Are you using multiprocessing (multiple workers in your DataLoader
) and CUDA tensors?
If so, this post might be the answer.
Are you using multiprocessing (multiple workers in your DataLoader
) and CUDA tensors?
If so, this post might be the answer.