pytorch实现LeNet,手写数字识别为例

import torch
import torchvision
import torch.utils.data as data
import torch.nn as nn
import matplotlib.pyplot as plt
import numpy as np

transform = torchvision.transforms.ToTensor()

train_dataset = torchvision.datasets.MNIST(root='./data', train=True, transform=transform, download=True)
test_dataset = torchvision.datasets.MNIST(root='./data', train=False, transform=transform, download=True)

batch_size = 128
train_loader = data.DataLoader(dataset=train_dataset, batch_size=batch_size, shuffle=True)
test_loader = data.DataLoader(dataset=test_dataset, batch_size=batch_size, shuffle=True)


class LeNet(nn.Module):
    def __init__(self):
        super(LeNet, self).__init__()
        self.conv1 = nn.Sequential(
            # 1,28,28
            nn.Conv2d(
                in_channels=1,
                out_channels=32,
                kernel_size=5,
                padding=2,
                stride=1
            ),
            # 16,28,28
            nn.ReLU(),
            nn.MaxPool2d(kernel_size=2, stride=2)
            # 16,14,14
        )
        self.conv2 = nn.Sequential(
            nn.Conv2d(
                in_channels=32,
                out_channels=16,
                kernel_size=5
            ),
            # 16,10,10
            nn.ReLU(),
            nn.MaxPool2d(kernel_size=2, stride=2)
            # 16,5,5
        )
        self.fc1 = nn.Sequential(
            nn.Linear(16 * 5 * 5, 120),
            nn.ReLU()
        )
        self.fc2 = nn.Sequential(
            nn.Linear(120, 84),
            nn.ReLU()
        )
        self.fc3 = nn.Linear(84, 10)

    def forward(self, x):
        x = self.conv1(x)
        x = self.conv2(x)
        x = x.view(x.size(0), -1)  # 拉伸
        x = self.fc1(x)
        x = self.fc2(x)
        x = self.fc3(x)
        return x


device = ('cuda' if torch.cuda.is_available() else 'cpu')
net = LeNet().to(device)

# 优化器
optimizer = torch.optim.Adam(net.parameters(), lr=1e-3)
# 损失函数
loss_func = nn.CrossEntropyLoss()

epoches = 1
costs = []

for epoch in range(epoches):
    sum_loss = 0
    net.train()
    for step, (batch_x, batch_y) in enumerate(train_loader):
        if torch.cuda.is_available():
            batch_x = batch_x.cuda()
            batch_y = batch_y.cuda()
        # 梯度清零
        optimizer.zero_grad()
        output = net(batch_x)
        loss = loss_func(output, batch_y)
        loss.backward()
        optimizer.step()
        costs.append(loss)
        sum_loss += loss
        if step % 100 == 0:
            print(f'epoch:{epoch + 1},mini_batch:{step + 1},mini_loss:{sum_loss / 100}')
            sum_loss = 0.0
    # 验证
    net.eval()
    correct = 0
    total = 0
    for (test_x, test_y) in test_loader:
        if torch.cuda.is_available():
            test_x = test_x.cuda()
            test_y = test_y.cuda()
        test_output = net(test_x)
        predicted = torch.max(test_output, 1)[1]
        total += test_y.size(0)
        correct += (predicted == test_y).sum()
    print(f'correct:{correct}')
    print(f'total:{total}')
    print(f'Test acc:{(correct / total * 100):.2f}%')

# 绘制函数
if torch.cuda.is_available():
    costs = [cost.cpu().detach().numpy() for cost in costs]
else:
    costs = [cost.numpy for cost in costs]

plt.plot(costs)
plt.xlabel('number of iteration')
plt.ylabel('loss in train')
plt.title('LeNet')
plt.show()

loss 函数：