To QA-Your-Docs, You will NEED to Fine-tune Your GPT, on Your Data.
That's my conclusion after:
1. playing with a Split-to-Chunks-&-Chain solution by #LangChain, and
2. running #vicuna on my home PC with #oobabooga WebUI (wow)
https://t.co/N6wVvlpbLA
https://t.co/ruCpnlVzez
How to query private docs with @OpenAI GPT & @langchain ?
Check out the notebook I played with, my notes & findings on this big challenge. Also included how-to and general background info
@hwchase17@ylecun@DmitryKan
https://t.co/UyC9n2TcIl
@Yampeleg Hi Yam, much respect for this critical project, well done! I'm having difficulties to run it on HF Inference Endpoint (AWS), getting "Endpoint failed to start" something to do with the expected q_dtype, any advise how to fix it?
@AlphaSignalAI@geoffreyhinton: "from the data, it figures out how to extract the meaning of the sentence. And it uses the meaning of the sentence to predict the next word.
It really does understand. And that's quite shocking."
https://t.co/AZGfqzcoUC
@yudapearl I believe that @ilyasut will not argue with the thesis of underlying causal relations, the claim is that by learning next token prediction of ~all the text ever produced, advanced LLM managed to understand these causal relations quite well. I think @geoffreyhinton also thinks so.
@Liad_Mudrik feels like something to be... I guess this is following Nagle's attempt 50 years ago to define consciousness. For me this definition is a good example how philosophy is more useless than not.
The video for my North Bay Python talk is out, and I've put together an accompanying edited transcript with annotated slides and links
https://t.co/0Dbkxz2BpI
If you haven't been completely immersed in this world for the last year, my hope is this can help catch you up!