Document why MLX is not shipped: 2.9x slower than Core ML and non-finite output 481ef57 verified msluszniak commited on 12 days ago