04-softmax-linear-regression-concise.py

import sys

import torch
from d2l import torch as d2l
from torch import nn

## 读取小批量数据
batch_size = 256
train_iter, test_iter = d2l.load_data_fashion_mnist(batch_size)
print(len(train_iter))  # train_iter的长度是235；说明数据被分成了234组大小为256的数据加上最后一组大小不足256的数据
print('11111111')
for X, y in train_iter:
    print(X, y)
    break  # 尝试打印第一组X, y的形状：torch.Size([256, 1, 28, 28])  torch.Size([256])

# 定义模型
net = nn.Sequential(nn.Flatten(), nn.Linear(784, 10))


# 初始化参数
def init_weight(m):
    if type(m) == nn.Linear:
        nn.init.normal_(m.weight, std=0.01)


net.apply(init_weight)

# 定义损失函数
loss = nn.CrossEntropyLoss()

# 定义优化算法
optimizer = torch.optim.SGD(net.parameters(), lr=0.1)

# 训练模型
num_epochs = 5


def accuracy(y_hat, y):
    return (y_hat.argmax(dim=1) == y).float().mean().item()


# 计算这个训练集的准确率
def evaluate_accuracy(data_iter, net):
    acc_sum, n = 0.0, 0
    for X, y in data_iter:
        acc_sum += (net(X).argmax(dim=1) == y).float().sum().item()
        n += y.shape[0]
    return acc_sum / n


def train_ch3(net, train_iter, test_iter, loss, num_epochs, batch_size,
              params=None, lr=None, optimizer=None):
    for epoch in range(num_epochs):
        train_l_sum, train_acc_sum, n = 0.0, 0.0, 0
        for X, y in train_iter:
            y_hat = net(X)
            l = loss(y_hat, y).sum()

            # 梯度清零
            if params is not None and params[0].grad is not None:
                for param in params:
                    param.grad.data.zero_()

            l.backward()
            # 执行优化方法
            if optimizer is not None:
                optimizer.step()
            else:
                d2l.sgd(params, lr, batch_size)

            train_l_sum += l.item()
            train_acc_sum += (y_hat.argmax(dim=1) == y).sum().item()
            n += y.shape[0]
        test_acc = evaluate_accuracy(test_iter, net)
        print('epoch %d, loss %.4f, train acc %.3f, test acc %.3f'
              % (epoch + 1, train_l_sum / n, train_acc_sum / n, test_acc))


# 开始训练模型
train_ch3(net, train_iter, test_iter, loss, num_epochs, 256,
          None, 0.3, optimizer)