llama.cpp

model: minicpm should use llm_build_granite

#13911

Merged