본문 바로가기

AI & Big Data/AI

ReXNet V1 summary

반응형
----------------------------------------------------------------
        Layer (type)               Output Shape         Param #
================================================================
            Conv2d-1         [-1, 32, 112, 112]             864
       BatchNorm2d-2         [-1, 32, 112, 112]              64
           Sigmoid-3         [-1, 32, 112, 112]               0
             Swish-4         [-1, 32, 112, 112]               0
            Conv2d-5         [-1, 32, 112, 112]             288
       BatchNorm2d-6         [-1, 32, 112, 112]              64
             ReLU6-7         [-1, 32, 112, 112]               0
            Conv2d-8         [-1, 16, 112, 112]             512
       BatchNorm2d-9         [-1, 16, 112, 112]              32
 LinearBottleneck-10         [-1, 16, 112, 112]               0
           Conv2d-11         [-1, 96, 112, 112]           1,536
      BatchNorm2d-12         [-1, 96, 112, 112]             192
          Sigmoid-13         [-1, 96, 112, 112]               0
            Swish-14         [-1, 96, 112, 112]               0
           Conv2d-15           [-1, 96, 56, 56]             864
      BatchNorm2d-16           [-1, 96, 56, 56]             192
            ReLU6-17           [-1, 96, 56, 56]               0
           Conv2d-18           [-1, 24, 56, 56]           2,304
      BatchNorm2d-19           [-1, 24, 56, 56]              48
 LinearBottleneck-20           [-1, 24, 56, 56]               0
           Conv2d-21          [-1, 144, 56, 56]           3,456
      BatchNorm2d-22          [-1, 144, 56, 56]             288
          Sigmoid-23          [-1, 144, 56, 56]               0
            Swish-24          [-1, 144, 56, 56]               0
           Conv2d-25          [-1, 144, 56, 56]           1,296
      BatchNorm2d-26          [-1, 144, 56, 56]             288
            ReLU6-27          [-1, 144, 56, 56]               0
           Conv2d-28           [-1, 32, 56, 56]           4,608
      BatchNorm2d-29           [-1, 32, 56, 56]              64
 LinearBottleneck-30           [-1, 32, 56, 56]               0
           Conv2d-31          [-1, 192, 56, 56]           6,144
      BatchNorm2d-32          [-1, 192, 56, 56]             384
          Sigmoid-33          [-1, 192, 56, 56]               0
            Swish-34          [-1, 192, 56, 56]               0
           Conv2d-35          [-1, 192, 28, 28]           1,728
      BatchNorm2d-36          [-1, 192, 28, 28]             384
AdaptiveAvgPool2d-37            [-1, 192, 1, 1]               0
           Conv2d-38             [-1, 16, 1, 1]           3,088
      BatchNorm2d-39             [-1, 16, 1, 1]              32
             ReLU-40             [-1, 16, 1, 1]               0
           Conv2d-41            [-1, 192, 1, 1]           3,264
          Sigmoid-42            [-1, 192, 1, 1]               0
               SE-43          [-1, 192, 28, 28]               0
            ReLU6-44          [-1, 192, 28, 28]               0
           Conv2d-45           [-1, 40, 28, 28]           7,680
      BatchNorm2d-46           [-1, 40, 28, 28]              80
 LinearBottleneck-47           [-1, 40, 28, 28]               0
           Conv2d-48          [-1, 240, 28, 28]           9,600
      BatchNorm2d-49          [-1, 240, 28, 28]             480
          Sigmoid-50          [-1, 240, 28, 28]               0
            Swish-51          [-1, 240, 28, 28]               0
           Conv2d-52          [-1, 240, 28, 28]           2,160
      BatchNorm2d-53          [-1, 240, 28, 28]             480
AdaptiveAvgPool2d-54            [-1, 240, 1, 1]               0
           Conv2d-55             [-1, 20, 1, 1]           4,820
      BatchNorm2d-56             [-1, 20, 1, 1]              40
             ReLU-57             [-1, 20, 1, 1]               0
           Conv2d-58            [-1, 240, 1, 1]           5,040
          Sigmoid-59            [-1, 240, 1, 1]               0
               SE-60          [-1, 240, 28, 28]               0
            ReLU6-61          [-1, 240, 28, 28]               0
           Conv2d-62           [-1, 48, 28, 28]          11,520
      BatchNorm2d-63           [-1, 48, 28, 28]              96
