Scaling Exponents Across Parameterizations and Optimizers #319

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

Closed

67372a opened this issue Jan 13, 2025 · 1 comment · Fixed by #321

Assignees

Labels

feature request

67372a commented Jan 13, 2025

Paper: https://arxiv.org/abs/2407.05872
Code: See C.5. ADAM-ATAN2 CODE CHANGE section in the paper, as well as https://github.com/LoganBooker/prodigy-plus-schedule-free/blob/1d2cfa2fe692a828d46a5a29b9667ec924961ac7/prodigyplus/core_optimiser.py#L412, where the clipping threshold is scaling for Prodigy purposes, but not strictly part of the technique.

67372a added the feature request label

67372a assigned kozistr

Owner

kozistr commented Jan 13, 2025

thanks!

kozistr mentioned this issue

[Feature] Implement OrthoGrad optimizer #321

Merged

1 task

kozistr closed this as completed in #321

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