Dataset · LDJnr
Capybara: 10k multi-turn dialogues
LDJnr/Capybara
Over 10k multi-turn examples for instruction fine-tuning in English.
What's inside
The official Capybara dataset: over 10,000 multi-turn examples in English. Gathered based on ideas from Evol-instruct, Alpaca, Orca, Vicuna, Lamini, and FLASK; single-turn seeds were manually selected. The data focuses on dialogue diversity and logical robustness.
What you can do with it
- SFT of conversational LLMs
- Fine-tuning models to follow complex instructions
- Evaluating multi-turn responses
- Tasks
- Question answering, Text generation
- Languages
- English
- Size
- 10K–100K rows
- Downloads this month
- 1.3K
- Likes
- 270
Models trained on it
- TheBloke/dolphin-2.7-mixtral-8x7b-GGUF 14.3K
- dphn/dolphin-2.6-mixtral-8x7bText generation 9.4K
- Weyaxi/Einstein-v6.1-Llama3-8BText generation 9.3K
- jondurbin/bagel-dpo-34b-v0.2Text generation 9.2K
- jondurbin/bagel-dpo-34b-v0.5Text generation 9.2K
- jondurbin/bagel-8b-v1.0Text generation 9.2K
- jondurbin/airoboros-70b-3.3Text generation 9.1K
- jondurbin/airoboros-dpo-70b-3.3Text generation 9.1K
How to load it
from datasets import load_dataset
ds = load_dataset("LDJnr/Capybara")