 LinearBottleneck-64           [-1, 48, 28, 28]               0
           Conv2d-65          [-1, 288, 28, 28]          13,824
      BatchNorm2d-66          [-1, 288, 28, 28]             576
          Sigmoid-67          [-1, 288, 28, 28]               0
            Swish-68          [-1, 288, 28, 28]               0
           Conv2d-69          [-1, 288, 14, 14]           2,592
      BatchNorm2d-70          [-1, 288, 14, 14]             576
AdaptiveAvgPool2d-71            [-1, 288, 1, 1]               0
           Conv2d-72             [-1, 24, 1, 1]           6,936
      BatchNorm2d-73             [-1, 24, 1, 1]              48
             ReLU-74             [-1, 24, 1, 1]               0
           Conv2d-75            [-1, 288, 1, 1]           7,200
          Sigmoid-76            [-1, 288, 1, 1]               0
               SE-77          [-1, 288, 14, 14]               0
            ReLU6-78          [-1, 288, 14, 14]               0
           Conv2d-79           [-1, 55, 14, 14]          15,840
      BatchNorm2d-80           [-1, 55, 14, 14]             110
 LinearBottleneck-81           [-1, 55, 14, 14]               0
           Conv2d-82          [-1, 330, 14, 14]          18,150
      BatchNorm2d-83          [-1, 330, 14, 14]             660
          Sigmoid-84          [-1, 330, 14, 14]               0
            Swish-85          [-1, 330, 14, 14]               0
           Conv2d-86          [-1, 330, 14, 14]           2,970
      BatchNorm2d-87          [-1, 330, 14, 14]             660
AdaptiveAvgPool2d-88            [-1, 330, 1, 1]               0
           Conv2d-89             [-1, 27, 1, 1]           8,937
      BatchNorm2d-90             [-1, 27, 1, 1]              54
             ReLU-91             [-1, 27, 1, 1]               0
           Conv2d-92            [-1, 330, 1, 1]           9,240
          Sigmoid-93            [-1, 330, 1, 1]               0
               SE-94          [-1, 330, 14, 14]               0
            ReLU6-95          [-1, 330, 14, 14]               0
           Conv2d-96           [-1, 63, 14, 14]          20,790
      BatchNorm2d-97           [-1, 63, 14, 14]             126
 LinearBottleneck-98           [-1, 63, 14, 14]               0
           Conv2d-99          [-1, 378, 14, 14]          23,814
     BatchNorm2d-100          [-1, 378, 14, 14]             756
         Sigmoid-101          [-1, 378, 14, 14]               0
           Swish-102          [-1, 378, 14, 14]               0
          Conv2d-103          [-1, 378, 14, 14]           3,402
     BatchNorm2d-104          [-1, 378, 14, 14]             756
AdaptiveAvgPool2d-105            [-1, 378, 1, 1]               0
          Conv2d-106             [-1, 31, 1, 1]          11,749
     BatchNorm2d-107             [-1, 31, 1, 1]              62
            ReLU-108             [-1, 31, 1, 1]               0
          Conv2d-109            [-1, 378, 1, 1]          12,096
         Sigmoid-110            [-1, 378, 1, 1]               0
              SE-111          [-1, 378, 14, 14]               0
           ReLU6-112          [-1, 378, 14, 14]               0
          Conv2d-113           [-1, 71, 14, 14]          26,838
     BatchNorm2d-114           [-1, 71, 14, 14]             142
LinearBottleneck-115           [-1, 71, 14, 14]               0
          Conv2d-116          [-1, 426, 14, 14]          30,246
     BatchNorm2d-117          [-1, 426, 14, 14]             852
         Sigmoid-118          [-1, 426, 14, 14]               0
           Swish-119          [-1, 426, 14, 14]               0
          Conv2d-120          [-1, 426, 14, 14]           3,834
     BatchNorm2d-121          [-1, 426, 14, 14]             852
AdaptiveAvgPool2d-122            [-1, 426, 1, 1]               0
          Conv2d-123             [-1, 35, 1, 1]          14,945
     BatchNorm2d-124             [-1, 35, 1, 1]              70
            ReLU-125             [-1, 35, 1, 1]               0
          Conv2d-126            [-1, 426, 1, 1]          15,336
         Sigmoid-127            [-1, 426, 1, 1]               0
              SE-128          [-1, 426, 14, 14]               0
           ReLU6-129          [-1, 426, 14, 14]               0
          Conv2d-130           [-1, 79, 14, 14]          33,654
