can‘t convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory

NoSuchKey

Guess you like

Origin blog.csdn.net/ResumeProject/article/details/121643345