3c
visual.layer4.6.conv3.weight_bias FLOAT[3072] efca3c1d6966
visual.layer4.7.conv1.weight FLOAT[768,3072,1,1] 5565dd2003f3
visual.layer4.7.conv1.weight_bias FLOAT[768] f81985899ba1
visual.layer4.7.conv2.weight FLOAT[768,768,3,3] 6e4b6e97f8f0
visual.layer4.7.conv2.weight_bias FLOAT[768] 6a2c3b51a87e
visual.layer4.7.conv3.weight FLOAT[3072,768,1,1] 9a1dad3c5d68
visual.layer4.7.conv3.weight_bias FLOAT[3072] 82daa5e93709
