CUDA version depencencies #786
Description
I have CUDA 12.9 installed on a Linux system. I can succesfully use convert.py for a recorded object, however it errors when I try to use train.py-- it just says ModuleNotFoundError: No module named 'torch' (it's installed, 2.8.0+cu129).
When I install pytorch with --index-url https://download.pytorch.org/whl/cu118, it installs, but then I get a proper warning that the pytorch version is a mismatch with the installed CUDA on my system when I try running train.py again.
Is there a roadmap to be able to use the latest CUDA drivers?