llama.cpp

784a14aa - convert : add support for Roberta embeddings (#10695)