llama.cpp

d12f7810 - llama : streamline embeddings from "non-embedding" models (#8087)