How do I verify the answer from the fine tuned model?

I am trying to fine tune a InstructGPT as I did not really understand what is going on. So I have tried to have my dataset like below:

training_data = [{
    "prompt": "Who is my boss? ->",
    "completion": "nicky.\n"
    "prompt":"Who is my buddy? ->",
    "completion": "patrick.\n"

I did not get the answer right when I was testing with OpenAI playground.
How does it works?

Fine-tuning is for teaching the model how to answer questions. If you want to impart new information, the best way currently is to use embeddings to put that information into the model’s context.

1 Like

Do u have any example on before and after result? Cause I don’t really see any big chances with all the example over the internet.

Examples of what? I’m not entirely clear about what it is exactly you’re trying to do.

I mean the result of after fine tuning a model.
I am trying to create something for data entry based on InstructGPT.