Merged
Conversation
awni
reviewed
Oct 14, 2025
awni
approved these changes
Oct 14, 2025
Member
awni
left a comment
There was a problem hiding this comment.
Looks great, thanks for fixing it!
faisalmemon
pushed a commit
to faisalmemon/mlx
that referenced
this pull request
Oct 30, 2025
* print the error & delete nccl group * timeout for nccl binding * typo * revert error * fixed a typo
1 task
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Proposed changes
During multi-node runs, the master host sometimes starts later than other nodes. As a result, other hosts wait until rank 0 binds. In the previous implementation the binding retried 30 times with 500 ms intervals, which was not always sufficient (and hard-coded).
env::nccl_timeout-- the binding will fail after the elapsed time exceedsnccl_timeout(default = 300 000 ms = 5 minutes, be overridden via theMLX_NCCL_TIMEOUT).ncclGroupEnd()call -- this was a bug.