Hi all,
Is there a way to run multiple torch.optim routines in parallel or batch its data?
For context: Inside my DataLoader I have an expensive optimization loop which depends on gaussian noise. Running the optimization either on GPU + 1 worker or CPU + n workers is very slow.