How to fine-tune to create items with characteristics

[lyfhn]

Hi,

I have a list of items from League of Legends. DaVinci knows them, but when asked about their characteristics it got it wrong. So I thought I would fine-tune the model to give it the updated characteristics. Here is an example of a prompt for an item:

• Serylda’s Grudge: 45 Attack Damage, 30% Armor Penetration, 20 Ability Haste. Bitter Cold: Damaging Abilities Slow enemies.

But I’m not sure how to format the dataset. I read the resource from OpenAI’s documentation and it wasn’t very helpful.

I read : How do I format my fine-tuning data? | OpenAI Help Center and I think it gets very close to what I need. But I would like a confirmation before I jumped into it. And some explanation as to what the indicator string like ‘->’ actually means and how adding such a random string in the prompt doesn’t mess with it

Thanks !

[ruby_coder]

@lyfhn

Thanks !

Study this tutorial and you should have a much better idea about fine-tuning:

Fine-Tuning In a Nutshell with a Single Line JSONL File and n_epochs

Those who frequent this community easily notice that developers struggle to get good fine-tunings results. In this multi-part tutorial, I’m going to use the davinci base model and fine-tune it multiple times with the same single-line prompt using different values for n_epochs. I will take a completion with each of these models, and demonstrate that if you set your n_epochs value high enough, you can get excellent results fine-tuning with a single-line prompt! Before we get started, I would …

HTH

[Artbyalex]

This sounds like something embedding could solve. Have you looked into that?