optimize LLAMA_CI_PATH, fix python package missing problem

[Orwlit]

LLAMA_CI_PATH is now initialized once you launch chatbot.py, so users don't have to init LLAMA_CI_PATH manually.

When I run this project on my Mac, I encountered several issues of missing python package. the flowing packages should be added into requirements.txt:

• rich
• accelerate
• bitsandbytes

Besides, I noticed that there is an error in your README.md: click this link --mode_path should be --model_path

I think the best way to implement a code interpreter with Llama2, is to use langchain with its components. I am interested in doing such thing but did not find a proper git repo yet. If you are interested in doing so please let me know, Thanks!

[SeungyounShin]

@Orwlit Thank you for your valuable input! I appreciate the pull request and your keen observations.

I've noted the missing python packages you mentioned and will be adding rich, accelerate, and bitsandbytes to the requirements.txt soon. Also, thank you for pointing out the error in the README.md; I'll correct the --mode_path to --model_path.

As for your suggestion about implementing a code interpreter with Llama2 using langchain, it sounds intriguing. I'm always open to exploring new tools and methods. However, my current focus is on data generation. Please also note that my latest feature on finetuning (had to be reset to the initial commit due to an inadvertent push of the Google Search API.)

While I see the potential of langchain in managing prompts and external tools, I'm leaning towards introducing a feature that manages external tools through code, instead of relying solely on a separate tool. Nonetheless, your suggestions have piqued my interest, and I will definitely review the code thoroughly to integrate any beneficial aspects.

Thank you once again for your contribution and interest in the project!