从头设计一个pytorch的神经网络

import torch
import cv2
import numpy as np
from torchvision import datasets, transforms
import torchvision

# 1. 使用torchvision加载数据集并做预处理
transform = transforms.Compose([transforms.ToTensor(),  # 把图像转换为tensor
                                transforms.Normalize((0.5,0.5,0.5), (0.5,0.5,0.5)),  # 归一化处理
                              ])
# 加载训练集和测试集
trainset = torchvision.datasets.CIFAR10(root='E:\\Machine Learning\\PyTorch\\CIFAR10', train=True, download=True, transform=transform)
trainloader = torch.utils.data.DataLoader(trainset, batch_size=4, shuffle=True, num_workers=2)
testset = torchvision.datasets.CIFAR10(root='E:\\Machine Learning\\PyTorch\\CIFAR10', train=False, download=True, transform=transform)
testloader = torch.utils.data.DataLoader(testset, batch_size=4, shuffle=False, num_workers=2)

# 2. 定义卷积神经网络
import torch.nn as nn
import torch.nn.functional as F

class MyModel(nn.Module):  #  继承nn.Module
    # 定义网络结构
    def __init__(self):
        super(MyModel, self).__init__()
        self.conv1 = nn.Conv2d(3,6,5)
        self.pool = nn.MaxPool2d(2,2)
        self.conv2 = nn.Conv2d(6,16,5)
        self.fc1 = nn.Linear(16*5*5, 120)
        self.fc2 = nn.Linear(120, 84)
        self.fc3 = nn.Linear(84, 10)
    # 定义前向传播过程
    def forward(self, x):
        x = self.pool(F.relu(self.conv1(x)))
        x = self.pool(F.relu(self.conv2(x)))
        x = x.view(-1, 16*5*5)
        x = F.relu(self.fc1(x))
        x = F.relu(self.fc2(x))
        x = self.fc3(x)

        return x

Net = MyModel()

'''这里解释一下为什么第一个全连接层的输入大小为16x5x5，16是因为最后的卷积层有16个filter（即输出的feature map有16个channel），后面的5x5并不是因为最后卷积层的kernel大小为5x5，而是因为最后的feature map大小为5x5'''
![在这里插入图片描述](https://img-blog.csdnimg.cn/d410f91184334a61ae2991e4e9be827f.png#pic_center)




# 3. 定义损失函数和优化器
import torch.optim as optim

criterion = nn.CrossEntropyLoss()  # 多分类问题，用交叉熵损失函数
optimizer = optim.SGD(Net.parameters(), lr=0.001, momentum=0.9)  # 用SGD优化器

# 4. 训练神经网络
epochs = 2  # 训练两个epoch，batch_size = 4 (batch_size的大小定义在第一步torch.utils.data.DataLoader中)
e1 = cv2.getTickCount()  # 记录训练时间
for epoch in range(epochs):
    total_loss = 0.0
    for i, data in enumerate(trainloader, 0):
        # 得到inputs
        inputs, labels = data
        optimizer.zero_grad()
        # forward + backward + optimize
        # 前向传播+反向传播+更新参数
        outputs = Net(inputs)  # 前向传播，得到outputs
        loss = criterion(outputs, labels)  # 得到损失函数
        loss.backward()  # 后向传播
        optimizer.step()  # 更新参数

        # 输出训练过程
        total_loss += loss.item()
        if (i+1) % 1000 == 0:  # 每1000次(就是4000张图像)输出一次loss
            print('第{}个epoch：第{:5d}次：目前的训练损失loss为：{:.3f}'.format(epoch+1, i+1, total_loss/1000))
            total_loss = 0.0

e2 = cv2.getTickCount()
print('用CPU训练总共用时：{} s'.format((e2-e1)/cv2.getTickFrequency()))


# 5. 测试模型准确率如何

correct = 0
total = 0

e1 = cv2.getTickCount()  # 记录测试时间
with torch.no_grad():  # 不跟踪梯度
    for data in testloader:
        images, labels = data
        outputs = Net(images)
        _, predicted = torch.max(outputs.data, 1)
        total += labels.size(0)
        correct += (predicted == labels).sum().item()
        
e2 = cv2.getTickCount()
print('用CPU测试总共用时：{} s'.format((e2-e1)/cv2.getTickFrequency()))
print('在测试集上的准确率为：{:.3f}%'.format(correct*100/total))