GPT3 fine-tune data

API
1

Hi everyone,

I am a graduate student, recently I am doing an research to do with novel writing. I want to fine-tune GPT3 model using some novels. Does anyone know what should my training data look like? What I’m doing now is follow the instruction on web:

{“prompt”: “”, “completion”: “”}
{“prompt”: “”, “completion”: “”}
{“prompt”: “”, “completion”: “”}

is the paragraph of the novel and is the next paragraph of it.

2

What are you trying to achieve?