Bugfix: support qk_head_dim != v_head_dim in FMHA#40
Closed
huanghua1994 wants to merge 1 commit intoNVIDIA:mainfrom
Closed
Bugfix: support qk_head_dim != v_head_dim in FMHA#40huanghua1994 wants to merge 1 commit intoNVIDIA:mainfrom
huanghua1994 wants to merge 1 commit intoNVIDIA:mainfrom
Conversation
The original implementation does not support qk_head_dim != v_head_dim, which is needed in Multi-head Latent Attention. Also fix some test code logic. Signed-off-by: Hua Huang <huah@nvidia.com>
Collaborator
|
@huanghua1994 thank you for your contribution. For the moment, we intent to keep the FMHA sample simple for educational purpose. If you are looking to contributing kernels, please check out the TileGym project: https://github.com/nvidia/tilegym Mind if I close this PR? |
Author
|
Sure. Thank you for reminding me this opening issue. |
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 original implementation does not support qk_head_dim != v_head_dim, which is needed in Multi-head Latent Attention. Also fix some test code logic.
Description
The original implementation does not support
qk_head_dim!=v_head_dim, which is needed in Multi-head Latent Attention. Problem sizes insamples/AttentionFMHA.pyare updated s.t.qk_head_dim!=v_head_dimandq_num_head!=kv_num_headto test a generic GQA case. Parameters and the way calling PyTorchscale_dot_product_attentionare also updated to avoid being unable to find a working backend.All tests have passed locally on a B200.
Checklist