Huggingface now owns llama.cpp, btw.

https://huggingface.co/blog/ggml-joins-hf