Fix issues with async TP#117
Open
LucasWilkinson wants to merge 2 commits into
Open
Conversation
Signed-off-by: Lucas Wilkinson <[email protected]>
…mputed When scheduler metadata is computed separately (skip_scheduler_metadata_computation=true), there may be other PDL users (e.g., symmetric memory all-reduce for async TP) between the scheduler call and the attention call. These can interfere with FA3's PDL signaling chain, causing hangs. This extends the previous fix (disabling prepare_varlen PDL) to also disable the main kernel -> combine kernel PDL when using pre-computed scheduler metadata. Signed-off-by: Lucas Wilkinson <[email protected]>
5f86b74 to
c427cae
Compare
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
FA3 always using PDL can cause a deadlock when combined with async TP which also uses PDL (in PyTorch's symmetric memory)