teknium1/GPTeacher
A collection of modular training datasets generated by GPT-4 for fine-tuning instruction-following language models.
Not currently ranked — collecting fresh signals.
star history
GPTeacher provides multiple instruction datasets designed for fine-tuning LLMs. The General-Instruct dataset contains ~20,000 examples including chain-of-thought reasoning and logic puzzles. The Roleplay-Instruct dataset (including a V2 version 2.5x larger) simulates diverse character interactions. The Code-Instruct dataset offers ~5,350 coding tasks across multiple languages. The Toolformer dataset prepares models for tool-use scenarios. All datasets follow Alpaca’s instruction-input-output format for compatibility with standard fine-tuning pipelines.
Frequently asked
- What is teknium1/GPTeacher?
- A collection of modular training datasets generated by GPT-4 for fine-tuning instruction-following language models.
- Is GPTeacher open source?
- Yes — teknium1/GPTeacher is open source, released under the MIT license.
- What language is GPTeacher written in?
- teknium1/GPTeacher is primarily written in Python.
- How popular is GPTeacher?
- teknium1/GPTeacher has 1.7k stars on GitHub.
- Where can I find GPTeacher?
- teknium1/GPTeacher is on GitHub at https://github.com/teknium1/GPTeacher.