llama.cpp

29ca8f25 - py quants: wip q4_k and q3_k quantize