[V0.9.1] Replace FA interface with FA_V2 to optimize perf in SelfAttention #1701
+14
−16
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
Signed-off-by: rjg-lyh <1318825571@qq.com>
ApsarasX
reviewed
Jul 14, 2025
| num_heads=self.num_heads, | ||
| num_kv_heads=self.num_kv_heads, | ||
| out=output) | ||
| torch_npu.atb._npu_flash_attention_v2( |
There was a problem hiding this comment.
Do you have the performance data for this PR?
|
Operating during the prefill phase can save on HBM activation and improve FA computation efficiency, especially in long sequence scenarios. |
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Outdated suggestions cannot be applied.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.
What this PR does / why we need it?
Due to the lack of support for passing compressed masks in the FA interface, performance significantly degraded in long-sequence scenarios, even leading to functional issues such as OOM errors. Therefore, I switched to using the FA_V2 interface for the selfattention computation, ensuring functionality while greatly improving performance.
Does this PR introduce any user-facing change?
No.
How was this patch tested?
CI passed with existing test.