在这篇技术博客中,我们将使用 PyTorch 框架构建一个卷积神经网络(CNN),以识别 MNIST 数据集中的手写数字。我们将重点介绍如何在 GPU 上运行模型,以提高训练和推理的速度。

环境准备与库导入

确保您已经安装了 PyTorch 和 torchvision。如果您使用 Google Colab,可以直接在代码单元中运行以下命令:

pip install torch torchvision

接下来,导入必要的库:

import torch
import torchvision
import matplotlib.pyplot as plt
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim

设备设置与超参数

检查是否有可用的 GPU,并相应地设置设备:

device = torch.device('cpu')
if torch.cuda.is_available():
    device = torch.device('cuda')
    print("Using CUDA!")

设置超参数,包括训练的轮数、批量大小、学习率等:

n_epochs = 3
batch_size_train = 64
batch_size_test = 1000
learning_rate = 0.01
momentum = 0.5
log_interval = 10

random_seed = 1
torch.manual_seed(random_seed)

数据加载与可视化

使用 torchvision 加载 MNIST 数据集,并进行归一化处理:

train_loader = torch.utils.data.DataLoader(
    torchvision.datasets.MNIST(root='./MNIST', train=True, download=True,
                                transform=torchvision.transforms.Compose([
                                    torchvision.transforms.ToTensor(),
                                    torchvision.transforms.Normalize((0.1307,), (0.3081,))
                                ])),
    batch_size=batch_size_train, shuffle=True)

test_loader = torch.utils.data.DataLoader(
    torchvision.datasets.MNIST(root='./MNIST', train=False, download=True,
                                transform=torchvision.transforms.Compose([
                                    torchvision.transforms.ToTensor(),
                                    torchvision.transforms.Normalize((0.1307,), (0.3081,))
                                ])),
    batch_size=batch_size_test, shuffle=True)

可视化一些测试数据,以便更好地理解数据集:

examples = enumerate(test_loader)
batch_idx, (example_data, example_targets) = next(examples)

fig = plt.figure()
for i in range(6):
    plt.subplot(2, 3, i + 1)
    plt.tight_layout()
    plt.imshow(example_data[i][0], cmap='gray', interpolation='none')
    plt.title("Ground Truth: {}".format(example_targets[i]))
    plt.xticks([])
    plt.yticks([])
plt.show()

模型定义与训练

定义一个简单的卷积神经网络:

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)

设置优化器和损失函数,并定义训练和测试函数:

network = Net().to(device)
optimizer = optim.SGD(network.parameters(), lr=learning_rate, momentum=momentum)

train_losses = []
train_counter = []
test_losses = []
test_counter = [i * len(train_loader.dataset) for i in range(n_epochs + 1)]

def train(epoch):
    network.train()
    for batch_idx, (data, target) in enumerate(train_loader):
        data, target = data.to(device), target.to(device)
        optimizer.zero_grad()
        output = network(data)
        loss = F.nll_loss(output, target)
        loss.backward()
        optimizer.step()
        if batch_idx % 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()))
            train_losses.append(loss.item())
            train_counter.append((batch_idx * 64) + ((epoch - 1) * len(train_loader.dataset)))

def test():
    network.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 = network(data)
            test_loss += F.nll_loss(output, target, reduction='sum').item()
            pred = output.data.max(1, keepdim=True)[1]
            correct += pred.eq(target.data.view_as(pred)).sum()
    test_loss /= len(test_loader.dataset)
    test_losses.append(test_loss)
    print('\nTest set: Avg. loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)\n'.format(
        test_loss, correct, len(test_loader.dataset),
        100. * correct / len(test_loader.dataset)))

进行模型训练和测试:

for epoch in range(1, n_epochs + 1):
    train(epoch)
    test()

可视化训练过程与模型预测

可视化训练和测试损失:

fig = plt.figure()
plt.plot(train_counter, train_losses, color='blue')
plt.scatter(test_counter, test_losses, color='red')
plt.legend(['Train Loss', 'Test Loss'], loc='upper right')
plt.xlabel('number of training examples seen')
plt.ylabel('negative log likelihood loss')
plt.show()

使用训练好的模型进行预测:

with torch.no_grad():
    example_data_device = example_data.to(device)
    output = network(example_data_device)
    output = output.cpu()

fig = plt.figure()
for i in range(6):
    plt.subplot(2, 3, i + 1)
    plt.tight_layout()
    plt.imshow(example_data[i][0], cmap='gray', interpolation='none')
    plt.title("Prediction: {}".format(output.data.max(1, keepdim=True)[1][i].item()))
    plt.xticks([])
    plt.yticks([])
plt.show()

继续训练模型

如果需要继续训练模型,可以加载之前保存的模型和优化器状态:

continued_network = Net().to(device)
continued_optimizer = optim.SGD(network.parameters(), lr=learning_rate, momentum=momentum)

network_state_dict = torch.load('./model.pth')
continued_network.load_state_dict(network_state_dict)

optimizer_state_dict = torch.load('./optimizer.pth')
continued_optimizer.load_state_dict(optimizer_state_dict)

for i in range(4, 6):
    test_counter.append(i * len(train_loader.dataset))
    train(i)
    test()

总结

在本文中,我们介绍了如何使用 PyTorch 构建一个简单的卷积神经网络,以识别 MNIST 数据集中的手写数字,并在 GPU 上运行以提高性能。希望这篇博客能帮助您更好地理解深度学习的基本概念和实践。

如需进一步学习,请参考 PyTorch 官方文档