feat: reduce CLIP memory usage with no embeddings #768
+52
−75
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
All handlers are constructed with the default clip_skip value, and it is always set during inference time, so there isn't much point in keeping it as a persistent attribute. Instead, just propagate the parameter value down from get_learned_condition*.
The CLIP weights need to be converted to f32 for textual inversions (fbd42b6), but that increases the amount of allocated VRAM even when embeddings aren't being used.
42f2fa6 to
489069c
Compare
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Outdated suggestions cannot be applied.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.
The CLIP weights need to be converted to f32 for textual inversions (fbd42b6, for #601), but that increases the amount of allocated VRAM even when embeddings aren't being used.
On a typical SDXL render on Vulkan, this change reduces peak VRAM usage around 190MB.
I'm including a refactor of the clip_skip handling, since this change would end up conflicting with it. Please let me know if I shall submit it separately instead.