Shouldn't prompt formats be an enumerator?

[danemadsen]

Hi i was looking into the code for gpt_params and I'm wondering why there's a boolean property for each prompt format (currently just instruct for Alpaca format and chatml for ChatML)?

Shouldn't this just be a single enumerator property named something like prompt_format or something?

(params.instruct == true && params.chatml == true) should never be possible so it doesn't make sense to me why they each have a property.

[ggerganov]

These parameters are old "hacks" - should be removed and replaced with llama_chat_apply_template() in order to use the correct chat template that comes with the model