Sole purpose of this repository is to play around with MLX, to see what are the possible speed ups on the Apple silicon.
So far it only contains simple matrix multiplication, but I have plans to grow this collection further.
Those simple tests are run on the 2021 14" MacBook Pro with M1 Pro (8 cores) processor and 32GB of memory.
Dot product using mlx turns out to be on the average (over 10 runs) 12.5 times faster than numpy implementation.
Simply install dependencies using poetry and run chosen test.