用预训练的densenet121模型训练cifar10数据集（可运行）

import torchvision.models as models
import ssl
ssl._create_default_https_context = ssl._create_unverified_context
import torch
import torch.nn as nn
import torchvision
import torchvision.transforms as transforms
import torch.optim as optim
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
num_classes = 10
num_epochs = 10
batch_size = 100
learning_rate = 0.001
model = models.densenet121(pretrained=True)
for param in model.parameters():
    param.requires_grad = True
model.fc = nn.Linear(512, 10)
model = model.to(device)
# Optimize only the classifier
transform = transforms.Compose(
    [transforms.ToTensor(),
     transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])
trainset = torchvision.datasets.CIFAR10(root='./data', train=True,
                                        download=True, transform=transform)
trainloader = torch.utils.data.DataLoader(trainset, batch_size=100,
                                          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=100,
                                         shuffle=False, num_workers=2)
classes = ('plane', 'car', 'bird', 'cat',
           'deer', 'dog', 'frog', 'horse', 'ship', 'truck')
criterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(model.parameters(), lr=1e-2, momentum=0.9)
"""train the model"""
total_step = len(trainloader)
for epoch in range(num_epochs):
        for i,(images,labels) in enumerate(trainloader):
                images, labels = images.to(device), labels.to(device)
                outputs = model(images)
                loss = criterion(outputs,labels)
                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()))
correct = 0
total = 0
with torch.no_grad():
    for data in testloader:
        images, labels = data
        images, labels = images.to(device), labels.to(device)
        outputs = model(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))
class_correct = list(0. for i in range(10))
class_total = list(0. for i in range(10))
with torch.no_grad():
    for data in testloader:
        images, labels = data
        images, labels = images.to(device), labels.to(device)
        outputs = model(images)
        _, predicted = torch.max(outputs, 1)
        c = (predicted == labels).squeeze()
        for i in range(4):
            label = labels[i]
            class_correct[label] += c[i].item()
            class_total[label] += 1
for i in range(10):
    print('Accuracy of %5s : %2d %%' % (\
    classes[i], 100 * class_correct[i] / class_total[i]))

 

为了演示，只训练了10轮，效果为55%
各个类的分类精度也给出了。