pytorch实现mnist分类的示例讲解

torchvision包 包含了目前流行的数据集，模型结构和常用的图片转换工具。

torchvision.datasets中包含了以下数据集

MNIST
COCO（用于图像标注和目标检测）(Captioning and Detection)
LSUN Classification
ImageFolder
Imagenet-12
CIFAR10 and CIFAR100
STL10

torchvision.models

torchvision.models模块的 子模块中包含以下模型结构。
AlexNet
VGG
ResNet
SqueezeNet
DenseNet You can construct a model with random weights by calling its constructor:

pytorch torchvision transform

对PIL.Image进行变换

from __future__ import print_function
import argparse #Python 命令行解析工具
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torchvision import datasets, transforms

class Net(nn.Module):
 def __init__(self):
   super(Net, self).__init__()
   self.conv1 = nn.Conv2d(1, 10, kernel_size=5)
   self.conv2 = nn.Conv2d(10, 20, kernel_size=5)
   self.conv2_drop = nn.Dropout2d()
   self.fc1 = nn.Linear(320, 50)
   self.fc2 = nn.Linear(50, 10)

def forward(self, x):
   x = F.relu(F.max_pool2d(self.conv1(x), 2))
   x = F.relu(F.max_pool2d(self.conv2_drop(self.conv2(x)), 2))
   x = x.view(-1, 320)
   x = F.relu(self.fc1(x))
   x = F.dropout(x, training=self.training)
   x = self.fc2(x)
   return F.log_softmax(x, dim=1)

def train(args, 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 ％ args.log_interval == 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(args, 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, size_average=False).item() # sum up batch loss
     pred = output.max(1, keepdim=True)[1] # 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():
 # Training settings
 parser = argparse.ArgumentParser(description='PyTorch MNIST Example')
 parser.add_argument('--batch-size', type=int, default=64, metavar='N',
           help='input batch size for training (default: 64)')
 parser.add_argument('--test-batch-size', type=int, default=1000, metavar='N',
           help='input batch size for testing (default: 1000)')
 parser.add_argument('--epochs', type=int, default=10, metavar='N',
           help='number of epochs to train (default: 10)')
 parser.add_argument('--lr', type=float, default=0.01, metavar='LR',
           help='learning rate (default: 0.01)')
 parser.add_argument('--momentum', type=float, default=0.5, metavar='M',
           help='SGD momentum (default: 0.5)')
 parser.add_argument('--no-cuda', action='store_true', default=False,
           help='disables CUDA training')
 parser.add_argument('--seed', type=int, default=1, metavar='S',
           help='random seed (default: 1)')
 parser.add_argument('--log-interval', type=int, default=10, metavar='N',
           help='how many batches to wait before logging training status')
 args = parser.parse_args()
 use_cuda = not args.no_cuda and torch.cuda.is_available()

torch.manual_seed(args.seed)

device = torch.device("cuda" if use_cuda else "cpu")

kwargs = {'num_workers': 1, 'pin_memory': True} if use_cuda else {}
 train_loader = torch.utils.data.DataLoader(
   datasets.MNIST('../data', train=True, download=True,
           transform=transforms.Compose([
             transforms.ToTensor(),
             transforms.Normalize((0.1307,), (0.3081,))
           ])),
   batch_size=args.batch_size, shuffle=True, **kwargs)
 test_loader = torch.utils.data.DataLoader(
   datasets.MNIST('../data', train=False, transform=transforms.Compose([
             transforms.ToTensor(),
             transforms.Normalize((0.1307,), (0.3081,))
           ])),
   batch_size=args.test_batch_size, shuffle=True, **kwargs)

model = Net().to(device)
 optimizer = optim.SGD(model.parameters(), lr=args.lr, momentum=args.momentum)

for epoch in range(1, args.epochs + 1):
   train(args, model, device, train_loader, optimizer, epoch)
   test(args, model, device, test_loader)

if __name__ == '__main__':
 main()

来源：https://blog.csdn.net/KyrieHe/article/details/80516737