Skip to content

Try to re-use already loaded libcuda/nvrtc/cudart#2

Open
Luthaf wants to merge 2 commits intometatensor:mainfrom
Luthaf:re-used-loaded-libs
Open

Try to re-use already loaded libcuda/nvrtc/cudart#2
Luthaf wants to merge 2 commits intometatensor:mainfrom
Luthaf:re-used-loaded-libs

Conversation

@Luthaf
Copy link
Copy Markdown
Member

@Luthaf Luthaf commented May 1, 2026

This way we don't load a different version from the rest of the code by mistake. This should also enable re-using libraries from torch installation when creating custom torch ops, and fix lab-cosmo/sphericart#230

@RMeli
Copy link
Copy Markdown
Member

RMeli commented May 4, 2026

cscs-ci run

@Luthaf Luthaf force-pushed the re-used-loaded-libs branch 2 times, most recently from 6373524 to 8d96684 Compare May 4, 2026 08:00
This way we don't load a different version from the
rest of the code by mistake. This should also enable
re-using libraries from torch installation when creating
custom torch ops
@Luthaf Luthaf force-pushed the re-used-loaded-libs branch from 8d96684 to f68eb31 Compare May 4, 2026 08:13
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.

Enable CUDA Windows support without CUDA toolkit

2 participants