Point mlx at Qwen3.6 decode matvec optimizations (~30 tok/s) - #63

Point mlx at Qwen3.6 decode matvec optimizations (~30 tok/s)#63

goniz merged 4 commits into

maingoniz/mlx-vulkan:mainfrom

cursor/qwen36-decode-opt-088dgoniz/mlx-vulkan:cursor/qwen36-decode-opt-088dCopy head branch name to clipboard