llama.cpp

5d16e81d - convert : keep quantization scales for nemotron --mtp export (#26903)