RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0

在对保存好的tensor文件进行调用时，打印发现device=‘cuda:0’

tensor([[[4.0218e-05, 4.0218e-05, 4.0218e-05,  ..., 9.4699e-06,
          1.6517e-05, 1.6517e-05],
         [4.0218e-05, 4.0218e-05, 4.0218e-05,  ..., 9.4699e-06,
          1.6517e-05, 1.6517e-05],
         [4.0218e-05, 4.0218e-05, 4.0218e-05,  ..., 9.4699e-06,
          1.6517e-05, 1.6517e-05]]], device='cuda:0')

而在对数据进行处理时却报错，说明这些操作默认在cpu上运行。

channel_mean += image.mean(2).sum(0)

为保证一致性，需要将tensor数据也放在cpu中

device = torch.device("cpu")
image = image.to(device)