train不过

[heham]

win下操作的，python版本3.9.10，安装库都顺利，到了train就不过，看着像是显卡加速的问题，我是amd的apu，要怎么不使用cuda只使用cpu？
raise AssertionError("Torch not compiled with CUDA enabled")
AssertionError: Torch not compiled with CUDA enabled

[heham]

尝试修改 device = torch.device("cuda")为device = torch.device("cpu")之后出现RuntimeError: expected scalar type Long but found Int，loss = criterion(outputs, labels)

[heham]

尝试修改 device = torch.device("cuda")为device = torch.device("cpu")之后出现RuntimeError: expected scalar type Long but found Int，loss = criterion(outputs, labels)

169行上面增加labels = labels.to(torch.long)

[heham]

不知道怎么搞了，改了之后训练是通过了，但是识别不正确

[Blacksnow123]

改下evaluate.py里面的cuda改成cpu就好了