（1）输入图像，1×32×32，通道数1，高32，宽32
（2）卷积层1，滤波器的shape为6×1×5×5，滤波器个数6，通道数1，高5，宽5。卷积层1的输出为6×28×28。
（3）relu层，输出和输入的shape相同
（4）池化层，滤波器的大小为2×2，stride为2×2，输出为6×14×14。
（5）卷积层2，滤波器的shape为16×6×5×5，滤波器个数16，通道数6，高5，宽5。卷积层2的输出为16×10×10。
（6）relu层，输出和输入的shape相同
（7）池化层，滤波器的大小为2×2，stride为2×2，输出为16×5×5。reshape为1×400后输入到全连接层1。
（8）全连接层1，权重的shape为400×120，输出为1×120
（9）relu层，输出和输入的shape相同
（10）全连接层2，权重的shape为120×84，输出为1×84
（11）relu层，输出和输入的shape相同
（12）全连接层3，权重的shape为84×10，输出为1×10

  这个网络的结构如下：

input -> conv2d -> relu -> maxpool2d -> conv2d -> relu -> maxpool2d
      -> flatten -> linear -> relu -> linear -> relu -> linear
      -> MSELoss
      -> loss

  Let’s define this network:

import torch
import torch.nn as nn
import torch.nn.functional as F


class Net(nn.Module):

    def __init__(self):
        super(Net, self).__init__()
        # 1 input image channel, 6 output channels, 5x5 square convolution
        # kernel
        self.conv1 = nn.Conv2d(1, 6, 5)
        self.conv2 = nn.Conv2d(6, 16, 5)
        # an affine operation: y = Wx + b
        self.fc1 = nn.Linear(16 * 5 * 5, 120)  # 5*5 from image dimension
        self.fc2 = nn.Linear(120, 84)
        self.fc3 = nn.Linear(84, 10)

    def forward(self, x):
        # Max pooling over a (2, 2) window
        x = F.max_pool2d(F.relu(self.conv1(x)), (2, 2))
        # If the size is a square, you can specify with a single number
        x = F.max_pool2d(F.relu(self.conv2(x)), 2)
        x = torch.flatten(x, 1) # flatten all dimensions except the batch dimension
        x = F.relu(self.fc1(x))
        x = F.relu(self.fc2(x))
        x = self.fc3(x)
        return x


net = Net()
print(net)

Net(
  (conv1): Conv2d(1, 6, kernel_size=(5, 5), stride=(1, 1))
  (conv2): Conv2d(6, 16, kernel_size=(5, 5), stride=(1, 1))
  (fc1): Linear(in_features=400, out_features=120, bias=True)
  (fc2): Linear(in_features=120, out_features=84, bias=True)
  (fc3): Linear(in_features=84, out_features=10, bias=True)
)

  这里写代码测试一下flatten函数的用法：

>>> data=[[[1,2,3],[4,5,6]], [[7,8,9],[10,11,12]]]
>>> x_data = torch.tensor(data)
>>> print(f"x_data: \n {x_data} \n")
x_data:
 tensor([[[ 1,  2,  3],
         [ 4,  5,  6]],

        [[ 7,  8,  9],
         [10, 11, 12]]])

>>> print(f"Shape of x_data: {x_data.shape}")
Shape of x_data: torch.Size([2, 2, 3])
>>> x_data = torch.flatten(x_data, 1)
>>> print(f"x_data: \n {x_data} \n")
x_data:
 tensor([[ 1,  2,  3,  4,  5,  6],
        [ 7,  8,  9, 10, 11, 12]])

>>> print(f"Shape of x_data: {x_data.shape}")
Shape of x_data: torch.Size([2, 6])

  该网络有2个卷积层和3个全连接层，有5个权重参数和5个偏置参数，共10个参数。

params = list(net.parameters())
print(len(params))
for i in range(len(params)):
    print(params[i].size())

  以上代码打印了每个参数的shape，如下：

10
torch.Size([6, 1, 5, 5])
torch.Size([6])
torch.Size([16, 6, 5, 5])
torch.Size([16])
torch.Size([120, 400])
torch.Size([120])
torch.Size([84, 120])
torch.Size([84])
torch.Size([10, 84])
torch.Size([10])