torchsummary输出网络结构

佛珠散了一地

已于 2023-01-01 20:50:55 修改

阅读量287

点赞数

分类专栏：网络结构查看文章标签： python 人工智能深度学习

于 2023-01-01 20:38:46 首次发布

本文链接：https://blog.csdn.net/qq_34974229/article/details/128515606

版权

使用了发现还挺好用，自己整理一下
和torchinfo进行了对比，还是这个更直观显示每一层的结构，但是不会像torchinfo一样显示模块部分。各有优劣。

    from torchsummary import summary
    model1 = MobileNetV2(n_class=2).cuda()
    summary(model1, input_size=(3, 256, 256))

输出显示

----------------------------------------------------------------
        Layer (type)               Output Shape         Param #
================================================================
            Conv2d-1         [-1, 32, 128, 128]             864
       BatchNorm2d-2         [-1, 32, 128, 128]              64
             ReLU6-3         [-1, 32, 128, 128]               0
            Conv2d-4         [-1, 32, 128, 128]             288
       BatchNorm2d-5         [-1, 32, 128, 128]              64
             ReLU6-6         [-1, 32, 128, 128]               0
            Conv2d-7         [-1, 16, 128, 128]             512
       BatchNorm2d-8         [-1, 16, 128, 128]              32
  InvertedResidual-9         [-1, 16, 128, 128]               0
           Conv2d-10         [-1, 96, 128, 128]           1,536
      BatchNorm2d-11         [-1, 96, 128, 128]             192
            ReLU6-12         [-1, 96, 128, 128]               0
           Conv2d-13           [-1, 96, 64, 64]             864
      BatchNorm2d-14           [-1, 96, 64, 64]             192
            ReLU6-15           [-1, 96, 64, 64]               0
           Conv2d-16           [-1, 24, 64, 64]           2,304
      BatchNorm2d-17           [-1, 24, 64, 64]              48
 InvertedResidual-18           [-1, 24, 64, 64]               0
           Conv2d-19          [-1, 144, 64, 64]           3,456
      BatchNorm2d-20          [-1, 144, 64, 64]             288
            ReLU6-21          [-1, 144, 64, 64]               0
           Conv2d-22          [-1, 144, 64, 64]           1,296
      BatchNorm2d-23          [-1, 144, 64, 64]             288
            ReLU6-24          [-1, 144, 64, 64]               0
           Conv2d-25           [-1, 24, 64, 64]           3,456
      BatchNorm2d-26           [-1, 24, 64, 64]              48
 InvertedResidual-27           [-1, 24, 64, 64]               0
           Conv2d-28          [-1, 144, 64, 64]           3,456
      BatchNorm2d-29          [-1, 144, 64, 64]             288
            ReLU6-30          [-1, 144, 64, 64]               0
           Conv2d-31          [-1, 144, 32, 32]           1,296
      BatchNorm2d-32          [-1, 144, 32, 32]             288
            ReLU6-33          [-1, 144, 32, 32]               0
           Conv2d-34           [-1, 32, 32, 32]           4,608
      BatchNorm2d-35           [-1, 32, 32, 32]              64
 InvertedResidual-36           [-1, 32, 32, 32]               0
           Conv2d-37          [-1, 192, 32, 32]           6,144
      BatchNorm2d-38          [-1, 192, 32, 32]             384
            ReLU6-39          [-1, 192, 32, 32]               0
           Conv2d-40          [-1, 192, 32, 32]           1,728
      BatchNorm2d-41          [-1, 192, 32, 32]             384
            ReLU6-42          [-1, 192, 32, 32]               0
           Conv2d-43           [