Skip to content
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.

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

Flash Attention Support #83

Closed
amrrs opened this issue May 5, 2024 · 2 comments · Fixed by #84
Closed

Flash Attention Support #83

amrrs opened this issue May 5, 2024 · 2 comments · Fixed by #84
Assignees
Labels
bug Something isn't working good first issue Good for newcomers

Comments

@amrrs
Copy link

amrrs commented May 5, 2024

Hey! Great work.

I think the latest code change for the SpeechToTextPipeline expects all GPUs to be Flash Attention 2compatible.

I'm not sure if there's anyway to override the kwargs.

https://github.com/kadirnar/whisper-plus/blob/487bfa05572a04eb39af260eb3197533ddcdcb0d/whisperplus/pipelines/whisper.py#L72C13-L72C58

I used it on P100 from Kaggle and got the error about Flash Attention

@kadirnar
Copy link
Owner

kadirnar commented May 5, 2024

Can you share the error message?

@kadirnar kadirnar self-assigned this May 5, 2024
@kadirnar kadirnar added bug Something isn't working good first issue Good for newcomers labels May 5, 2024
@kadirnar kadirnar linked a pull request May 5, 2024 that will close this issue
@kadirnar
Copy link
Owner

kadirnar commented May 5, 2024

I added flash-attention2 as a parameter for you to turn off. You can look at Readme.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
bug Something isn't working good first issue Good for newcomers
Projects
None yet
Development

Successfully merging a pull request may close this issue.

2 participants