0

I want to fine tune an OpenAI model so that it can answer questions to a document that it has seen while training. What should my training data look like?

I read that it needs to look like prompt and completion pairs, but I'm not sure on how to actually do it. Do I have to use the completions as answers to my questions? How can I provide the text so that the model can learn from it?

Talha Tayyab
  • 8,111
  • 25
  • 27
  • 44
JoLa
  • 11
  • So I can't really continues training the OpenAI model to learn my document? I can only use context based answering (prompt engineering) ? – JoLa Apr 04 '23 at 07:46

0 Answers0