chore: bump mlx-swift for compile scoped-device fix (#176) - #177
Merged
Merged
Conversation
Picks up SharpAI/mlx-swift#17, which makes compiled functions run on the task-scoped default device. Fixes the --gpu-layers Metal GPU timeout (#176). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Bumps
mlx-swiftfrom cbb6a09 to 384de38 to pick up SharpAI/mlx-swift#17.That PR makes
compiled functions run on the task-scoped default device. Before it, compiled MoE helpers (weightedExpertSum, the GELU/SiLU product) that were first traced in a GPU layer were reused inside CPU layers under--gpu-layers. The GPU command buffer then waited on the CPUGatherQMMpast the Metal watchdog, and the server aborted withkIOGPUCommandBufferCallbackErrorTimeout. Details are in #176.The other commits in the range are CI-only (SharpAI/mlx-swift#18, selecting Xcode 26 for the SwiftLM integration jobs) and the merge commit for SharpAI/mlx-swift#16.
Tested on M5 Pro
--model mlx-community/gemma-4-26b-a4b-it-4bit --gpu-layers 23: a 490-token request no longer crashes and returns a correct answer.testCompileHonorsScopedDefaultDevicefails without the fix and passes with it.CPU-resident MoE layers are still very slow (single-threaded, ~0.4 tok/s prefill). That is noted in #176 and the README.
Fixes #176
🤖 Generated with Claude Code