llama.cpp

fix(gguf-py): special tokens are no longer skipped when add_<token>_token is set to false

#5487

Merged