Fix Z-Image FP16 overflow via downscaling #11187
Open
+11
−3
Conversation
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
vanDuven
requested review from
Kosinkadink,
comfyanonymous and
guill
as code owners
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.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.
This PR improves FP16 stability for Z-Image by using scaling instead of clamping.
Because the tensor passes through
LinearandRMSNorm, the fp16 tensor can be scaled down to prevent overflow.The scale value(2^x) is based on testing.
No noticeable impact on inference speed.
Tested with: Z-Image and Lumina 2.
The
clamp_fp16function can be safely removed or stay just in case.workflow.json