If I do Fine Tuning or Embedding with retrieval to improve the model to serve my needs - would these changes be permanent to the model or I would lose all such changes after I logout the session. Please help me to understand as I am very new to this platform/ technology. Many thanks…
Hi and welcome to the developer forum!
Fine tuning is a way to narrow the focus of the model and also show it “how” to act, it’s not suitable to teach the model new information or facts, that is where embeddings come in. You can use both, you could use embeddings to retrieve relevant context which is then processed by a fine tuned model to to give a particular style of output.
You can find documentation on both here:
1 Like
Adding on a bit since this is a common misconception: “Embedding with retrieval” is just a fancy way to perform a relevancy search over your content and return chunks of your content. You then add these chunks to your prompt along with whatever question is being asked, same as if you just copied and pasted the sections manually.
So it doesn’t teach or change the model in any way. It’s just a method of finding the right bits of content to include in your prompt.
2 Likes