Multiprocessing using torch.multiprocessing

@Enumaris why didn’t you just use DDP (DistributedDataParallel — PyTorch 1.7.1 documentation)?