Qui ci sono dei siti con informazioni importanti per sfruttare al meglio CUDA o per la riproducibilità:
- https://pytorch.org/docs/stable/data.html#data-loading-randomness
- https://pytorch.org/docs/stable/notes/randomness.html
- https://pytorch.org/docs/stable/notes/multiprocessing.html
- https://discuss.pytorch.org/t/dataloader-worker-failed/140518/2
- https://stackoverflow.com/questions/71261347/runtimeerror-dataloader-worker-exited-unexpectedly/74384157#74384157
- https://stackoverflow.com/questions/64837376/how-to-efficiently-run-multiple-pytorch-processes-models-at-once-traceback