Llama2 fine-tuned chat version

[ksmdnl]

Hi, thanks for the conversion work on the Llama models!

I have a question regarding the version of the Llama models. So there are two versions for Llama2 models, the standard one and the one fine-tuned for chat completion. Given the prompt examples in generate.py, I suppose that this repository supports the standard version, i.e. for text completion task and needs to be modified for the chat completion task?

Looking forward for a discussion!