一个简单的深度学习例子（CIFAR10数据集，训练与预测，迁移学习）

1.首先是写一个nn

from torch import nn
import torch
class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.model=nn.Sequential(
            nn.Conv2d(3,32,5,1,2),
            nn.MaxPool2d(2),
            nn.Conv2d(32,32, 5,1,2),
            nn.MaxPool2d(2),
            nn.Conv2d(32,64,5,1,2),
            nn.MaxPool2d(2),
            nn.Flatten(),
            nn.Linear(64*4*4,64),
            nn.Linear(64,10)
        )

    def forward(self, x):
        x = self.model(x)
        return x

2.第二部加载数据集，定义模型，进行训练，训练过之后会保存一个权重文件，你可以加载这个权重文件进行再训练，迁移学习。

import torch
import torchvision
import torchvision.transforms as transforms
import torch.nn as nn
import torch.optim as optim
#from model import Net
# 加载CIFAR-10数据集
trainset = torchvision.datasets.CIFAR10(root='./data', train=True,
                                        download=True, transform=transforms.ToTensor())
trainloader = torch.utils.data.DataLoader(trainset, batch_size=128)

testset = torchvision.datasets.CIFAR10(root='./data', train=False,
                                       download=True, transform=transforms.ToTensor())
testloader = torch.utils.data.DataLoader(testset, batch_size=128)
print(len(trainset),len(testset))
# 定义模型
class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.model=nn.Sequential(
            nn.Conv2d(3,32,5,1,2),
            nn.MaxPool2d(2),
            nn.Conv2d(32,32, 5,1,2),
            nn.MaxPool2d(2),
            nn.Conv2d(32,64,5,1,2),
            nn.MaxPool2d(2),
            nn.Flatten(),
            nn.Linear(64*4*4,64),
            nn.Linear(64,10)
        )

    def forward(self, x):
        x = self.model(x)
        return x

net = Net()
#如果导入了Net，则不需要再写一遍模型
net=torch.load('model_net.pth')  #加载训练过的模型，如果没有这一行需要注释
# print(net)
if torch.cuda.is_available():
    net=net.cuda()
    print('1')
# 定义损失函数和优化器
criterion = nn.CrossEntropyLoss()
if torch.cuda.is_available():
    criterion=criterion.cuda()
optimizer = optim.SGD(net.parameters(), lr=0.001, momentum=0.9)

i=0
# 训练模型
for epoch in range(10):  # 多次循环遍历数据集
    running_loss = 0.0
    print(f'----{epoch+1}轮---')
    for data in trainloader:
        inputs, labels = data
        if torch.cuda.is_available():
            inputs=inputs.cuda()
            labels=labels.cuda()
        # print(len(inputs),len(labels))
        # 梯度清零
        optimizer.zero_grad()

        # 前向传播 + 反向传播 + 优化
        outputs = net(inputs)
        loss = criterion(outputs, labels)
        loss.backward()
        optimizer.step()
        i+=1
        print(f'训练第{i}次,loss={loss}')

print('Finished Training')

# 测试模型
correct = 0
total = 0
with torch.no_grad():
    for data in testloader:
        images, labels = data
        if torch.cuda.is_available():
            images=images.cuda()
            labels=labels.cuda()
        outputs = net(images)
        _, predicted = torch.max(outputs.data, 1)
        total += labels.size(0)
        correct += (predicted == labels).sum().item()

print('Accuracy of the network on the 10000 test images: %d %%' % (
    100 * correct / total))
# 保存模型
torch.save(net, 'model_net.pth')

print('Finished Training, model saved as model_net.pth')

3.训练出权重文件，开始进行预测（需要导入Net或者再写一遍模型），out的输出是数据集中的类别序号

import torchvision
import torch
from model import Net
from PIL import Image
image_path='dee.png'
img=Image.open(image_path)
print(img)
r, g, b, a = img.split()

# 创建一个新的RGB图像
rgb_image = Image.merge('RGB', (r, g, b))

transform=torchvision.transforms.Compose(
        [torchvision.transforms.Resize((32, 32)),
         torchvision.transforms.ToTensor()]
)
img = transform(rgb_image)
print(img.shape)
img = torch.unsqueeze(img, dim=0)
print(img.shape)
model=torch.load('model_net.pth')
print(model)
model.eval()
output=model(img)
print(output)
out=torch.argmax(output)
print(out)