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

Add images with flash attention 2 #651

Merged
merged 26 commits into from
Oct 10, 2023
Merged

Add images with flash attention 2 #651

merged 26 commits into from
Oct 10, 2023

Conversation

dakinggg
Copy link
Collaborator

@dakinggg dakinggg commented Oct 7, 2023

CI passes when run manually with flash attention 2

=============================================================================== 471 passed, 2 skipped, 239 deselected, 73 xfailed, 144 warnings in 309.04s (0:05:09) ================================================================================
pip show flash-attn
Name: flash-attn
Version: 2.3.2
Summary: Flash Attention: Fast and Memory-Efficient Exact Attention
Home-page: https://github.com/Dao-AILab/flash-attention
Author: Tri Dao
Author-email: trid@cs.stanford.edu
License: 
Location: /usr/lib/python3/dist-packages
Requires: einops, ninja, packaging, torch
Required-by: 

Will do more testing after this PR goes in so that I can add proper CI and use properly built images instead of just test ones.

setup.py Outdated Show resolved Hide resolved
@dakinggg dakinggg marked this pull request as ready for review October 10, 2023 01:19
@dakinggg dakinggg changed the title [WIP] Add images with flash attention 2 Add images with flash attention 2 Oct 10, 2023
Copy link
Collaborator

@mvpatel2000 mvpatel2000 left a comment

Choose a reason for hiding this comment

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

It's hard to reason about the docker changes. I am assuming you tested this

.github/workflows/docker.yaml Show resolved Hide resolved
@dakinggg
Copy link
Collaborator Author

@mvpatel2000 I tested in that working images get pushed to the staging repo from this pr. I suppose I haven't tested that I didn't break the push to main workflow that releases the production images.

@dakinggg dakinggg merged commit ba6b880 into main Oct 10, 2023
15 checks passed
@dakinggg dakinggg deleted the flash2-upstream branch October 10, 2023 17:28
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.

None yet

3 participants