import torch
import torch.nn as nn
import torch.optim as optim
import torchvision
import torchvision.transforms as transforms
import matplotlib.pyplot as plt
import numpy as np
数据预处理
transform = transforms.Compose([
transforms.ToTensor(), # 将图像转换为张量
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5)) # 标准化,使数据分布更适合模型训练
])
加载训练集和测试集
trainset = torchvision.datasets.CIFAR10(root='./data', train=True,
download=True, transform=transform)
trainloader = torch.utils.data.DataLoader(trainset, batch_size=4,
shuffle=True, num_workers=2)
testset = torchvision.datasets.CIFAR10(root='./data', train=False,
download=True, transform=transform)
testloader = torch.utils.data.DataLoader(testset, batch_size=4,
shuffle=False, num_workers=2)
类别标签
classes = ('plane', 'car', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck')
定义网络模型
class Net(nn.Module):
def init(self):
super(Net, self).init()
self.conv1 = nn.Conv2d(3, 6, 5) # 卷积层1,输入通道3,输出通道6,卷积核大小5
self.pool = nn.MaxPool2d(2, 2) # 最大池化层,池化核大小2,步长2
self.conv2 = nn.Conv2d(6, 16, 5) # 卷积层2,输入通道6,输出通道16,卷积核大小5
self.fc1 = nn.Linear(16 * 5 * 5, 120) # 全连接层1,输入特征数1655,输出120
self.fc2 = nn.Linear(120, 84) # 全连接层2,输入120,输出84
self.fc3 = nn.Linear(84, 10) # 全连接层3,输入84,输出10(对应10个类别)
def forward(self, x):x = self.pool(torch.relu(self.conv1(x))) # 卷积1 -> 激活 -> 池化x = self.pool(torch.relu(self.conv2(x))) # 卷积2 -> 激活 -> 池化x = x.view(-1, 16 * 5 * 5) # 展平,为全连接层做准备x = torch.relu(self.fc1(x)) # 全连接1 -> 激活x = torch.relu(self.fc2(x)) # 全连接2 -> 激活x = self.fc3(x) # 全连接3,输出类别得分return x
net = Net()
定义损失函数和优化器
criterion = nn.CrossEntropyLoss() # 交叉熵损失函数,适用于分类问题
optimizer = optim.SGD(net.parameters(), lr=0.001, momentum=0.9) # 随机梯度下降优化器,学习率0.001,动量0.9
训练网络
for epoch in range(2): # 训练2个epoch
running_loss = 0.0
for i, data in enumerate(trainloader, 0):
# 获取输入数据
inputs, labels = data
# 梯度清零
optimizer.zero_grad()
# 前向传播
outputs = net(inputs)
# 计算损失
loss = criterion(outputs, labels)
# 反向传播
loss.backward()
# 更新参数
optimizer.step()
# 打印训练信息
running_loss += loss.item()
if i % 2000 == 1999: # 每2000个mini - batch打印一次
print(f'[{epoch + 1}, {i + 1:5d}] loss: {running_loss / 2000:.3f}')
running_loss = 0.0
print('Finished Training')
保存模型
PATH = './cifar_net.pth'
torch.save(net.state_dict(), PATH)
测试网络
correct = 0
total = 0
with torch.no_grad(): # 测试时不需要计算梯度
for data in testloader:
images, labels = data
outputs = net(images)
_, predicted = torch.max(outputs.data, 1) # 获取预测的类别
total += labels.size(0)
correct += (predicted == labels).sum().item() # 统计正确预测的数量
print(f'Accuracy of the network on the 10000 test images: {100 * correct / total} %')