llama.cpp

f56d6077 - Add byte token type when tokenizer.model is not exists (#4641)