Add POD-Attention to FlashInfer #858
Open
+2,837
−143
Conversation
This file contains 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
added 19 commits
February 14, 2025 17:02
…r to POD instead of batched decode.
…y decode kernel is manually called and needs to be merged into POD kernel
15 tasks
|
Sorry, renamed the branch on my fork which accidentally closed this PR. Reopened now. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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 POD-Attention kernel (https://arxiv.org/abs/2410.18038) with all necessary files.
Both AOT and JIT are supported.
NOTE: prefill.cuh and decode.cuh have been changed.
All helper functions that use threadIdx and blockIdx now use function arguments that supply these instead. Since these device functions are inline, this shouldn't impact performance. The prefill kernels (SinglePrefillWithKVCacheKernel and BatchPrefillWithPagedKVCacheKernel) have been changed into device functions (called by POD-Attention) and new wrapper kernels for the device function have been created.