I am currently creating a chat bot that answers questions about a small lake and the 60 communities that surround it. Ideally the chatbot would answer general questions about the lake and each community. But also I would like it to match a user based on input to the perfect community to buy a house in. I have the data set all made in excel for each community as well as prompts and completions but I'm not sure how to implement. Fine tune with the prompts and completions and also use embedding for the data? The embedding data has many columns per row and from what I have read you can only have 2 columns per row. If someone can help me or point me in the right direction I will compensate them no problem.
#Best way to fine tune or embed gpt3 model with new information
river nacelle
vague halo
I am currently creating a chat bot that answers questions about a small lake and...
You can likely handle this with just embeddings and figuring out a good initial prompt. Here's a guide on answering questions using embeddings: https://github.com/openai/openai-cookbook/blob/main/examples/Question_answering_using_embeddings.ipynb.
When you are uploading that embedding data into whatever database you want to use you can add whatever metadata you want to each vector to handle the extra information regarding it.
river nacelle
So no matter what I probably gonna need a vector database? Have you ever dealt with Geospatial data with these models to give the model an idea of what the lake looks like? Also geographic coordinates to answer how far questions. Any pointers?