Skip to content

Conversation

@VoXoXoX
Copy link
Contributor

@VoXoXoX VoXoXoX commented Jul 23, 2025

It can avoid issue when there are package updates and guarantee everyone will have the same package versions (too many issue happen with requirements files) "uv lock --upgrade" can be used to upgrade the dependencies

Easy usage to install package for all cases:

  • uv sync --extra cpu
  • uv sync --extra cuda
  • uv sync --extra cuda-rtx50
  • uv sync --extra rocm

@VoXoXoX
Copy link
Contributor Author

VoXoXoX commented Jul 23, 2025

I only test it on Linux. A test on Windows could be better to confirm if it works too

@ack00gar
Copy link
Owner

Thanks again, I will test it on Mac when I can :)
We'll see if someone can test on Windows.

@VoXoXoX
Copy link
Contributor Author

VoXoXoX commented Jul 28, 2025

By the way, I choose UV instead of poetry because poetry doesn't support easily the installation for pytorch for different source (cuda, cuda-rtx, rocm)

@VoXoXoX VoXoXoX requested a review from Rakly3 August 1, 2025 19:06
@VoXoXoX VoXoXoX force-pushed the use_uv branch 2 times, most recently from 4eb0693 to 9b9b6ab Compare August 10, 2025 18:43
It can avoid issue when there are package updates and guarantee
everyone will have the same package versions (too many issue happen with requirements files)
"uv lock --upgrade" can be used to upgrade the dependencies

Easy usage to install package for all cases:
- uv sync --extra cpu
- uv sync --extra cuda
- uv sync --extra cuda-rtx50
- uv sync --extra rocm
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