llama.cpp

feat: add support for Roberta embeddings

#10695

Merged