llama.cpp

58cba76a - gguf-py : fix TemplateProcessing pair when bos/eos is missing (#14312)