Open
Conversation
angeloskath
requested changes
Mar 3, 2026
Member
angeloskath
left a comment
angeloskath
left a comment
There was a problem hiding this comment.
Looks good!
I left a few comments, lmk what you think. Especially about removing the communication type 🤔. I think it 's just unnecessary complexity at this point.
python/mlx/nn/utils.py
Outdated
| rank = group.rank() | ||
| world = mx.distributed.init() | ||
| N = world.size() | ||
| fsdp_group = fsdp_group or world |
Member
There was a problem hiding this comment.
That is interestingly incorrect.
mx.distributed.init() doesn't return the world communicator but the first one instantiated. There is no guarantee that it would even be the same type ie world could be TCP ring and fsdp_group could be NCCL.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Added hybrid sharding to fsdp.
I refactored the code, because I thought it would be a bit more consistent, but I can revert it back and simply call
mx.distributed.all_sum.