Skip to content

Implement Flash Attention Option #19

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Closed
dustydecapod opened this issue Mar 11, 2023 · 4 comments
Closed

Implement Flash Attention Option #19

dustydecapod opened this issue Mar 11, 2023 · 4 comments
Labels
enhancement New feature or request

Comments

@dustydecapod
Copy link

Would love to see a faster, more memory efficient attention implemented like Flash Attention. :)
@ggerganov ggerganov added the enhancement New feature or request label Mar 12, 2023
@ggerganov
Copy link
Member

In whisper.cpp I tried using FA in the Decoder and it did not help (it does help a lot in the Encoder).
I guess it is a matter of the tensor sizes, but of course, maybe I didn't implement it properly.

ggml-org/whisper.cpp#284

@Orevantum
Copy link

Is it possible to implement multi-query attention then?

In whisper.cpp I tried using FA in the Decoder and it did not help (it does help a lot in the Encoder). I guess it is a matter of the tensor sizes, but of course, maybe I didn't implement it properly.

ggerganov/whisper.cpp#284

@xloem
Copy link
Contributor

xloem commented Apr 5, 2023

also note in flexgen they use top 10% sparse attention

SlyEcho pushed a commit to SlyEcho/llama.cpp that referenced this issue Jun 11, 2023
@digiwombat
Merge pull request ggml-org#19 from lesaun/master
d6d263f 
Clarify build instructions in README.
rooprob pushed a commit to rooprob/llama.cpp that referenced this issue Aug 2, 2023
@karpathy
Merge pull request ggml-org#19 from mcognetta/master
7d62088 
Simplify Softmax
@jamesbiederbeck
Copy link

also note in flexgen they use top 10% sparse attention

Sparse attention is cool, but lossy. Flash attention is exact.

Deadsg pushed a commit to Deadsg/llama.cpp that referenced this issue Dec 19, 2023
@abetlen
Add verbose flag. Closes ggml-org#19
c137789
@Bearsaerker Bearsaerker mentioned this issue Mar 12, 2025
Open
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
enhancement New feature or request
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
5 participants
@xloem @jamesbiederbeck @ggerganov @dustydecapod @Orevantum