Comment by leodriesch

How does this compare to MLX? As far as I understand MLX is equivalent to PyTorch but optimized for Apple Silicon.

Is this meant for training MLX models in a distributed manner? Or what is its purpose?