Difference between AXLearn and MLX

[ManishAradwad]

I was reading about apple foundation models and it mentioned that the pretraining of the local llms used is done on axlearn. So just wanted to know why do we have two different frameworks for machine learning on apple silicon? Is mlx going to replace axlearn eventually? or is it mostly similar to tensorflow and jax situation?

[awni]

Axlearn is a tool built on top of Jax and is primarily designed and used for training large language models on clusters of TPUs.

MLX is an array framework designed specifically for Apple silicon.