Pytorch学习系列（5）：卷积神经网络（Convolutional Neural Network）

 

导包

# 包
import torch
import torch.nn as nn
import torchvision
import torchvision.transforms as transforms

 

超参数设置

# 设备配置
# 有cuda就用cuda
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

# 超参数设置
input_size = 784
hidden_size = 500
num_classes = 10
num_epochs = 5
batch_size = 100
learning_rate = 0.001

 

MINIST 数据集加载

train_dataset = torchvision.datasets.MNIST(root='./data/minist',
                                           train=True,
                                           transform=transforms.ToTensor(),
                                           download=True)

test_dataset = torchvision.datasets.MNIST(root='./data/minist',
                                          train=False,
                                          transform=transforms.ToTensor())
# 数据加载器（data loader）
train_loader = torch.utils.data.DataLoader(dataset=train_dataset,
                                           batch_size=batch_size,
                                           shuffle=True)

test_loader = torch.utils.data.DataLoader(dataset=test_dataset,
                                          batch_size=batch_size,
                                          shuffle=False)

 

自定义 卷积神经网络

# 搭建卷积神经网络模型
# 两个卷积层
class ConvNet(nn.Module):
    def __init__(self, num_classes=10):
        super(ConvNet, self).__init__()
        self.layer1 = nn.Sequential(
            # 卷积层计算
            nn.Conv2d(1, 16, kernel_size=5, stride=1, padding=2),
            #  批归一化
            nn.BatchNorm2d(16),
            #ReLU激活函数
            nn.ReLU(),
            # 池化层：最大池化
            nn.MaxPool2d(kernel_size=2, stride=2))
        self.layer2 = nn.Sequential(
            nn.Conv2d(16, 32, kernel_size=5, stride=1, padding=2),
            nn.BatchNorm2d(32),
            nn.ReLU(),
            nn.MaxPool2d(kernel_size=2, stride=2))
        self.fc = nn.Linear(7*7*32, num_classes)

    # 定义前向传播顺序
    def forward(self, x):
        out = self.layer1(x)
        out = self.layer2(out)
        out = out.reshape(out.size(0), -1)
        out = self.fc(out)
        return out


# 实例化一个模型，并迁移至gpu
model = ConvNet(num_classes).to(device)

# 定义损失函数和优化器
criterion = nn.CrossEntropyLoss()
optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)

 

训练模型

total_step = len(train_loader)
for epoch in range(num_epochs):
    for i, (images, labels) in enumerate(train_loader):
        # 注意模型在GPU中，数据也要搬到GPU中
        images = images.to(device)
        labels = labels.to(device)

        # 前向传播
        outputs = model(images)
        loss = criterion(outputs, labels)

        # 反向传播和优化
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        if (i+1) % 100 == 0:
            print ('Epoch [{}/{}], Step [{}/{}], Loss: {:.4f}' 
                   .format(epoch+1, num_epochs, i+1, total_step, loss.item()))

Epoch [1/5], Step [100/600], Loss: 0.0962
Epoch [1/5], Step [200/600], Loss: 0.0651
Epoch [1/5], Step [300/600], Loss: 0.0299
Epoch [1/5], Step [400/600], Loss: 0.1865
Epoch [1/5], Step [500/600], Loss: 0.0479
Epoch [1/5], Step [600/600], Loss: 0.1068
Epoch [2/5], Step [100/600], Loss: 0.0208
Epoch [2/5], Step [200/600], Loss: 0.0552
Epoch [2/5], Step [300/600], Loss: 0.0692
Epoch [2/5], Step [400/600], Loss: 0.0148
Epoch [2/5], Step [500/600], Loss: 0.0457
Epoch [2/5], Step [600/600], Loss: 0.1268
Epoch [3/5], Step [100/600], Loss: 0.0358
Epoch [3/5], Step [200/600], Loss: 0.0237
Epoch [3/5], Step [300/600], Loss: 0.0158
Epoch [3/5], Step [400/600], Loss: 0.0140
Epoch [3/5], Step [500/600], Loss: 0.0057
Epoch [3/5], Step [600/600], Loss: 0.0032
Epoch [4/5], Step [100/600], Loss: 0.0061
Epoch [4/5], Step [200/600], Loss: 0.0359
Epoch [4/5], Step [300/600], Loss: 0.1519
Epoch [4/5], Step [400/600], Loss: 0.0326
Epoch [4/5], Step [500/600], Loss: 0.0390
Epoch [4/5], Step [600/600], Loss: 0.0707
Epoch [5/5], Step [100/600], Loss: 0.0118
Epoch [5/5], Step [200/600], Loss: 0.0223
Epoch [5/5], Step [300/600], Loss: 0.0230
Epoch [5/5], Step [400/600], Loss: 0.0024
Epoch [5/5], Step [500/600], Loss: 0.0026
Epoch [5/5], Step [600/600], Loss: 0.0097

 

测试模型

# 切换成评估测试模式
# 这是因为在测试时，与训练时的dropout和batch normalization的操作是不同的
model.eval()
# 节省计算资源，不去计算梯度
with torch.no_grad():
    correct = 0
    total = 0
    for images, labels in test_loader:
        images = images.to(device)
        labels = labels.to(device)
        outputs = model(images)
        _, predicted = torch.max(outputs.data, 1)
        total += labels.size(0)
        correct += (predicted == labels).sum().item()

    print('Test Accuracy of the model on the 10000 test images: {} %'.format(100 * correct / total))

Test Accuracy of the model on the 10000 test images: 98.88 %

 

保存模型

#  保存模型
torch.save(model.state_dict(), 'model.ckpt')