Skip to content

Dataset · openbmb

Multi-turn dialogues generated by Turbo API

openbmb/UltraChat

Large English multi-turn dialogue dataset generated by two Turbo APIs.

What's inside

An open large-scale multi-turn dialogue dataset in English containing 100K-1M examples. For privacy, prompts are not sourced from the internet. Generation used two separate ChatGPT Turbo APIs: one created queries, the other answers.

What you can do with it

  • SFT of chat models on multi-turn dialogues
  • Instruction-tuning dialogue assistants
  • Evaluating chat response generation quality
Tasks
Text generation
Languages
English
Size
100K–1M rows
Downloads this month
3.9K
Likes
522

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("openbmb/UltraChat")