sd: sync to master-607-fd1a279#2212
Merged
LostRuins merged 1 commit intoMay 17, 2026
Merged
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
HiDream worked out-of-the-box, after I rebuilt everything (I really wish we had a working Makefile...). This is the BF16 with offloading to CPU:
(I've tried making a kobold, but it kept adding antlers to it 🤷♀️)
Despite the model size, it's pretty fast. And just as a reminder for anyone who wishes to try it: upstream still has issues with high contrast / noise, and the model can be very sensitive to quantization, especially on Vulkan.
Apart from HiDream, we have sampling code refactor that shouldn't change anything on our side.