28.pdf

    1. import  torch
    2. import  torch.nn as nn
    3. import  torch.nn.functional as F
    4. import  torch.optim as optim
    5. from    torchvision import datasets, transforms
    8. batch_size=200
    9. learning_rate=0.01
    10. epochs=10
    12. train_loader = torch.utils.data.DataLoader(
    13.     datasets.MNIST('../data', train=True, download=True,
    14.                    transform=transforms.Compose([
    15.                        transforms.ToTensor(),
    16.                        transforms.Normalize((0.1307,), (0.3081,))
    17.                    ])),
    18.     batch_size=batch_size, shuffle=True)
    19. test_loader = torch.utils.data.DataLoader(
    20.     datasets.MNIST('../data', train=False, transform=transforms.Compose([
    21.         transforms.ToTensor(),
    22.         transforms.Normalize((0.1307,), (0.3081,))
    23.     ])),
    24.     batch_size=batch_size, shuffle=True)
    31. class MLP(nn.Module):
    33.     def __init__(self):
    34.         super(MLP, self).__init__()
    36.         self.model = nn.Sequential(
    37.             nn.Linear(784, 200),
    38.             nn.LeakyReLU(inplace=True),
    39.             nn.Linear(200, 200),
    40.             nn.LeakyReLU(inplace=True),
    41.             nn.Linear(200, 10),
    42.             nn.LeakyReLU(inplace=True),
    43.         )
    45.     def forward(self, x):
    46.         x = self.model(x)
    48.         return x
    50. device = torch.device('cuda:0')
    51. net = MLP().to(device)
    52. optimizer = optim.SGD(net.parameters(), lr=learning_rate)
    53. criteon = nn.CrossEntropyLoss().to(device)
    55. for epoch in range(epochs):
    57.     for batch_idx, (data, target) in enumerate(train_loader):
    58.         data = data.view(-1, 28*28)
    59.         data, target = data.to(device), target.cuda()
    61.         logits = net(data)
    62.         loss = criteon(logits, target)
    64.         optimizer.zero_grad()
    65.         loss.backward()
    66.         # print(w1.grad.norm(), w2.grad.norm())
    67.         optimizer.step()
    69.         if batch_idx % 100 == 0:
    70.             print('Train Epoch: {} [{}/{} ({:.0f}%)]\tLoss: {:.6f}'.format(
    71.                 epoch, batch_idx * len(data), len(train_loader.dataset),
    72.                        100. * batch_idx / len(train_loader), loss.item()))
    75.     test_loss = 0
    76.     correct = 0
    77.     for data, target in test_loader:
    78.         data = data.view(-1, 28 * 28)
    79.         data, target = data.to(device), target.cuda()
    80.         logits = net(data)
    81.         test_loss += criteon(logits, target).item()
    83.         pred = logits.data.max(1)[1]
    84.         correct += pred.eq(target.data).sum()
    86.     test_loss /= len(test_loader.dataset)
    87.     print('\nTest set: Average loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)\n'.format(
    88.         test_loss, correct, len(test_loader.dataset),
    89.         100. * correct / len(test_loader.dataset)))