The newest llama.cpp provides a new ggml encoding format, and no modern models support this version. Can you update the llama.cpp version? Thank you!