llama.cpp

83f3004c - convert: allow using quantized Mistral weight