Dataset · BAAI
Infinity-Instruct: a large instruction dataset
BAAI/Infinity-Instruct
Millions of English and Chinese instructions for LLM fine-tuning and evaluation.
What's inside
Infinity-Instruct by BAAI is a large text-generation instruction dataset in English and Chinese. Volume ranges from 10 to 100 million examples, with the card emphasizing the importance of instruction data quality and scale. Formats are tabular and text.
What you can do with it
- Supervised fine-tuning of a large instruct model on millions of examples
- LoRA fine-tuning for English and Chinese assistants
- Evaluating model robustness across different instruction types
- Tasks
- Text generation
- Languages
- English, Chinese
- Size
- 10M–100M rows
- Downloads this month
- 11.7K
- Likes
- 771
- Access
- on request on Hugging Face
Models trained on it
- BAAI/Infinity-Instruct-7M-Gen-Llama3_1-8B 17.3K
- BAAI/Infinity-Instruct-3M-0625-Llama3-8BText generation 9.5K
- BAAI/Infinity-Instruct-3M-0625-Yi-1.5-9BText generation 9.4K
- Zyphra/Zamba2-1.2B-instructText generation 7.6K
- mradermacher/Hermes-2-Pro-Mistral-7B-infinity-i1-GGUF 1.4K
- mradermacher/Infinity-Instruct-7M-Gen-mistral-7B-i1-GGUF 1.3K
- mradermacher/Infinity-Instruct-7M-0729-mistral-7B-i1-GGUF 1.3K
- mradermacher/Antlia-Learn-7B-i1-GGUFText generation 1.1K
How to load it
from datasets import load_dataset
ds = load_dataset("BAAI/Infinity-Instruct")