DFlash/Domino: Add support for the FlexAttention FLASH backend#714
Open
zou3519 wants to merge 1 commit into
Open
DFlash/Domino: Add support for the FlexAttention FLASH backend#714zou3519 wants to merge 1 commit into
zou3519 wants to merge 1 commit into
Conversation
Adds an option to turn on FAv4 backend for FlexAttention. This is significantly faster on Blackwell than the default backend (which is triton). We unfortunately need to monkey-patch PyTorch (in 2.11, which is what SGLang is pinned to right now). Whenever SpecForge can upgrade to PyTorch >= 2.13 then we can get rid of the monkeypatches (and also the correctness test). This PR also cleans up the attention implementation a little.
Contributor
|
Warning You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again! |
zou3519
marked this pull request as ready for review
July 21, 2026 19:31
zou3519
requested review from
FlamingoPg,
FrankLeeeee,
shuaills and
sleepcoo
as code owners
July 21, 2026 19:31
Contributor
|
Caution The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased. |
Author
|
cc @maocheng23 |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Adds an option to turn on FAv4 backend for FlexAttention (it remains off by default). This is significantly faster on Blackwell than the default backend (which is triton). We unfortunately need to monkey-patch PyTorch (in 2.11, which is what SGLang is pinned to right now). Whenever SpecForge can upgrade to PyTorch >= 2.13 then we can get rid of the monkeypatches (and also the correctness test).
This PR also cleans up the attention implementation a little.