refactor: apg: normalize diff_norm calculation by tensor size#1620
Merged
Conversation
stduhpf
changed the title
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
I've noticed a small flaw in the APG paper (that was implemented in #593).
The norm threshold for the APG update uses the L2 norm of the whole latent image diff, but since this norm scales with the square root of the size, the effect of the norm threshold will depend on the image resolution and number of latent channels.
This PR re-normalizes the L2 norm so the effect of the norm_threshold always match the effect it would have on a standard SDXL 1024x1024 generation.
Related Issue / Discussion
N/A
Additional Information
Technically, this is no longer a proper implementation of the original APG paper, but it is an overall improvement.
It's also technically a breaking change, the effect of
apg_norm_thresholdis changed, except in the specific case of images that have exactly 128x128x4 elements in their latent representations.Checklist