Skip to content

Dataset · teknium

OpenHermes: 242k SFT examples

teknium/openhermes

242k predominantly GPT-4 examples from open sources for instruction tuning.

What's inside

OpenHermes consists of 242,000 records predominantly generated by GPT-4 based on open datasets. The card mentions GPTeacher, Roleplay v1/v2, and Code Instruct. Language is English, size 100K–1M, format text.

What you can do with it

  • Train a chat model via SFT
  • Configure LoRA for instructions
  • Test instruction following
Tasks
Text generation
Languages
English
Size
100K–1M rows
Downloads this month
1.1K
Likes
245

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("teknium/openhermes")