[summary] 显示的shape第一维的值和pdb调试打印的 shape 不一致 ？

[vfdff]

根据 summary 描述进行了简单的验证，确实对于其中的测试用例输出了 forward() 部份印出的模型

import torch
import torch.nn as nn
import torch.nn.functional as F
from torchsummary import summary
# Model
class CNN(nn.Module):
    def __init__(self, classes):
        super(CNN, self).__init__()
        self.conv_1 = nn.Conv2d(in_channels=1,
                              out_channels=16,
                              kernel_size=5,
                              stride=1,
                              padding=0)

        self.conv_2 = nn.Conv2d(in_channels=16,
                                out_channels=32,
                                kernel_size=5,
                                stride=1,
                                padding=0)
        self.relu = nn.ReLU()
        self.max_pool = nn.MaxPool2d(kernel_size=2)
        self.fc = nn.Linear(32*4*4, classes)

    def forward(self, x):
        x = self.conv_1(x)
        x = self.relu(x)
        x = self.max_pool(x)
        x = self.conv_2(x)
        x = self.relu(x)
        x = self.max_pool(x)
        x = x.view(x.size(0), -1)
        x = self.fc(x)
        return x


if __name__ == '__main__':
    cnn = CNN(3000)
    print(cnn)
    summary(cnn, (1, 28, 28))

当前的显示结果中第一维使用了 -1，即无法推断吗 ？

----------------------------------------------------------------
        Layer (type)               Output Shape         Param #
            Conv2d-1           [-1, 16, 24, 24]             416
              ReLU-2           [-1, 16, 24, 24]               0
         MaxPool2d-3           [-1, 16, 12, 12]               0
            Conv2d-4             [-1, 32, 8, 8]          12,832
              ReLU-5             [-1, 32, 8, 8]               0
         MaxPool2d-6             [-1, 32, 4, 4]               0
            Linear-7                 [-1, 3000]       1,539,000

但是，直接基于pdb调试打印，我们能够看到第一维的大小是 2，不知道这个差异是否可以改进？

(Pdb) l
 27  	         self.max_pool = nn.MaxPool2d(kernel_size=2)
 28  	         self.fc = nn.Linear(32*4*4, classes)
 29  	
 30  	     def forward(self, x):
 31 B	         x = self.conv_1(x)
 32  ->	         x = self.relu(x)
 33  	         x = self.max_pool(x)
 34  	         x = self.conv_2(x)
 35  	         x = self.relu(x)
 36  	         x = self.max_pool(x)
 37  	         x = x.view(x.size(0), -1)
(Pdb) p x.shape
torch.Size([2, 16, 24, 24])

[vfdff]

oh, I think this is similar to sksq96/pytorch-summary#168