llama.cpp

92383bfa - quantize : store metadata for prior weight used for imatrix