Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

don't support both options fsdp and gradient_checkpointing . why? #52

Open
yanmj0601 opened this issue Nov 15, 2024 · 2 comments
Open

don't support both options fsdp and gradient_checkpointing . why? #52

yanmj0601 opened this issue Nov 15, 2024 · 2 comments

Comments

@yanmj0601
Copy link

assert not (train_args.fsdp and train_args.gradient_checkpointing), "currently, we don't support both options. open an issue for details."

why??
@tianweiy
Copy link
Owner

Using pure PyTorch's FSDP would offer better compatibility with gradient checkpointing and potentially much faster training, especially with newer PyTorch versions (unfortunately, we're currently stuck on PyTorch 2.0.1 due to Accelerate).

I plan to implement an Accelerate-free version in the future. However, I'm currently occupied with interviews, so I don't have a clear timeline for this yet :(

@yanmj0601
Copy link
Author

ok, thanks

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@tianweiy @yanmj0601