llama.cpp

model: fix llama arch implementation

#17665

Merged