Fix GPU assignment for Slurm-launched Ray clusters#1592
Open
agolajko wants to merge 3 commits intoNovaSky-AI:mainfrom
Open
Fix GPU assignment for Slurm-launched Ray clusters#1592agolajko wants to merge 3 commits intoNovaSky-AI:mainfrom
agolajko wants to merge 3 commits intoNovaSky-AI:mainfrom
Conversation
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
![devin-ai-integration[bot]](https://avatars.githubusercontent.com/in/811515?s=60&v=4){kind=small}
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Fixes #1577
Problem
DistributedTorchRayActor.__init__assumed Ray always narrowsCUDA_VISIBLE_DEVICESto a single GPU per actor, makingLOCAL_RANK="0"safe. On Slurm clusters, Ray inheritsCUDA_VISIBLE_DEVICES="0,1,...,7"from Slurm and never narrows it, so every worker callsset_device(0)and piles onto physical GPU 0, causing silent OOMs.Fix
We explicitly narrow
CUDA_VISIBLE_DEVICESto the Ray-assigned physical GPU, makingLOCAL_RANK="0"unconditionally correct. This replaces theray_noset_visible_devices()ternary which only checked whether the user disabled Ray's narrowing, not whether Ray actually performed it.Scenarios
"3"(Ray narrowed)"3"(no-op)"0"✓"0,1,...,7""3""0"✓"0,1,...,7""3""0"✓