PyTorch——AlexNet实现（附完整代码）

参考链接

1. https://tangshusen.me/Dive-into-DL-PyTorch/#/chapter05_CNN/5.6_alexnet

AlexNet网络结构

AlexNet v.s. LeNet

第一，与相对较小的LeNet相比，AlexNet包含8层变换，其中有5层卷积和2层全连接隐藏层，以及1个全连接输出层。

lexNet第一层中的卷积窗口形状是11×11。因为ImageNet中绝大多数图像的高和宽均比MNIST图像的高和宽大10倍以上，ImageNet图像的物体占用更多的像素，所以需要更大的卷积窗口来捕获物体。第二层中的卷积窗口形状减小到5×5，之后全采用3×3。此外，第一、第二和第五个卷积层之后都使用了窗口形状为3×3、步幅为2的最大池化层。而且，AlexNet使用的卷积通道数也大于LeNet中的卷积通道数数十倍。

第二，AlexNet将sigmoid激活函数改成了更加简单的ReLU激活函数。一方面，ReLU激活函数的计算更简单，例如它并没有sigmoid激活函数中的求幂运算。另一方面，ReLU激活函数在不同的参数初始化方法下使模型更容易训练。这是由于当sigmoid激活函数输出极接近0或1时，这些区域的梯度几乎为0，从而造成反向传播无法继续更新部分模型参数；而ReLU激活函数在正区间的梯度恒为1。因此，若模型参数初始化不当，sigmoid函数可能在正区间得到几乎为0的梯度，从而令模型无法得到有效训练。

第三，AlexNet通过dropout来控制全连接层的模型复杂度。而LeNet并没有使用dropout。

第四，AlexNet引入了大量的图像增广，如翻转、裁剪和颜色变化，从而进一步扩大数据集来缓解过拟合。

加载数据

使用“Fashion-MNIST”数据集。读取数据的时候我们额外做了一步将图像高和宽扩大到AlexNet使用的图像高和宽224。这个可以通过torchvision.transforms.Resize实例来实现。也就是说，我们在ToTensor实例前使用Resize实例，然后使用Compose实例来将这两个变换串联。

def load_data_fashion_mnist(batch_size, resize=None, root='~/Datasets/FashionMNIST'):
    if sys.platform.startswith('win'):
        num_workers = 0
    else:
        num_workers = 4
    trans = []
    if resize:
        trans.append(torchvision.transforms.Resize(size=resize))
    trans.append(torchvision.transforms.ToTensor())

    transform = torchvision.transforms.Compose(trans)
    mnist_train = torchvision.datasets.FashionMNIST(root=root, train=True, download=True, transform=transform)
    mnist_test = torchvision.datasets.FashionMNIST(root=root, train=False, download=True, transform=transform)

    train_iter = torch.utils.data.DataLoader(mnist_train, batch_size=batch_size, shuffle=True, num_workers=num_workers)
    test_iter = torch.utils.data.DataLoader(mnist_test, batch_size=batch_size, shuffle=False, num_workers=num_workers)

    return train_iter, test_iter

batch_size = 128
train_iter, test_iter = load_data_fashion_mnist(batch_size, resize=224)

构建模型

class AlexNet(nn.Module):
    def __init__(self):
        super(AlexNet, self).__init__()
        self.conv = nn.Sequential(
            nn.Conv2d(1, 96, 11, 4), # in_channels, out_channels, kernel_size, stride, padding
            nn.ReLU(),
            nn.MaxPool2d(3, 2), # kernel_size, stride
            # 减小卷积窗口，使用填充为2来使得输入与输出的高和宽一致，且增大输出通道数
            nn.Conv2d(96, 256, 5, 1, 2),
            nn.ReLU(),
            nn.MaxPool2d(3, 2),
            # 连续3个卷积层，且使用更小的卷积窗口。除了最后的卷积层外，进一步增大了输出通道数。
            nn.Conv2d(256, 384, 3, 1, 1),
            nn.ReLU(),
            nn.Conv2d(384, 384, 3, 1, 1),
            nn.ReLU(),
            nn.Conv2d(384, 256, 3, 1, 1),
            nn.ReLU(),
            nn.MaxPool2d(3, 2)
        )
        self.fc = nn.Sequential(
            nn.Linear(256*5*5, 4096),
            nn.ReLU(),
            nn.Dropout(0.5),
            nn.Linear(4096, 4096),
            nn.ReLU(),
            nn.Dropout(0.5),
            nn.Linear(4096, 10),
        )

    def forward(self, img):
        feature = self.conv(img)
        output = self.fc(feature.view(img.shape[0], -1))
        return output

