llama.cpp

Extend llm_build_ffn() to support _scale tensors

#8103

Merged