PyTorch深度学习实践（11. Advanced CNN-2）

 源自课程：《PyTorch深度学习实践》完结合集

Chapter11 卷积神经网络（高级篇）

"""
Advanced CNN:resnet
"""
import torch
import torch.nn.functional as F
from torch.optim import optimizer
from torch.utils.data import DataLoader
from torchvision import datasets
from torchvision import transforms


# resnet常用模块搭建
class ResnetBlock(torch.nn.Module):
    def __init__(self, channels):
        super(ResnetBlock, self).__init__()
        self.channels = channels
        # 因为resnet要求通道数一致,因此只有channel即可
        self.conv1 = torch.nn.Conv2d(channels, channels, kernel_size=3, padding=1)
        self.conv2 = torch.nn.Conv2d(channels, channels, kernel_size=3, padding=1)

    def forward(self, x):
        # 一次卷积没有用,卷积的本质还是线性运算
        # 两次相同的卷积,第一次结束后激活
        y = F.relu(self.conv1(x))
        y = self.conv2(x)

        return F.relu(x + y)


# resnet网络具体的网络结构搭建
class Net(torch.nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.conv1 = torch.nn.Conv2d(1, 16, kernel_size=5)
        self.conv2 = torch.nn.Conv2d(16, 32, kernel_size=5)

        self.resnb1 = ResnetBlock(16)
        self.resnb2 = ResnetBlock(32)

        self.mp = torch.nn.MaxPool2d(2)
        self.fc = torch.nn.Linear(512, 10)

    def forward(self, x):
        in_size = x.size(0)

        x = F.relu(self.mp(self.conv1(x)))
        x = self.resnb1(x)

        x = F.relu(self.mp(self.conv2(x)))
        x = self.resnb2(x)

        x = x.view(in_size, -1)
        x = self.fc(x)
        return x


# 实例化模型,如果需要迁移到GPU上,则要使用to(device)语句
# 一共要将"模型  训练数据  测试数据"转换到GPU上
model = Net()
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
model.to(device)

# 设置batch_size大小，以及使用transform对样本数据进行标准化
batch_size = 64
transform = transforms.Compose([
    transforms.ToTensor(),
    transforms.Normalize((0.1307,), (0.3081,))
])

# 加载下载公开数据集的训练、测试数据（主要区别是Train=True/False）
# 注意：创建Dataset类的子类用于读取数据，而datasets则是加载下载公开数据集
train_dataset = datasets.MNIST(root="New_program//MNIST",
                               train=True,
                               download=True,
                               transform=transform)

# 训练集中的数据一般shuffle=True，而测试集中为False
train_loader = DataLoader(train_dataset,
                          shuffle=True,
                          batch_size=batch_size,
                          num_workers=2)

test_dataset = datasets.MNIST(root="New_program//MNIST",
                              train=False,
                              download=True,
                              transform=transform)

test_loader = DataLoader(test_dataset,
                         shuffle=False,
                         batch_size=batch_size,
                         num_workers=2)

# 构建损失函数和优化器，此处使用的是交叉熵损失，因此网络最后一层无需激活函数
criterion = torch.nn.CrossEntropyLoss()
# SGD带冲量，旨在帮助避免陷入局部最优点
optimizer = torch.optim.SGD(model.parameters(), lr=0.01, momentum=0.5)


# 开始进行迭代训练
def train(epoch):
    running_loss = 0
    for batch_idx, data in enumerate(train_loader, 0):
        # 正向传播、计算损失
        inputs, target = data
        inputs, target = inputs.to(device), target.to(device)
        outputs = model.forward(inputs)

        loss = criterion(outputs, target)

        # 梯度清零、反向传播、参数更新
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        running_loss = running_loss + loss.item()

        # 以100次作为一个iteration
        if batch_idx % 100 == 99:
            print('[%d, %5d] loss: %.3f' % (epoch + 1, batch_idx + 1, running_loss / 100))
            running_loss = 0.0


# 进行模型测试
def test():
    correct = 0
    total = 0
    # 测试时无需对梯度进行跟踪
    with torch.no_grad():
        # 测试时不需要batch_idx，单个样本测试即可
        for data in test_loader:
            images, labels = data
            images, labels = images.to(device), labels.to(device)
            outputs = model(images)
            _, predicted = torch.max(outputs.data, dim=1)
            total += labels.size(0)
            correct += (predicted == labels).sum().item()
            print('Accuracy on test set: %.3f %%' % (100 * correct / total))


if __name__ == '__main__':
    for epoch in range(10):
        train(epoch)
        test()