rudr937 / gpt-3-training-data

A set of questions & answers used to train a chatGPT model.

GPT-3 training data

This repository holds the training.jsonl file.
This file contains a set of questions & answers that are used to train a chatGPT model.

This way it's possible to train chatGPT to respond to questions regarding Astro, helping users understand and use the bot without the need to wait for human support.

Contribute

To contribute simply fork this repository, add your set of questions and answers and then make a pull request.

✨ You will receive a special @Contributor role in the Support Server of Astro for helping with the project!

File format

The file training.jsonl has a json object for each line.
Each json object MUST have the following form:

{ "prompt": "<question>\n\n===\n\n", "completion": " <answer> END" }

It's very important to keep the \n\n===\n\n and the spaces between the <answer>.

Example

{ "prompt": "What is Astro?\n\n===\n\n", "completion": " Astro is a Discord bot that can create temporary voice channels and assign temporary roles to users inside voice channels END" }