llama.cpp

6eeb4d90 - convert: remove most of the n_mult usage in convert.py (#3098)