苍穹外卖

使用pytorch框架手写数字识别

用mnist数据集训练一个手写数字图片识别的神经网络。

该神经网络使用了三层cnn卷积神经网络和两层全连接构建神经网络，并使用了一层Dropout防止过拟合。神经网络使用pytorch框架搭建，并调用gpu提高运算速度，最终模型保存于custom_cnn_model.pth中。

import torch
import torch.nn as nn
import torch.optim as optim
import torchvision.transforms as transforms
from torch.utils.data import DataLoader
from torchvision.datasets import MNIST
import matplotlib.pyplot as plt

# 检查GPU是否可用
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

# 定义CNN模型
class CustomCNN(nn.Module):
    def __init__(self):
        super(CustomCNN, self).__init__()
        #使用3个卷积层和3个池化层构建CNN模型
        self.conv1 = nn.Conv2d(1, 32, kernel_size=3, stride=1, padding=1)
        self.relu1 = nn.ReLU()
        self.pool1 = nn.MaxPool2d(kernel_size=2, stride=2)

        self.conv2 = nn.Conv2d(32, 64, kernel_size=3, stride=1, padding=1)
        self.relu2 = nn.ReLU()
        self.pool2 = nn.MaxPool2d(kernel_size=2, stride=2)

        self.conv3 = nn.Conv2d(64, 128, kernel_size=3, stride=1, padding=1)
        self.relu3 = nn.ReLU()
        self.pool3 = nn.MaxPool2d(kernel_size=2, stride=2)

        self.flatten = nn.Flatten()# 将多维张量展平为一维张量
        self.fc1 = nn.Linear(128 * 3 * 3, 512)# 全连接层
        self.relu4 = nn.ReLU()
        self.dropout = nn.Dropout(0.5)  # 添加Dropout
        self.fc2 = nn.Linear(512, 10)

    def forward(self, x):
        x = self.pool1(self.relu1(self.conv1(x)))
        x = self.pool2(self.relu2(self.conv2(x)))
        x = self.pool3(self.relu3(self.conv3(x)))
        x = self.flatten(x)
        x = self.relu4(self.fc1(x))
        x = self.dropout(x)
        x = self.fc2(x)
        return x

# 设置训练参数
batch_size = 64
learning_rate = 0.001
epochs = 10

# 加载MNIST数据集
transform = transforms.Compose([transforms.ToTensor(), transforms.Normalize((0.5,), (0.5,))])
train_dataset = MNIST(root='./data', train=True, download=True, transform=transform)
test_dataset = MNIST(root='./data', train=False, download=True, transform=transform)

train_loader = DataLoader(dataset=train_dataset, batch_size=batch_size, shuffle=True)
test_loader = DataLoader(dataset=test_dataset, batch_size=batch_size, shuffle=False)

# 初始化模型、损失函数和优化器，并将它们移动到GPU上
model = CustomCNN().to(device)
criterion = nn.CrossEntropyLoss()# 交叉熵损失函数
optimizer = optim.Adam(model.parameters(), lr=learning_rate)# Adam优化器

# 记录训练和测试集的损失值
train_loss_values = []
test_loss_values = []

# 训练模型
for epoch in range(epochs):
    model.train()
    total_train_loss = 0.0

    for i, (images, labels) in enumerate(train_loader):
        images, labels = images.to(device), labels.to(device)# 将数据移动到GPU上

        optimizer.zero_grad()# 梯度清零
        outputs = model(images)# 前向传播
        loss = criterion(outputs, labels)# 计算损失
        loss.backward()
        optimizer.step()

        total_train_loss += loss.item()# 累加损失值

    # 计算平均训练集损失
    average_train_loss = total_train_loss / len(train_loader)
    train_loss_values.append(average_train_loss)

    # 在每个epoch结束时计算测试集损失
    model.eval()
    total_test_loss = 0.0

    with torch.no_grad():
        for images, labels in test_loader:
            images, labels = images.to(device), labels.to(device)
            outputs = model(images)
            loss = criterion(outputs, labels)
            total_test_loss += loss.item()

    # 计算平均测试集损失
    average_test_loss = total_test_loss / len(test_loader)
    test_loss_values.append(average_test_loss)

    print('Epoch [{}/{}], Train Loss: {:.4f}, Test Loss: {:.4f}'.format(epoch + 1, epochs, average_train_loss, average_test_loss))

# 绘制损失函数变化折线图
plt.plot(train_loss_values, label='Train Loss')
plt.plot(test_loss_values, label='Test Loss')
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.title('Train and Test Loss Curve')
plt.legend()
plt.show()

# 保存模型
torch.save(model.state_dict(), 'custom_cnn_model.pth')

运行截图：

每个epoch后训练集和测试集的损失函数值：

---

转载请注明来源，欢迎对文章中的引用来源进行考证，欢迎指出任何有错误或不够清晰的表达。可以在下面评论区评论