给我写一个resnet50的pytorch程序

当然可以！以下是一个简单的ResNet50 PyTorch程序示例：

import torch.nn as nn
import torch.optim as optim
from torchvision import models

resnet50 = models.resnet50(pretrained=True)
for param in resnet50.parameters():
    param.requires_grad = False
num_ftrs = resnet50.fc.in_features
resnet50.fc = nn.Linear(num_ftrs, 10)  # 假设输出10个类别
criterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(resnet50.fc.parameters(), lr=0.001, momentum=0.9)

# 接下来是训练循环，可以根据具体数据集进行修改
for epoch in range(num_epochs):
    for inputs, labels in dataloader:
        optimizer.zero_grad()
        outputs = resnet50(inputs)
        loss = criterion(outputs, labels)
        loss.backward()
        optimizer.step()

希望这可以帮到您！