【pytorch学习笔记03】pytorch完整模型训练套路

B站我是土堆视频学习笔记，链接：https://www.bilibili.com/video/BV1hE411t7RN/?spm_id_from=333.999.0.0

1. 准备数据集

train_data = torchvision.datasets.CIFAR10(root='./dataset',
                                          train=True,
                                          transform=torchvision.transforms.ToTensor(),
                                          download=True)
test_data = torchvision.datasets.CIFAR10(root='./dataset',
                                         train=False,
                                         transform=torchvision.transforms.ToTensor(),
                                         download=True)

# length
train_data_size = len(train_data)
test_data_size = len(test_data)
print(f'训练数据集的长度为：{train_data_size}')
print(f'测试数据集的长度为：{test_data_size}')

2. 加载数据集

train_dataloader = DataLoader(train_data, batch_size=64)
test_dataloader = DataLoader(test_data, batch_size=64)

3. 构建模型

class MyModel(nn.Module):
    def __init__(self):
        super(MyModel, self).__init__()
        self.model = nn.Sequential(
            nn.Conv2d(3, 32, 5, 1, 2),
            nn.MaxPool2d(2),
            nn.Conv2d(32, 32, 5, 1, 2),
            nn.MaxPool2d(2),
            nn.Conv2d(32, 64, 5, 1, 2),
            nn.MaxPool2d(2),
            nn.Flatten(),
            nn.Linear(64 * 4 * 4, 64),
            nn.Linear(64, 10)
        )

    def forward(self, x):
        x = self.model(x)
        return x

# 测试
if __name__ == '__main__':
    model = MyModel()
    input = torch.ones(64, 3, 32, 32)
    output = model(input)
    print(output.shape)

实例化

model = MyModel()

4. 定义损失函数、优化器等

# 创建损失函数
loss_func = nn.CrossEntropyLoss()  # 这里是一个多分类问题

# 定义优化器
learning_rate = 1e-2
optimizer = torch.optim.SGD(model.parameters(), lr=learning_rate)

# 设置训练网络的参数
total_train_step = 0  # 训练次数
total_test_step = 0  # 测试次数
epoch = 10  # 训练轮数

5. 训练+测试+保存模型

# 添加 tensorboard
writer = SummaryWriter('./logs_train')

# 训练
for i in range(epoch):
    print(f'--------第{i+1}轮训练开始---------')

    # 训练
    model.train()
    for data in train_dataloader:
        imgs, labels = data
        outputs = model(imgs)
        loss = loss_func(outputs, labels)
        # 优化器优化模型参数
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        total_train_step += 1
        if total_train_step % 100 ==0:
            print(f'step：{total_train_step}, loss：{loss.item()}')
            writer.add_scalar('train_loss', loss.item(), total_train_step)

    # 测试
    model.test()
    total_test_loss = 0
    total_accuracy = 0
    with torch.no_grad():  # 测试的时候不需要再调参
        for data in test_dataloader:
            imgs, labels = data
            outputs = model(imgs)
            loss = loss_func(outputs, labels)
            total_test_loss += loss.item()
            accuracy = (outputs.argmax(1) == labels).sum()
            total_accuracy += accuracy
    print(f'整体测试集上的loss：{total_test_loss}, 准确率：{total_accuracy/test_data_size}')
    writer.add_scalar('test_loss', total_test_loss, total_test_step)
    writer.add_scalar('test_accuracy', total_accuracy/test_data_size, total_test_step)
    total_test_step += 1

    # 保存模型参数
    torch.save(model, f'model_{i}.pth')
    print('模型已保存')

writer.close()

6. 注意

model.train()和eval()

在训练前和测试前不是必要添加这两个函数的

官方文档解释

只对某些模块有影响，例如Dropout, BatchNorm等。所以保险起见两个都加？