Add local attention mask support to CUTLASS Blackwell decode op #58
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
|
@Aya-ZIbra has exported this pull request. If you are a Meta employee, you can view the originating Diff in D89192917. |
Summary: Extends the xformers CUTLASS Blackwell FwOpDecode operator to support BlockDiagonalLocalAttentionPaddedKeysMask, enabling sliding window attention for decode workloads. This builds on the gen kernel sliding window implementation to provide end-to-end local attention support through the xformers API. Also for consistent shapes, I changed the default split_k_size from 1024 to 0 in the interface to use no split-K sizing. TODO: Add merge_attentions step to cutlass op when number of splits >1 Differential Revision: D89192917
Aya-ZIbra
force-pushed
the
export-D89192917
branch
from
5204841 to
722f6dc
Compare
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Outdated suggestions cannot be applied.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.
Summary:
Extends the xformers CUTLASS Blackwell FwOpDecode operator to support BlockDiagonalLocalAttentionPaddedKeysMask, enabling sliding window attention for decode workloads. This builds on the gen kernel sliding window implementation to provide end-to-end local attention support through the xformers API.
Also for consistent shapes, I changed the default split_k_size from 1024 to 0 in the interface to use no split-K sizing.
TODO: Add merge_attentions step to cutlass op when number of splits >1
Differential Revision: D89192917