Pytorch实现ResNet

目标：用pytorch实现下图所示的网络

 代码：

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

class ResBlock(nn.Module): #残差块的实现也是继承nn.module后实现一个类，同样的要实现__init__()方法和forward方法
    def __init__(self, n_chans):
        super().__init__()
        self.conv = nn.Conv2d(n_chans, n_chans, kernel_size=3, padding=1, bias=False)
        self.batch_norm = nn.BatchNorm2d(n_chans)
        torch.nn.init.kaiming_normal_(self.conv.weight, nonlinearity='relu') #参数初始化
        torch.nn.init.constant_(self.batch_norm.weight, 0.5)
        torch.nn.init.zeros_(self.batch_norm.bias)

    def forward(self,x):
        out = self.conv(x)
        out = self.batch_norm(out)
        out = F.relu(out)
        return out + x

class NetResDepp(nn.Module):
    def __init__(self, n_chans1=32, num_blocks=100):
        super().__init__()
        self.n_chans1 = n_chans1
        self.num_blocks = num_blocks
        self.conv = nn.Conv2d(3, n_chans1, kernel_size=3, padding=1)
        self.resblocks = nn.Sequential(*(num_blocks * [*ResBlock(n_chans=n_chans1)])) # 注意这里的100个Resblock是通过先对ResBlock解包放到列表里，再用100乘这个列表就实现了将列表复制100倍，再解包就实现了100个ResBlock
        self.fc1 = nn.Linear(8 * 8 * n_chans1, 32)
        self.fc2 = nn.Linear(32,2)
    def forward(self, x):
        out = F.relu(self.conv(x))
        out = F.max_pool2d(out, 2)
        out = self.resblocks(out)
        out = F.max_pool2d(out, 2)
        out = out.view(-1, 8 * 8 * self.n_chans1)
        out = self.fc1(out)
        out = self.fc2(out)
        return out

参考资料：

pytorch深度学习实战（伊莱史蒂文斯）