My two cents. Unless you have jargon in your field you just need RAG. Fine-tuning is for adjusting the token associations to predict the most likely next token.
this post was submitted on 30 Nov 2023
1 points (100.0% liked)
LocalLLaMA
1 readers
1 users here now
Community to discuss about Llama, the family of large language models created by Meta AI.
founded 10 months ago
MODERATORS
Ok, so I’ll look more into RAG. I just saw Mathew Bermans video on creating agents with YouAI, but I’m hoping for something that’ll connect to our database (or even just a couple specific tables) instead of having to manually export csv files from my db to then upload.
You have to do some type of search for the correct information for the question. It doesn't matter how you get the information.
I agree with the other guy, you just need semantic search of some flavor unless you have industry specific language that the LLM won't understand well enough to write
Ok great. So if we aren’t getting great results with RAG, then fine tune with our own dataset on our language to see if that helps.
I’m hoping to find an automated way to RAG, ideally with a db connection but maybe could set up an api script to export data from the db and upload to the model. I assume at least the second option exists, so I’ll be looking.
Fine tune + RAG seems to be the method we’re expected to go in. Need to fine tune for terminology and RAG for policy and procedure knowledge.