Currently packages like Pytorch are installed as CPU only. With Pixi environments, we could install both CPU and GPU versions, and have the modulefile pick which environment to load dynamically based on the resources on the node.
https://pixi.sh/latest/python/pytorch/#installing-from-conda-forge
Currently packages like Pytorch are installed as CPU only. With Pixi environments, we could install both CPU and GPU versions, and have the modulefile pick which environment to load dynamically based on the resources on the node.
https://pixi.sh/latest/python/pytorch/#installing-from-conda-forge