How can we use vicuna for information retrievel from bunch of docs?

[alan-ai-learner]

trafficstars

Hi @yantao0527 @infwinston @Mearman @zhisbug @jegonzal , I've got a question, I have let's say a book on NLP and I want to use Vicuna in such a way that if user asks any question, the model can answer from the book only. Is it possible using Vicuna or other llms?

Any suggestion would be very helpful! Thanks!

[yantao0527]

I recommend a youtube video to learn how to use OpenAI's new GPT-4 api to 'chat' with and analyze multiple PDF files.

[alan-ai-learner]

@yantao0527 , i will look into it, that is my last and expensive option , but before that i was thinking is it possible with fine tuning any open source model like vicuna.

Thanks!

[merrymercy]

You can combine other tools (e.g. LangChain) and Vicuna's OpenAI-compatible API (https://github.com/lm-sys/FastChat#openai-compatible-restful-apis--sdk).

I think it does not work out-of-the-box now. Contributions are welcome.

[Marenz]

Could fine-tuning on the book be done easily? How would you need to structure the data?

[hitchclimber]

I would also be interested to know how one could fine-tune Vicuna with books or any other large bodies information

[surak]

While this is good, it's not really an issue of fastchat, right?

I would recommend you have a look at this colab on how to fine-tune llama2 as an example: https://colab.research.google.com/drive/1Ly01S--kUwkKQalE-75skalp-ftwl0fE?usp=sharing