llama.cpp

d23bdd57 - convert : handle int-quantized models