美文网首页
Pytorch_3:可执行的最小程序

Pytorch_3:可执行的最小程序

作者: 闪电侠悟空 | 来源:发表于2019-08-15 21:39 被阅读0次
    from __future__ import print_function
    import torch
    import torch.nn as nn
    import torch.nn.functional as F
    import torch.optim as optim
    from torchvision import datasets, transforms, models
    
    def train(model, device, train_loader, optimizer, epoch):
        model.train()
        for batch_idx, (data, target) in enumerate(train_loader):
            data, target = data.to(device), target.to(device)
            optimizer.zero_grad()
            output = model(data)
            loss = F.nll_loss(output, target)
            loss.backward()
            optimizer.step()
            if batch_idx % 100 == 0:
                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(model, device, test_loader):
        model.eval()
        test_loss = 0
        correct = 0
        with torch.no_grad():
            for data, target in test_loader:
                data, target = data.to(device), target.to(device)
                output = model(data)
                test_loss += F.nll_loss(output, target, reduction='sum').item()  # sum up batch loss
                pred = output.argmax(dim=1, keepdim=True)  # get the index of the max log-probability
                correct += pred.eq(target.view_as(pred)).sum().item()
    
        test_loss /= len(test_loader.dataset)
    
        print('\nTest set: Average loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)\n'.format(
            test_loss, correct, len(test_loader.dataset),
            100. * correct / len(test_loader.dataset)))
    
    
    def main():
        # 0. Settings
        torch.manual_seed(100)
        device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
        batch_size = 32
        epochs = 20
        lr = 0.01
        momentum = 0.90
    
        # 1. Dataset
        dataset_train = datasets.CIFAR10('../data', train=True, download=True,
                                         transform=transforms.Compose([
                                             transforms.ToTensor(),
                                             transforms.Normalize((0.1307,), (0.3081,))
                                         ]))
        dataset_test = datasets.CIFAR10('../data', train=False, download=True,
                                        transform=transforms.Compose([
                                            transforms.ToTensor(),
                                            transforms.Normalize((0.1307,), (0.3081,))
                                        ]))
        train_loader = torch.utils.data.DataLoader(dataset_train, batch_size=batch_size, shuffle=True, num_workers=4)
        test_loader = torch.utils.data.DataLoader(dataset_test, batch_size=batch_size, shuffle=True, num_workers=4)
    
        # 2. Model
        model = models.resnet18(pretrained=True, progress=True)  # 从模型库中调用ResNet18
        print(model)
        # num_ftrs = model.fc.in_features
        # model.fc = nn.Linear(num_ftrs, 10)  # 得到新的模型,更改fc层
        # model = model.to(device)
        # print(model)
    
        # 3. Optimizer
        optimizer = optim.SGD(model.parameters(), lr=lr, momentum=momentum)
    
        # 4. Training
        for epoch in range(epochs):
            train(model, device, train_loader, optimizer, epoch)
            test(model, device, test_loader)
            # save the current model
            if (epoch + 1) % 10 == 0:
                torch.save(model, "train_cnn_%3d.pt" % (epoch + 1))
    
    
    if __name__ == '__main__':
        main()
    
    

    相关文章

      网友评论

          本文标题:Pytorch_3:可执行的最小程序

          本文链接:https://www.haomeiwen.com/subject/pigtsctx.html