现代卷积网络实战系列2：训练函数、PyTorch构建LeNet网络

4、训练函数

4.1 调用训练函数

train(epochs, net, train_loader, device, optimizer, test_loader, true_value)

因为每一个epoch训练结束后，我们需要测试一下这个网络的性能，所有会在训练函数中频繁调用测试函数，所有测试函数中所有需要的参数，训练函数都需要
这七个参数，是训练一个神经网络所需要的最少参数

4.2 训练函数

训练函数中，所有训练集进行多次迭代，而每次迭代又会将数据分成多个批次进行迭代

def train(epochs, net, train_loader, device, optimizer, test_loader, true_value):
    for epoch in range(1, epochs + 1):
        net.train()
        all_train_loss = []
        for batch_idx, (data, target) in enumerate(train_loader):
            data = data.to(device)
            target = target.to(device)
            optimizer.zero_grad()
            output = net(data)
            loss = F.cross_entropy(output, target)
            loss.backward()
            optimizer.step()
            cur_train_loss = loss.item()
            all_train_loss.append(cur_train_loss)
        train_loss = np.round(np.mean(all_train_loss) * 1000, 2)
        print('\nepoch step:', epoch)
        print('training loss: ', train_loss)
        test(net, test_loader, device, true_value, epoch)
    print("\nTraining finished")

1. 定义训练函数
2. 安装epochs迭代数据
3. 进入pytorch的训练模式
4. all_train_loss 存放训练集5万张图片的损失值
5. 按照batch取数据
6. 数据进入GPU
7. 标签进入GPU
8. 梯度清零
9. 当前batch进入网络后得到输出
10. 根据输出得到当前损失
11. 反向传播
12. 梯度下降
13. 获取损失的损失值（PyTorch框架中的数据）
14. 把当前batch的损失加入all_train_loss数组中，结束batch的迭代
15. 将5张图片的损失计算出来并且进行求平均，这里乘以1000是因为我觉得计算出的损失太小了，所以乘以1000，方便看损失的变化，保留两位有效数字
16. 打印当前epoch
17. 打印损失
18. 调用测试函数，测试当前训练的网络的性能，结束epoch的迭代
19. 打印训练完成

5、LeNet

5.1 网络结构

LeNet可以说是首次提出卷积神经网络的模型
主要包含下面的网络层：

1. 5*5的二维卷积
2. sigmoid激活函数（这里使用了relu）
3. 5*5的二维卷积
4. sigmoid激活函数
5. 数据一维化
6. 全连接层
7. 全连接层
8. softmax分类器

将网络结构打印出来：

LeNet(
-------(conv1): Conv2d(1, 10, kernel_size=(5, 5), stride=(1, 1))
-------(conv2): Conv2d(10, 20, kernel_size=(5, 5), stride=(1, 1))
-------(conv2_drop): Dropout2d(p=0.5, inplace=False)
-------(fc1): Linear(in_features=320, out_features=50, bias=True)
-------(fc2): Linear(in_features=50, out_features=10, bias=True)
)

5.2 PyTorch构建LeNet

class LeNet(nn.Module):
    def __init__(self, num_classes):
        super(LeNet, self).__init__()
        self.conv1 = nn.Conv2d(1, 10, kernel_size=5)
        self.conv2 = nn.Conv2d(10, 20, kernel_size=5)
        self.conv2_drop = nn.Dropout2d()
        self.fc1 = nn.Linear(320, 50)
        self.fc2 = nn.Linear(50, num_classes)

    def forward(self, x):
        x = F.relu(F.max_pool2d(self.conv1(x), 2))
        x = F.relu(F.max_pool2d(self.conv2_drop(self.conv2(x)), 2))
        x = x.view(-1, 320)
        x = F.relu(self.fc1(x))
        x = F.dropout(x, training=self.training)
        x = self.fc2(x)
        return F.log_softmax(x, dim=1)

这个时候已经是一个完整的项目了，看看10个epoch训练过程的打印：

D:\conda\envs\pytorch\python.exe A:\0_MNIST\train.py

Reading data…
train_data: (60000, 28, 28) train_label (60000,)
test_data: (10000, 28, 28) test_label (10000,)

Initialize neural network
test loss: 2301.68
test accuracy: 11.3 %

epoch step: 1
training loss: 634.74
test loss: 158.03
test accuracy: 95.29 %

epoch step: 2
training loss: 324.04
test loss: 107.62
test accuracy: 96.55 %

epoch step: 3
training loss: 271.25
test loss: 88.43
test accuracy: 97.04 %

epoch step: 4
training loss: 236.69
test loss: 70.94
test accuracy: 97.61 %

epoch step: 5
training loss: 211.05
test loss: 69.69
test accuracy: 97.72 %

epoch step: 6
training loss: 199.28
test loss: 62.04
test accuracy: 97.98 %

epoch step: 7
training loss: 187.11
test loss: 59.65
test accuracy: 97.98 %

epoch step: 8
training loss: 178.79
test loss: 53.89
test accuracy: 98.2 %

epoch step: 9
training loss: 168.75
test loss: 51.83
test accuracy: 98.43 %

epoch step: 10
training loss: 160.83
test loss: 50.35
test accuracy: 98.4 %

Training finished
进程已结束，退出代码为 0

可以看出基本上只要一个epoch就可以得到很好的训练效果了，后续的epoch中的提升比较小