I’ve started creating a GPT with the GPT-Builder. It uses spec/description files uploaded to the Knowledge to understand the format of a JSON input. It’s working well and is delivering consistent results.
Next - I would like to test against gpt-3.5 and work on some fine tuning.
Using the same prompt and uploaded files on the Playground, I get different results and the prompt cannot parse the JSON format.
How is GPT-Builder different from the playground?
Can I transition from GPT-Builder to using the playground or is it a full refactor?