llama.cpp

9733104b - drop quantize.py (now that models are using a single file)