Fix GQA support in varlen attention by adding enable_gqa=True#2905
Closed
daniellepintz wants to merge 1 commit intomainfrom
Closed
Fix GQA support in varlen attention by adding enable_gqa=True#2905daniellepintz wants to merge 1 commit intomainfrom
daniellepintz wants to merge 1 commit intomainfrom
Conversation
The attention refactor in #2761 moved GQA head expansion out of the attention modules, but the varlen_attn/varlen_attn_out calls were not updated to pass enable_gqa=True. This causes a ValueError when query and key/value have different numbers of heads (e.g. Hq=8, Hkv=4). Fixes both the core VarlenAttention (trainer path) and the RL PyTorchFlashAttentionImpl (vLLM generator path).
daniellepintz
requested review from
fegin,
tianyu-l,
wconstab and
wwwjn
as code owners
meta-cla
bot
added
the
CLA Signed
Contributor
Author
|
Fixed in #2891 |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The attention refactor in #2761 moved GQA head expansion out of the attention modules, but the varlen_attn/varlen_attn_out calls were not updated to pass enable_gqa=True. This causes a ValueError when query and key/value have different numbers of heads (e.g. Hq=8, Hkv=4).
Fixes both the core VarlenAttention (trainer path) and the RL PyTorchFlashAttentionImpl (vLLM generator path).