pytorch图像分类

这里写一个用pytorch图像分类实例，包含早停写法。其中的pytorchtools调用仓库

import torch
import torch.nn as nn
import torch.nn.functional as F
import torchvision
import torchvision.transforms as transforms
import torch.optim as optim
from pytorchtools import EarlyStopping
# cifar-10官方提供的数据集是用numpy array存储的
# 下面这个transform会把numpy array变成torch tensor，然后把rgb值归一到[0, 1]这个区间
patience = 20
early_stopping = EarlyStopping(patience, verbose=True)
class LeNet(nn.Module):
    # 一般在__init__中定义网络需要的操作算子，比如卷积、全连接算子等等
    def __init__(self):
        super(LeNet, self).__init__()
        # Conv2d的第一个参数是输入的channel数量，第二个是输出的channel数量，第三个是kernel size
        self.conv1 = nn.Conv2d(3, 6, 5)
        self.conv2 = nn.Conv2d(6, 16, 5)
        # 由于上一层有16个channel输出，每个feature map大小为5*5，所以全连接层的输入是16*5*5
        self.fc1 = nn.Linear(16*5*5, 120)
        self.fc2 = nn.Linear(120, 84)
        # 最终有10类，所以最后一个全连接层输出数量是10
        self.fc3 = nn.Linear(84, 10)
        self.pool = nn.MaxPool2d(2, 2)
    # forward这个函数定义了前向传播的运算，只需要像写普通的python算数运算那样就可以了
    def forward(self, x):
        x = F.relu(self.conv1(x))
        x = self.pool(x)
        x = F.relu(self.conv2(x))
        x = self.pool(x)
        # 下面这步把二维特征图变为一维，这样全连接层才能处理
        x = x.view(-1, 16*5*5)
        x = F.relu(self.fc1(x))
        x = F.relu(self.fc2(x))
        x = self.fc3(x)
        return x
transform = transforms.Compose(
    [transforms.ToTensor(),
     transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])

# 在构建数据集的时候指定transform，就会应用我们定义好的transform
# root是存储数据的文件夹，download=True指定如果数据不存在先下载数据
cifar_train = torchvision.datasets.CIFAR10(root='./data', train=True,
                                           download=True, transform=transform)
cifar_test = torchvision.datasets.CIFAR10(root='./data', train=False,
                                          transform=transform)
print(cifar_train)
print(cifar_test)
trainloader = torch.utils.data.DataLoader(cifar_train, batch_size=64, shuffle=True)
testloader = torch.utils.data.DataLoader(cifar_test, batch_size=10000, shuffle=True)
print(trainloader)
net = LeNet()
criterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(net.parameters(), lr=0.001, momentum=0.9)
for  data in (testloader):
    X_val,y_val=data
print("Start Training...")
for epoch in range(50):
    correct = 0
    total = 0
    # 我们用一个变量来记录每100个batch的平均loss
    loss100 = 0.0
    # 我们的dataloader派上了用场
    net.train()
    for i, data in enumerate(trainloader):
        inputs, labels = data
        inputs, labels = inputs, labels # 注意需要复制到GPU
        optimizer.zero_grad()
        outputs = net(inputs)
        loss = criterion(outputs, labels)
        loss.backward()
        optimizer.step()
        loss100 += loss.item()
        if i % 100 == 99:
            print('[Epoch %d, Batch %5d] loss: %.3f' %
                  (epoch + 1, i + 1, loss100 / 100))
            loss100 = 0.0
    net.eval()
    outputs = net(X_val)
    # 我们的网络输出的实际上是个概率分布，去最大概率的哪一项作为预测分类
    _, predicted = torch.max(outputs.data, 1)
    total += y_val.size(0)
    correct += (predicted == y_val).sum().item()
    valid_output = outputs
    valid_loss = criterion(valid_output, y_val)  # 注意这里的输入参数维度要符合要求，我这里为了简单，并未考虑这一点

    early_stopping(valid_loss, net)
    # 若满足 early stopping 要求
    if early_stopping.early_stop:
        print("Early stopping")
        # 结束模型训练
        break
    print('Accuracy of the network on the 10000 test images: %d %%' % (
            100 * correct / total))