This ArchLinux User Repository (AUR) is aimed at distribution of software for artificial intelligence, machine learning, and scientific applications in general. The rationale behind it is to maintain single and uniform environment with support of containerization technologies across different hosts for training and experimenting with modern neural networks without much pain.
In order to start using it, update pacman config and add the following
lines to pacman.conf.
# /etc/pacman.conf
[ai]
SigLevel = Optional TrustAll
Server = https://arch.daskol.xyz/$repo/os/$archThen refresh package databases with pacman -Sy. Now, you can use it: see list
of prebuilt packages or install JAX with CUDA devices support.
pacman -Sl ai # List packages in repo.
pacman -S python-jax python-jaxlib-cudaAlso, there is a pre-built docker images on top of corresponding base of
archlinux:base and archlinux:base-devel images.
docker pull ghcr.io/archlinux-ai/base # small and simple
docker pull ghcr.io/archlinux-ai/base-devel # devtools inside