     BatchNorm2d-131           [-1, 79, 14, 14]             158
LinearBottleneck-132           [-1, 79, 14, 14]               0
          Conv2d-133          [-1, 474, 14, 14]          37,446
     BatchNorm2d-134          [-1, 474, 14, 14]             948
         Sigmoid-135          [-1, 474, 14, 14]               0
           Swish-136          [-1, 474, 14, 14]               0
          Conv2d-137          [-1, 474, 14, 14]           4,266
     BatchNorm2d-138          [-1, 474, 14, 14]             948
AdaptiveAvgPool2d-139            [-1, 474, 1, 1]               0
          Conv2d-140             [-1, 39, 1, 1]          18,525
     BatchNorm2d-141             [-1, 39, 1, 1]              78
            ReLU-142             [-1, 39, 1, 1]               0
          Conv2d-143            [-1, 474, 1, 1]          18,960
         Sigmoid-144            [-1, 474, 1, 1]               0
              SE-145          [-1, 474, 14, 14]               0
           ReLU6-146          [-1, 474, 14, 14]               0
          Conv2d-147           [-1, 87, 14, 14]          41,238
     BatchNorm2d-148           [-1, 87, 14, 14]             174
LinearBottleneck-149           [-1, 87, 14, 14]               0
          Conv2d-150          [-1, 522, 14, 14]          45,414
     BatchNorm2d-151          [-1, 522, 14, 14]           1,044
         Sigmoid-152          [-1, 522, 14, 14]               0
           Swish-153          [-1, 522, 14, 14]               0
          Conv2d-154          [-1, 522, 14, 14]           4,698
     BatchNorm2d-155          [-1, 522, 14, 14]           1,044
AdaptiveAvgPool2d-156            [-1, 522, 1, 1]               0
          Conv2d-157             [-1, 43, 1, 1]          22,489
     BatchNorm2d-158             [-1, 43, 1, 1]              86
            ReLU-159             [-1, 43, 1, 1]               0
          Conv2d-160            [-1, 522, 1, 1]          22,968
         Sigmoid-161            [-1, 522, 1, 1]               0
              SE-162          [-1, 522, 14, 14]               0
           ReLU6-163          [-1, 522, 14, 14]               0
          Conv2d-164           [-1, 95, 14, 14]          49,590
     BatchNorm2d-165           [-1, 95, 14, 14]             190
LinearBottleneck-166           [-1, 95, 14, 14]               0
          Conv2d-167          [-1, 570, 14, 14]          54,150
     BatchNorm2d-168          [-1, 570, 14, 14]           1,140
         Sigmoid-169          [-1, 570, 14, 14]               0
           Swish-170          [-1, 570, 14, 14]               0
          Conv2d-171            [-1, 570, 7, 7]           5,130
     BatchNorm2d-172            [-1, 570, 7, 7]           1,140
AdaptiveAvgPool2d-173            [-1, 570, 1, 1]               0
          Conv2d-174             [-1, 47, 1, 1]          26,837
     BatchNorm2d-175             [-1, 47, 1, 1]              94
            ReLU-176             [-1, 47, 1, 1]               0
          Conv2d-177            [-1, 570, 1, 1]          27,360
         Sigmoid-178            [-1, 570, 1, 1]               0
              SE-179            [-1, 570, 7, 7]               0
           ReLU6-180            [-1, 570, 7, 7]               0
          Conv2d-181            [-1, 103, 7, 7]          58,710
     BatchNorm2d-182            [-1, 103, 7, 7]             206
LinearBottleneck-183            [-1, 103, 7, 7]               0
          Conv2d-184            [-1, 618, 7, 7]          63,654
     BatchNorm2d-185            [-1, 618, 7, 7]           1,236
         Sigmoid-186            [-1, 618, 7, 7]               0
           Swish-187            [-1, 618, 7, 7]               0
          Conv2d-188            [-1, 618, 7, 7]           5,562
     BatchNorm2d-189            [-1, 618, 7, 7]           1,236
AdaptiveAvgPool2d-190            [-1, 618, 1, 1]               0
          Conv2d-191             [-1, 51, 1, 1]          31,569
     BatchNorm2d-192             [-1, 51, 1, 1]             102
            ReLU-193             [-1, 51, 1, 1]               0
          Conv2d-194            [-1, 618, 1, 1]          32,136
         Sigmoid-195            [-1, 618, 1, 1]               0
              SE-196            [-1, 618, 7, 7]               0
           ReLU6-197            [-1, 618, 7, 7]               0
