feat: add chat template to existed gguf file

[bruceunx]

feat: add chat template to existed gguf file

follow the issue 5897

add gguf_add_chat_template.py to gguf-py/scripts

[ngxson]

Thanks for the contribution. However, I'm quite doubt if this file may be confused for new users. Some users may misunderstand that this script allow them to use which ever templates they want, without not knowing that llama.cp can only support a predefine set of templates

I think it would be nice to include some kind of warning. And would be even nicer if you can take time to allow the script to change other attributes in metadata.

[bruceunx]

the chat template actually trained in the model, and normally included in the tokenizer_config.json in original model file, while some old gguf missed the chat template in metadata while quantized.
Actually It's not proper to add any other template to model which was not trained in the model.
I thought this script just for gguf which quantized before without including chat template in it, this gguf file is all OK just missing the proper chat template. :)

[bruceunx]

Maybe I can add predefined chat templates for known popular open source llms models, which maybe solve this concern