损失函数

损失函数使用交叉熵损失。

loss = torch.nn.CrossEntropyLoss()

优化方法

优化方法使用Adam算法。

optimizer = torch.optim.Adam(net.parameters(), lr=lr)

完整代码

import time
import torch
from torch import nn, optim
import torchvision
import sys

device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

def load_data_fashion_mnist(batch_size, resize=None, root='~/Datasets/FashionMNIST'):
    if sys.platform.startswith('win'):
        num_workers = 0
    else:
        num_workers = 4
    trans = []
    if resize:
        trans.append(torchvision.transforms.Resize(size=resize))
    trans.append(torchvision.transforms.ToTensor())

    transform = torchvision.transforms.Compose(trans)
    mnist_train = torchvision.datasets.FashionMNIST(root=root, train=True, download=True, transform=transform)
    mnist_test = torchvision.datasets.FashionMNIST(root=root, train=False, download=True, transform=transform)

    train_iter = torch.utils.data.DataLoader(mnist_train, batch_size=batch_size, shuffle=True, num_workers=num_workers)
    test_iter = torch.utils.data.DataLoader(mnist_test, batch_size=batch_size, shuffle=False, num_workers=num_workers)

    return train_iter, test_iter

batch_size = 128
train_iter, test_iter = load_data_fashion_mnist(batch_size, resize=224)

class AlexNet(nn.Module):
    def __init__(self):
        super(AlexNet, self).__init__()
        self.conv = nn.Sequential(
            nn.Conv2d(1, 96, 11, 4), # in_channels, out_channels, kernel_size, stride, padding
            nn.ReLU(),
            nn.MaxPool2d(3, 2), # kernel_size, stride
            # 减小卷积窗口，使用填充为2来使得输入与输出的高和宽一致，且增大输出通道数
            nn.Conv2d(96, 256, 5, 1, 2),
            nn.ReLU(),
            nn.MaxPool2d(3, 2),
            # 连续3个卷积层，且使用更小的卷积窗口。除了最后的卷积层外，进一步增大了输出通道数。
            nn.Conv2d(256, 384, 3, 1, 1),
            nn.ReLU(),
            nn.Conv2d(384, 384, 3, 1, 1),
            nn.ReLU(),
            nn.Conv2d(384, 256, 3, 1, 1),
            nn.ReLU(),
            nn.MaxPool2d(3, 2)
        )
        self.fc = nn.Sequential(
            nn.Linear(256*5*5, 4096),
            nn.ReLU(),
            nn.Dropout(0.5),
            nn.Linear(4096, 4096),
            nn.ReLU(),
            nn.Dropout(0.5),
            nn.Linear(4096, 10),
        )

    def forward(self, img):
        feature = self.conv(img)
        output = self.fc(feature.view(img.shape[0], -1))
        return output

net = AlexNet()

def evaluate_accuracy(data_iter, net, device=None):
    if device is None and isinstance(net, torch.nn.Module):
        # 如果没指定device就使用net的device
        device = list(net.parameters())[0].device
    acc_sum, n = 0.0, 0
    with torch.no_grad():
        for X, y in data_iter:
            net.eval() # 评估模式, 这会关闭dropout
            acc_sum += (net(X.to(device)).argmax(dim=1) == y.to(device)).float().sum().cpu().item()
            net.train() # 改回训练模式
            n += y.shape[0]
    return acc_sum / n


def train(net, train_iter, test_iter, batch_size, optimizer, device, num_epochs):
    net = net.to(device)
    print("training on ", device)
    loss = torch.nn.CrossEntropyLoss()
    for epoch in range(num_epochs):
        train_l_sum, train_acc_sum, n, batch_count, start = 0.0, 0.0, 0, 0, time.time()
        for X, y in train_iter:
            X = X.to(device)
            y = y.to(device)
            y_hat = net(X)
            l = loss(y_hat, y)
            optimizer.zero_grad()
            l.backward()
            optimizer.step()
            train_l_sum += l.cpu().item()
            train_acc_sum += (y_hat.argmax(dim=1) == y).sum().cpu().item()
            n += y.shape[0]
            batch_count += 1
        test_acc = evaluate_accuracy(test_iter, net)
        print('epoch %d, loss %.4f, train acc %.3f, test acc %.3f, time %.1f sec'
              % (epoch + 1, train_l_sum / batch_count, train_acc_sum / n, test_acc, time.time() - start))

lr, num_epochs = 0.001, 5
optimizer = torch.optim.Adam(net.parameters(), lr=lr)
train(net, train_iter, test_iter, batch_size, optimizer, device, num_epochs)