          Conv2d-198            [-1, 110, 7, 7]          67,980
     BatchNorm2d-199            [-1, 110, 7, 7]             220
LinearBottleneck-200            [-1, 110, 7, 7]               0
          Conv2d-201            [-1, 660, 7, 7]          72,600
     BatchNorm2d-202            [-1, 660, 7, 7]           1,320
         Sigmoid-203            [-1, 660, 7, 7]               0
           Swish-204            [-1, 660, 7, 7]               0
          Conv2d-205            [-1, 660, 7, 7]           5,940
     BatchNorm2d-206            [-1, 660, 7, 7]           1,320
AdaptiveAvgPool2d-207            [-1, 660, 1, 1]               0
          Conv2d-208             [-1, 55, 1, 1]          36,355
     BatchNorm2d-209             [-1, 55, 1, 1]             110
            ReLU-210             [-1, 55, 1, 1]               0
          Conv2d-211            [-1, 660, 1, 1]          36,960
         Sigmoid-212            [-1, 660, 1, 1]               0
              SE-213            [-1, 660, 7, 7]               0
           ReLU6-214            [-1, 660, 7, 7]               0
          Conv2d-215            [-1, 118, 7, 7]          77,880
     BatchNorm2d-216            [-1, 118, 7, 7]             236
LinearBottleneck-217            [-1, 118, 7, 7]               0
          Conv2d-218            [-1, 708, 7, 7]          83,544
     BatchNorm2d-219            [-1, 708, 7, 7]           1,416
         Sigmoid-220            [-1, 708, 7, 7]               0
           Swish-221            [-1, 708, 7, 7]               0
          Conv2d-222            [-1, 708, 7, 7]           6,372
     BatchNorm2d-223            [-1, 708, 7, 7]           1,416
AdaptiveAvgPool2d-224            [-1, 708, 1, 1]               0
          Conv2d-225             [-1, 59, 1, 1]          41,831
     BatchNorm2d-226             [-1, 59, 1, 1]             118
            ReLU-227             [-1, 59, 1, 1]               0
          Conv2d-228            [-1, 708, 1, 1]          42,480
         Sigmoid-229            [-1, 708, 1, 1]               0
              SE-230            [-1, 708, 7, 7]               0
           ReLU6-231            [-1, 708, 7, 7]               0
          Conv2d-232            [-1, 126, 7, 7]          89,208
     BatchNorm2d-233            [-1, 126, 7, 7]             252
LinearBottleneck-234            [-1, 126, 7, 7]               0
          Conv2d-235            [-1, 756, 7, 7]          95,256
     BatchNorm2d-236            [-1, 756, 7, 7]           1,512
         Sigmoid-237            [-1, 756, 7, 7]               0
           Swish-238            [-1, 756, 7, 7]               0
          Conv2d-239            [-1, 756, 7, 7]           6,804
     BatchNorm2d-240            [-1, 756, 7, 7]           1,512
AdaptiveAvgPool2d-241            [-1, 756, 1, 1]               0
          Conv2d-242             [-1, 63, 1, 1]          47,691
     BatchNorm2d-243             [-1, 63, 1, 1]             126
            ReLU-244             [-1, 63, 1, 1]               0
          Conv2d-245            [-1, 756, 1, 1]          48,384
         Sigmoid-246            [-1, 756, 1, 1]               0
              SE-247            [-1, 756, 7, 7]               0
           ReLU6-248            [-1, 756, 7, 7]               0
          Conv2d-249            [-1, 134, 7, 7]         101,304
     BatchNorm2d-250            [-1, 134, 7, 7]             268
LinearBottleneck-251            [-1, 134, 7, 7]               0
          Conv2d-252            [-1, 896, 7, 7]         120,064
     BatchNorm2d-253            [-1, 896, 7, 7]           1,792
         Sigmoid-254            [-1, 896, 7, 7]               0
           Swish-255            [-1, 896, 7, 7]               0
AdaptiveAvgPool2d-256            [-1, 896, 1, 1]               0
         Dropout-257            [-1, 896, 1, 1]               0
          Conv2d-258              [-1, 4, 1, 1]           3,588
================================================================
Total params: 1,939,058
Trainable params: 1,939,058
Non-trainable params: 0
----------------------------------------------------------------
Input size (MB): 0.57
Forward/backward pass size (MB): 184.59
Params size (MB): 7.40
Estimated Total Size (MB): 192.56
----------------------------------------------------------------
반응형