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

uv2 #492

Merged
merged 6 commits into from
Jan 10, 2025
Merged

uv2 #492

merged 6 commits into from
Jan 10, 2025

Conversation

vwxyzjn
Copy link
Collaborator

@vwxyzjn vwxyzjn commented Dec 20, 2024

This PR does a few things:

  • Bump the dependencies (torch, etc)
  • (dev only) introduce uv as the package manager to only lock dependencies (requirements.txt); people should still use the existing pip install workflow until further notice.
  • Optimize the docker file further: we will now use a base ghcr.io/allenai/oi-cuda-no-conda:12.1-cudnn8-dev-ubuntu20.04, so build and load time should be substantially lower.

@vwxyzjn vwxyzjn marked this pull request as ready for review January 9, 2025 15:36
Copy link
Contributor

@ValentinaPy ValentinaPy left a comment

Choose a reason for hiding this comment

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

that's great, thank you!

@hamishivi
Copy link
Collaborator

Is there a particular process to update the uv lockfile?
Also, how does building the docker image work for non-ai2 folks? is the beaker image accessible, or do they have to go through more steps? I glanced at the documentation but didn't see anything describing that (sorry if I missed it).

@vwxyzjn
Copy link
Collaborator Author

vwxyzjn commented Jan 9, 2025

Is there a particular process to update the uv lockfile?

Yes. Three ways:

  • uv add packageA
  • uv lock --upgrade-package packageA
  • directly editing the main files, and run uv lock or uv sync

For now, people will prob continue using the pip workflow, and they can just edit the requirements.txt locally for more flexibility.

Also, how does building the docker image work for non-ai2 folks?

I made the image and it's public, so anyone should be able to download it. They can also build their own image, too, via

 docker build --build-arg CUDA=12.1.0 --build-arg TARGET=cudnn8-devel --build-arg DIST=ubuntu20.04 -f  Dockerfile.base . -t cuda-no-conda:12.1-cudnn8-dev-ubuntu20.04
image

Copy link
Collaborator

@natolambert natolambert left a comment

Choose a reason for hiding this comment

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

Slick stuff! Thanks! Inspires me to try and add tests :D

Copy link
Collaborator

@hamishivi hamishivi left a comment

Choose a reason for hiding this comment

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

thanks for the explanation!

@hamishivi
Copy link
Collaborator

This probably closes #280

Copy link
Contributor

@fabrahman fabrahman left a comment

Choose a reason for hiding this comment

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

This looks great! thanks

@vwxyzjn vwxyzjn merged commit c95fbbe into main Jan 10, 2025
3 checks passed
@vwxyzjn vwxyzjn mentioned this pull request Jan 10, 2025
8 tasks
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.

5 participants