Python学习Day40

学习来源：@浙大疏锦行

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

# # 设置中文字体支持
 plt.rcParams["font.family"] = ["SimHei"]
 plt.rcParams['axes.unicode_minus'] = False  # 解决负号显示问题

# # 1. 数据预处理
 transform = transforms.Compose([
    transforms.ToTensor(),  # 转换为张量并归一化到[0,1]
    transforms.Normalize((0.1307,), (0.3081,))  # MNIST数据集的均值和标准差
 ])

# # 2. 加载MNIST数据集
 train_dataset = datasets.MNIST(
     root='./data',
     train=True,
     download=True,
     transform=transform
 )

 test_dataset = datasets.MNIST(
     root='./data',
     train=False,
     transform=transform
 )

# # 3. 创建数据加载器
 batch_size = 64  # 每批处理64个样本
 train_loader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True)
 test_loader = DataLoader(test_dataset, batch_size=batch_size, shuffle=False)

# # 4. 定义模型、损失函数和优化器
class MLP(nn.Module):
     def __init__(self):
         super(MLP, self).__init__()
         self.flatten = nn.Flatten()  # 将28x28的图像展平为784维向量
         self.layer1 = nn.Linear(784, 128)  # 第一层：784个输入，128个神经元
         self.relu = nn.ReLU()  # 激活函数
         self.layer2 = nn.Linear(128, 10)  # 第二层：128个输入，10个输出（对应10个数字类别）
        
     def forward(self, x):
         x = self.flatten(x)  # 展平图像
         x = self.layer1(x)   # 第一层线性变换
         x = self.relu(x)     # 应用ReLU激活函数         x = self.layer2(x)   # 第二层线性变换，输出logits
         return x

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

# # 初始化模型
 model = MLP()
 model = model.to(device)  # 将模型移至GPU（如果可用）

 criterion = nn.CrossEntropyLoss()  # 交叉熵损失函数，适用于多分类问题
 optimizer = optim.Adam(model.parameters(), lr=0.001)  # Adam优化器

# # 5. 训练模型（记录每个 iteration 的损失）
 def train(model, train_loader, test_loader, criterion, optimizer, device, epochs):
     model.train()  # 设置为训练模式
    
     # 新增：记录每个 iteration 的损失
     all_iter_losses = []  # 存储所有 batch 的损失
iter_indices = []     # 存储 iteration 序号（从1开始）
    
     for epoch in range(epochs):
         running_loss = 0.0
         correct = 0
         total = 0
        
       for batch_idx, (data, target) in enumerate(train_loader):
            data, target = data.to(device), target.to(device)  # 移至GPU(如果可用)
            
             optimizer.zero_grad()  # 梯度清零
             output = model(data)  # 前向传播
             loss = criterion(output, target)  # 计算损失
             loss.backward()  # 反向传播
             optimizer.step()  # 更新参数
            
             # 记录当前 iteration 的损失（注意：这里直接使用单 batch 损失，而非累加平均）
             iter_loss = loss.item()#             
all_iter_losses.append(iter_loss)
             iter_indices.append(epoch * len(train_loader) + batch_idx + 1)  # iteration 序号从1开始
            
             # 统计准确率和损失（原逻辑保留，用于 epoch 级统计）
             running_loss += iter_loss
             _, predicted = output.max(1)
             total += target.size(0)
             correct += predicted.eq(target).sum().item()
            
             # 每100个批次打印一次训练信息（可选：同时打印单 batch 损失）
             if (batch_idx + 1) % 100 == 0:
                 print(f'Epoch: {epoch+1}/{epochs} | Batch: {batch_idx+1}/{len(train_loader)} '
                       f'| 单Batch损失: {iter_loss:.4f} | 累计平均损失: {running_loss/(batch_idx+1):.4f}')
        
         # 原 epoch 级逻辑（测试、打印 epoch 结果）不变
         epoch_train_loss = running_loss / len(train_loader)
         epoch_train_acc = 100. * correct / total
         epoch_test_loss, epoch_test_acc = test(model, test_loader, criterion, device)
        
         print(f'Epoch {epoch+1}/{epochs} 完成 | 训练准确率: {epoch_train_acc:.2f}% | 测试准确率: {epoch_test_acc:.2f}%')
    
     # 绘制所有 iteration 的损失曲线
     plot_iter_losses(all_iter_losses, iter_indices)
     # 保留原 epoch 级曲线（可选）
     # plot_metrics(train_losses, test_losses, train_accuracies, test_accuracies, epochs)
    
#     return epoch_test_acc  # 返回最终测试准确率

# # 6. 测试模型
 def test(model, test_loader, criterion, device):
     model.eval()  # 设置为评估模式
     test_loss = 0
     correct = 0
     total = 0
    
     with torch.no_grad():  # 不计算梯度，节省内存和计算资源
         for data, target in test_loader:
             data, target = data.to(device), target.to(device)
             output = model(data)
             test_loss += criterion(output, target).item()
            
             _, predicted = output.max(1)
             total += target.size(0)
             correct += predicted.eq(target).sum().item()
    
     avg_loss = test_loss / len(test_loader)
     accuracy = 100. * correct / total
     return avg_loss, accuracy  # 返回损失和准确率

# # 7.绘制每个 iteration 的损失曲线
 def plot_iter_losses(losses, indices):
     plt.figure(figsize=(10, 4))
     plt.plot(indices, losses, 'b-', alpha=0.7, label='Iteration Loss')
     plt.xlabel('Iteration（Batch序号）')
     plt.ylabel('损失值')
     plt.title('每个 Iteration 的训练损失')
    plt.legend()
     plt.grid(True)
     plt.tight_layout()
     plt.show()

# # 8. 执行训练和测试（设置 epochs=2 验证效果）
 epochs = 2  
 print("开始训练模型...")
 final_accuracy = train(model, train_loader, test_loader, criterion, optimizer, device, epochs)
 print(f"训练完成！最终测试准确率: {final_accuracy:.2f}%")