Add Lion optimizer and a "skip step" version of it #43
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
2015aroras
reviewed
Sep 6, 2024
| if self._grad_norms: | ||
| grad_norm_std = torch.std(torch.stack(self._grad_norms[:-1])) | ||
| return ( | ||
| self.latest_loss <= self.sigma_factor * loss_std |
There was a problem hiding this comment.
The Karpathy thing actually filters out when loss/grad norm is sigma_factor standard deviations above the mean (this does above 0). A number that is s standard deviations above the mean is said to have a "z-score" of s.
There was a problem hiding this comment.
Oh, duh, that's embarrassing after having spent 4 years in a stats PhD program
There was a problem hiding this comment.
Forgot the .abs() but just pushed a fix for that.
2015aroras
approved these changes
Sep 6, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
I don't really intend to use the Lion optimizer, this is more meant as a proof of concept for two things: