Repository navigation
chore: bump mlx-swift-lm to 4f1fdaa (preload non-expert weights when streaming) - #218
Merged
Merged
Conversation
…streaming) Picks up mlx-swift-lm#80: with --stream-experts, materialize non-expert weights after load so the GPU command buffer does not wait on cold safetensors reads (kIOGPUCommandBufferCallbackErrorTimeout, see #216). Co-Authored-By: Claude Sonnet 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Bumps the
mlx-swift-lmsubmodule9da07be→4f1fdaa(mlx-swift-lm#80).#80 materializes the non-expert weights right after load when
--stream-expertsis on (lazyLoadleaves them as pending safetensors reads, so a cold page cache could stall a GPU command buffer past the watchdog).Notes
kIOGPUCommandBufferCallbackErrorTimeoutin --stream-experts: intermittent kIOGPUCommandBufferCallbackErrorTimeout (server crash), correlates with cold page cache #216. Evidence so far is one very-cold run that passed (n=1); the crash is probabilistic, so please do not treat --stream-experts: intermittent kIOGPUCommandBufferCallbackErrorTimeout (server crash), correlates with cold page cache #216 as closed until it is seen to hold on more cold starts.🤖 Generated with Claude Code