But I am also doing local model stuff, which means I can only run on CPU. I know on Windows there are containers that will forward CUDA stuff to the host system to run, but there is no such thing on MacOS.
The builtin devcontainer stuff is based on Docker, but maybe someone has hacked something together to get native virtualization to present itself like docker so i could get this to work?
What do other people do?