Fix diffusion GPU memory leak: reuse InferenceFunction#110
Merged
Conversation
The diffusion pipeline loaded a fresh InferenceFunction on every inference call (~30 per generation) as a workaround for a framework buffer caching bug that has since been fixed. The workaround caused GPU memory to accumulate across generations, leading to SIGABRT after ~20 images. Now reuses the stored function (matching how LLM engines work). Also wraps model loading and inference in do/catch to surface actionable errors instead of crashing on GPU memory exhaustion. Addresses apple#77.
kevchengcodes
approved these changes
Jul 16, 2026
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The diffusion pipeline loaded a fresh InferenceFunction on every inference call (~30 per generation) as a workaround for an MPSGraph buffer caching bug. That bug is fixed since macOS 27 Beta 3+. The workaround caused GPU memory to accumulate across generations, leading to SIGABRT after ~20 images.
Now reuses the stored function (matching how LLM engines work). Also wraps model loading and inference in do/catch to surface actionable errors instead of crashing on GPU memory exhaustion.
Addresses #77.