Find answers from the community

Updated 3 months ago

Models

So in this example text-davinci-003 is used to generate the text but text-ada-002 is already defaulted to create the embedding?

I'm sorry, I don't really understand that part.
Attachment
image.png
L
3 comments
In this example, the text-davinci-003 is specifically used to generate text yes. If you did not pass in llm_predictor to the index constructor, it would still default to that.

And yes, in this example, text-ada-002 is being used to generate the embeddings
Basically, llama index uses an llm_predictor for text completion and an embed_model for generating embeddings
Usually, the token usage for embeddings is very high. So using a cheap model like text-ada-002 is a good default, because it's very cheap and works very well for generating embeddings
Add a reply
Sign up and join the conversation on Discord