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

MoH: Multi-Head Attention as Mixture-of-Head Attention #39

Open

johndpope opened this issue Oct 19, 2024 · 2 comments

Owner

johndpope commented Oct 19, 2024 •

edited

Loading

https://arxiv.org/html/2410.11842
https://wandb.ai/snoozie/IMF/runs/3f2e9892?nw=nwusersnoozie
https://github.com/johndpope/IMF/tree/feat/moh

trained to 3000 steps - ~ 10 mins

vanilla training now to compare
https://wandb.ai/snoozie/IMF/runs/vc26civb

Owner Author

johndpope commented Oct 19, 2024

vanilla

VS MoH

pretty good - needs longer training - but this is an improvement.

Owner Author

johndpope commented Oct 20, 2024

fyi - @hologerry

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment