llama.cpp

b248f4a3 - gguf-py : ByteLevel processing defaults bos/eos to False (#29422)