PyTorch如何在只有标量损失的情况下训练神经网络？

import torch
import torch.nn as nn
import torch.optim as optim

# 定义一个简单的神经网络
class SimpleNet(nn.Module):
    def __init__(self):
        super(SimpleNet, self).__init__()
        self.fc = nn.Linear(784, 10)

    def forward(self, x):
        x = x.view(-1, 784)
        x = self.fc(x)
        return x

# 创建模型实例
model = SimpleNet()

# 定义损失函数和优化器
criterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(model.parameters(), lr=0.01)

# 模拟数据
inputs = torch.randn(64, 784)
targets = torch.randint(0, 10, (64,))

# 训练循环
for epoch in range(10):  # 迭代10个epoch
    # 前向传播
    outputs = model(inputs)
    loss = criterion(outputs, targets)

    # 反向传播和优化
    optimizer.zero_grad()  # 清零梯度
    loss.backward()        # 计算梯度
    optimizer.step()       # 更新参数

    print(f'Epoch {epoch+1}, Loss: {loss.item()}')