llama.cpp

common,server : fix custom preset dedup against cached models

#25235

Open