Skip to content

fix(setup): Move Linux and Windows install to UV wheelnext environment for auto GPU detection. (#13)#15

Merged
Tok merged 1 commit into
Tok:devfrom
danrossi:setup-uv-wheelnext
Jan 20, 2026
Merged

fix(setup): Move Linux and Windows install to UV wheelnext environment for auto GPU detection. (#13)#15
Tok merged 1 commit into
Tok:devfrom
danrossi:setup-uv-wheelnext

Conversation

@danrossi
Copy link
Copy Markdown
Contributor

UV won't install the GPU Torch in Windows but the wheelnext UV should automatically detect the GPU (nvidia,amd,intel) and install the right version of Torch.

I think UV is the best method for GPU support over pip.

UV can be used to install the required Python 3.12.

  • Install UV wheelnext in Linux script and force uv over pip.
  • Install UV wheelnext in Windows setup script.
  • Use UV to install and pin Python 3.12
  • Add the check cuda helper script.

The nvidia provider plugin is broken on Windows for now.

wheelnext/nvidia-variant-provider#8

…uto GPU detection. (Tok#13)

- Install UV wheelnext in Linux script and force uv over pip.
- Install UV wheelnext in Windows setup script.
- Use UV to install and pin Python 3.12
- Add the check cuda helper script.
@codecov
Copy link
Copy Markdown

codecov Bot commented Jan 20, 2026

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 92.56%. Comparing base (c714df5) to head (9586fc0).
⚠️ Report is 1 commits behind head on dev.

Additional details and impacted files
@@           Coverage Diff           @@
##              dev      #15   +/-   ##
=======================================
  Coverage   92.56%   92.56%           
=======================================
  Files          38       38           
  Lines        2974     2974           
=======================================
  Hits         2753     2753           
  Misses        221      221           
Flag Coverage Δ
unittests 92.56% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:
  • ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

@Tok Tok merged commit 62d17e6 into Tok:dev Jan 20, 2026
5 checks passed
@Tok
Copy link
Copy Markdown
Owner

Tok commented Jan 20, 2026

thanks

@danrossi
Copy link
Copy Markdown
Contributor Author

Let me know what you think. For ease of gpu detection of packages install wheelnext uv in the setup on Linux and use uv, for now install python 3.12 with uv and pin it. I need multiple python versions on Windows. This way it may be able to support automatically nvidia, amd and intel gpu for any package with gpu support.

normal UV failed to install GPU Torch on Windows via pip. Wheelnext is the better solution but the nvidia plugin has an issue right now I am working on to get Windows working. I am not sure what my resource issue still is preventing the script completing. Either WSL as it crashes WSL or system wide. I won't know until I get Powershell method working on Windows.

wheelnext/nvidia-variant-provider#8 (comment)

@danrossi
Copy link
Copy Markdown
Contributor Author

Wheelmext is very alpha. The issue with the installation of GPU torch and the nvidia wheels on Windows. Is because they have a marker filter for Linux only. It should be fixed soon after I reported the install log.

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.

2 participants