Use this topic to ask your questions to Sylvain Gugger during his talk: Supercharge your PyTorch training loop with Accelerate
Can Accelarate help with hyperparameter optimization, same way Ray Tune does?
Can we do gradient accumulation with Accelerate
This is not provided out of the box right now, but it might be a future development.
Yes, you can use anything you usually do in your training loop
Thanks for the talk. It’s really a great tool. My question is:
Is Accelerate supported on Window platform? or just currently on Linux?