pytorch实现mnist检测

import torch
import torch.nn as nn
import torch.nn.functional as fun
import argparse
from torchvision import datasets, transforms
import torch.optim as optim


class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.conv1 = nn.Conv2d(1, 20, 5, 1)
        self.conv2 = nn.Conv2d(20, 50, 5, 1)
        self.fc1 = nn.Linear(4 * 4 * 50, 500)  # 图像经过卷及池化后的尺寸，再压缩
        self.fc2 = nn.Linear(500, 10)

    def forward(self, x):
        x = fun.relu(self.conv1(x))
        x = fun.max_pool2d(x, 2, 2)  # kernel_size, stride
        x = fun.relu(self.conv2(x))
        x = fun.max_pool2d(x, 2, 2)
        x = x.view(-1, 4 * 4 * 50)
        x = fun.relu(self.fc1(x))
        x = self.fc2(x)

        result = fun.log_softmax(x, dim=1)
        return result


def train(args, model, device, train_loader, optimizer, epoch):
    model.train()  # Sets the module in training mode.
    for batch_idx, (data, target) in enumerate(train_loader):
        # to_GPU
        data, target = data.to(device), target.to(device)
        # 去除积累的梯度
        optimizer.zero_grad()
        # predicted
        predicted = model(data)
        # loss+backward
        loss = fun.nll_loss(predicted, target)  # 负对数似然损失函数
        loss.backward()
        # 执行一次优化
        optimizer.step()

        # 打印记录
        if batch_idx % args.log_interval == 0:  # 记录间隔
            print('Train Epoch: {} [{}/{} ({:.0f}%)]\tLoss: {:.6f}'.format(
                epoch, batch_idx * len(data), len(train_loader.dataset),
                       100. * batch_idx / len(train_loader), loss.item()))


def test(args, model, device, test_loader):
    model.eval()
    test_loss = 0
    correct = 0

    with torch.no_grad():  # 不需要计算梯度,也不会进行反向传播
        for data, target in test_loader:
            data, target = data.to(device), target.to(device)
            output = model(data)
            test_loss += fun.nll_loss(output, target,
                                      reduction='sum').item()  # sum up batch loss将整个batch的loss都加起来，再取平均，表示这一次迭代的loss
            pred = output.argmax(dim=1, keepdim=True)  # 取最大概率所在项作为预测结果
            correct += pred.eq(target.view_as(pred)).sum().item()  # 整理target与pred的维度相同，与pred对比，累计本次迭代所有正确个数
    test_loss /= len(test_loader.dataset)  # test_loader装载着一个数据集

    print('\nTest set: Average loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)\n'.format(
        test_loss, correct, len(test_loader.dataset),
        100. * correct / len(test_loader.dataset)))


def argument_parser():
    # training settings
    # metavar:为帮助消息中的可选参数提供了不同的名称
    parse = argparse.ArgumentParser(description='Pytorch MNIST Example')
    parse.add_argument('--batch-size', type=int, default=64, metavar='N',
                       help='input batch size for training (defaule:64)')
    parse.add_argument('--test-batch-size', type=int, default=1000)
    parse.add_argument('--epochs', type=int, default=10)
    parse.add_argument('--lr', type=int, default=0.01)
    parse.add_argument('--momentum', type=float, default=0.5)
    parse.add_argument('--no-cuda', action='store_true', default=False)
    parse.add_argument('--seed', type=int, default=1)
    parse.add_argument('--log-interval', type=int, default=1000, metavar='N',
                       help='how many batches to wait before logging training status')
    parse.add_argument('--save-model', action='store_true', default=True)

    args = parse.parse_args()

    return args


def run():
    # 输入参数
    args = argument_parser()
    # GPU
    use_cuda = not args.no_cuda and torch.cuda.is_available()
    # random seed
    torch.manual_seed(args.seed)  # 生成随机数种子，以使得结果是确定的
    # device
    device = torch.device('cuda' if use_cuda else 'cpu')

    kwargs = {'num_workers': 1, 'pin_memory': True} if use_cuda else {}
    """加载数据。组合数据集和采样器，提供数据上的单或多进程迭代器
    参数：
    dataset：Dataset类型，从其中加载数据
    batch_size：int，可选。每个batch加载多少样本
    shuffle：bool，可选。为True时表示每个epoch都对数据进行洗牌
    sampler：Sampler，可选。从数据集中采样样本的方法。
    num_workers：int，可选。加载数据时使用多少子进程。默认值为0，表示在主进程中加载数据。
    collate_fn：callable，可选。
    pin_memory：bool，可选
    drop_last：bool，可选。True表示如果最后剩下不完全的batch,丢弃。False表示不丢弃。
    """
    # data_loader
    train_set = datasets.MNIST('./data', train=True, download=True, transform=transforms.Compose(
        [transforms.ToTensor(), transforms.Normalize((0.1307,), (0.3081,))]))
    test_set = datasets.MNIST('./data', train=False, transform=transforms.Compose(
        [transforms.ToTensor(), transforms.Normalize((0.1307,), (0.3081,))]))
    train_loader = torch.utils.data.DataLoader(train_set, batch_size=args.batch_size,
                                               shuffle=True, **kwargs)
    test_loader = torch.utils.data.DataLoader(test_set, batch_size=args.test_batch_size,
                                              shuffle=True, **kwargs)
    # 实例化
    model = Net().to(device)
    # 定义优化器
    optimizer = optim.SGD(model.parameters(), lr=args.lr, momentum=args.momentum)

    # train
    for epoch in range(1, args.epochs + 1):
        train(args, model, device, train_loader, optimizer, epoch)
        test(args, model, device, test_loader)
    # model save
    if (args.save_model):
        torch.save(model.state_dict(), "./mnist_cnn.pt")


if __name__ == '__main__':
    run()