Skip to content

Conversation

@kjh-giesbertz
Copy link

@kjh-giesbertz kjh-giesbertz commented Jan 22, 2026

Currently the user needs to install torch him-/herself to be able to use skala. libtorch can actually be downloaded from the pytorch site, so the PR adds this functionality to cmake.

  • The cpu version is pretty straightforward.
    Tested on:
    • Mac
    • Linux (wsl)
    • Windows
  • The gpu version is more tricky as the download needs to match the particular cuda version. The script is supposed to test whether there is a match and if not, falls back to the cpu version. The versions are quite strict though: 12.6, 12.8 and 13.0.
    • Linux (cuda 12.8)
    • Windows

Klaas Giesbertz and others added 2 commits January 19, 2026 13:31
…sion is probably ok (at leat on my mac it works), but gpu still needs to filter versions correctly.
…loads are available. Still needs to be tested.
@kjh-giesbertz kjh-giesbertz marked this pull request as draft January 22, 2026 09:55
@kjh-giesbertz kjh-giesbertz marked this pull request as ready for review January 22, 2026 09:55
@kjh-giesbertz kjh-giesbertz marked this pull request as draft January 22, 2026 09:55
@kjh-giesbertz kjh-giesbertz marked this pull request as ready for review January 22, 2026 14:31
…nd_package(CUDA REQUIRED) needed to included. The CUDA_VERSION_STRING conflicts, so renamed that to CUDA_VERSION_NO_DOT.
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.

1 participant