EMLX
An Nx backend that runs GPU-accelerated tensor operations on Apple Silicon through Apple's MLX library, for numerical computing and machine learning in Elixir.
EMLX is a backend for Nx, Elixir's numerical computing library, that implements Nx's tensor operations on top of Apple's MLX library, enabling GPU-accelerated tensor computation on Apple Silicon from Elixir code. The same repository also holds EMLXAxon, which rewrites supported nodes of Axon, Elixir's neural network library, to use fast Metal shader implementations, accelerating LLM inference on Apple Silicon.
Each of the two projects, EMLX and EMLXAxon, has its own README with further details, and the maintaining organization's page gives a general introduction to doing machine learning in Elixir.