Skip to content

Conversation

@Narsil
Copy link
Contributor

@Narsil Narsil commented Aug 14, 2023

Co-Authored-By: Victor Sanh [email protected]

What does this PR do?

Fixes # (issue)

Before submitting

  • This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
  • Did you read the contributor guideline,
    Pull Request section?
  • Was this discussed/approved via a Github issue or the forum? Please add a link
    to it if that's the case.
  • Did you make sure to update the documentation with your changes? Here are the
    documentation guidelines, and
    here are tips on formatting docstrings.
  • Did you write any new necessary tests?

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.

Narsil and others added 5 commits August 15, 2023 12:53
Next:
- precompute cos/sin
- Fix KV layout
- Refactor everything for flash (long).
Fused KV (QKV cannot be done for cross attention because shapes are
different) in attention
Fused Gate + up.

~at 40ms
@Narsil Narsil requested review from VictorSanh and gante August 15, 2023 16:06
Copy link
Contributor

@VictorSanh VictorSanh left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

yaaaay! thanks for the all the improvements!

Copy link
Member

@gante gante left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM 👍

I have suggested a few minor nits in the transformers PR (here) as a result of this review, which may optionally be ported here as well :)

@Narsil Narsil merged commit bce5e22 into main Aug 17, 2023
@Narsil Narsil deleted the idefics branch August 17, 2023 12:38
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants