llama.cpp

019ba1dc - convert : fix Baichuan2 models by using vocab size in config.json (#3299)