You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Referring to past issues, I was able to get the code working by replacing all instances of "text-davinci-00x" (x = 1 to 3) in defunct_run_gpt_prompt.py, gpt_structure.py, and run_gpt_prompt.py with "gpt-3.5-turbo-instruct". I really appreciate the insights shared by others — they were very helpful.
As far as I understand, this setup now works fine with gpt-3.5-turbo-instruct.
That said, has anyone tried running this with newer models like gpt-4-turbo or gpt-4o (especially 4o)?
If you're aware of any changes needed in the code (such as adjustments in API call structure or response handling), I’d be very grateful for your advice or pointers.