fix nccl future execution #126
Merged
Conversation
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
facebook-github-bot
added
the
CLA Signed
H-Huang
force-pushed
the
diloco
branch
3 times, most recently
from
7753bbb to
b23a598
Compare
H-Huang
force-pushed
the
diloco
branch
2 times, most recently
from
70d648a to
fa61e1d
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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Outdated suggestions cannot be applied.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.
Async NCCL collectives dont block the CPU which means their futures will also get executed. This causes an issue in
manager.pywhen we modify theallreducetensor in a callback (https://github.com/pytorch/torchft/blob/main/torchft/manager.py#L292) as the tensor may not have finishedallreduce.The fix is to add an event after the allreduce has been
wait()ed and then wait on this event before setting the future.Created a test to repro:
pytest torchft/manager_integ_test.py -vsk test_manager_allreduce