llama.cpp

08df6bd7 - openai: test features of templates at runtime, to make sure no bits of intel are lost