Hi everyone,
EDIT: I’m looking to do the following (example I did in the playground with text-davinci-003)
I assumed in red is embedding from what I read in the different forum post.
How can I proceed to do this on a fine-tuned model ?
And for updating, I was thinking to inform my model through an API by passing a new line, seems like it work fine.
Thank you.
Welcome to the forums.
This thread on fine-tuning might help…
Hi @PaulBellow ,
Yes I’ve come across that post ! It’s super interesting.
Basically, Fine tuning is for structure and embedding is for knowledge.
My problem is that I think I need both.
-
I need to inform with specific data. Ex: The company ABC Construction is located at 3900 green street with the postal code 80500.
-
Then, I need to be able to do a prompt that returns a structured response, such as:
Example: Find the customer ABC Construction
Output: {“customer_name”:“ABC Construction”,“address”:“3900 green street”,“postal_code”:“80500”}
For doing the step 2, I can use a base model Davinci and do fine tuning.
What about the step 1 ? I’ve been looking the whole day and I am not sure how I can get both working together ?
Thank you.
Here is an example in the Playground that work well with text-davinci-003. I assume in red was embedding. How can I reproduce this?
And for updating, I was thinking to inform my model through an API by passing a new line, seems like it work fine.