Generalize SEEDS samplers for RF#8529
Merged
comfyanonymous merged 1 commit intoComfy-Org:masterfrom Jun 14, 2025
Merged
Conversation
Restore VP algorithm for RF and refactor noise_coeffs and half-logSNR calculations
adlerfaulkner
pushed a commit
to LucaLabsInc/ComfyUI
that referenced
this pull request
Oct 16, 2025
Restore VP algorithm for RF and refactor noise_coeffs and half-logSNR calculations
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
"Sigma" usually refers to two types: the plain sigma and the scaled one, such as (sigma / alpha). RF uses the former so that the SNR-related samplers currently do not capture the alpha signal.
By restoring the VP algorithm to include alpha and computing the actual half-logSNR, SDE samplers may become compatible with RF models. However, starting sigma at 1.0 leads to an invalid logSNR. To avoid this, the first sigma value needs to be offset during sampling. For other model types, this should not change their behavior because the computation of lambda remains the same.
Other changes are refactorings.
scale * some.expm1().neg().sqrt()seems more numerically stable.Tested with SD 3.5 medium, CFG = 3.0.
