This change introduces a pure JAX implementation of flash attention to Maxtext, designed as a drop-in replacement for the existing Pallas kernel. In this cl we set up the stage by integrating it with maxtext in fsdp mode. We have plans for further optimizations to close the gap with pallas using different techniques such as: #2793
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
…o Maxtext, designed as a drop-in replacement for the existing Pallas kernel. In this cl we set up the stage by integrating it with maxtext in fsdp mode. We have plans for further optimizations to close the gap with pallas using different techniques such as: iteration skipping, must_fuse, and memory space coloring. The new implementation is located in maxtext/src/maxtext/kernels/jax_flash_attention.py and can be enabled with the use_jax_splash config flag. To validate the implementation and compare it against the Tokamax kernel and the baseline dot-product attention, this change also introduces: A new test suite in google_mla_attention_test.py for correctness and performance comparison, particularly for FSDP cases. Refactored common MLA test utilities into attention_test_util.py. PiperOrigin-RevId: 834764107
copybara-service
bot
requested review from
A9isha,
NicoGrande,
NuojCheng,
RissyRan,
SurbhiJainUSC,
aireenmei,
bvandermoon,
gagika,
gobbleturk,
hengtaoguo,
jiangjy1982,
khatwanimohit,
parambole,
richjames0,
shralex,
shuningjin,
suexu1025 and
vipannalla
as code owners
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Outdated suggestions cannot be applied.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.
This change introduces a pure JAX implementation of flash attention to Maxtext, designed as a drop-in replacement for the existing Pallas kernel. In this cl we set up the stage by integrating it with maxtext in fsdp mode. We have plans for further optimizations to close the gap with pallas using different techniques such as:
iteration skipping, must_fuse, and memory space coloring.
The new implementation is located in maxtext/src/maxtext/kernels/jax_flash_attention.py and can be enabled with the use_jax_splash config flag.
To validate the implementation and compare it against the Tokamax kernel and the baseline dot-product attention, this change also introduces:
A new test suite in google_mla_attention_test.py for correctness and performance comparison, particularly for FSDP cases.
Refactored common MLA test utilities into attention_test_util.py.