llama.cpp

llama : fix llm_build_k_shift to use correct n_rot

#4889

Merged