Skip to content

Dataset · BAAI

Millions of GPT-4-generated visual instructions

BAAI/SVIT

A large English dataset for training multimodal models (VQA) on instructions from GPT-4.

What's inside

Contains 1 to 10 million English examples for visual-question-answering (VQA) tasks. The data was generated using GPT-4 to scale visual instruction tuning. Published by the BAAI research academy.

What you can do with it

  • Large-scale SFT training of multimodal models (VLMs) for image understanding
  • Adapting language models to answer questions about images
Tasks
Visual question answering
Languages
English
Size
1M–10M rows
Downloads this month
268
Likes
35
Access
on request on Hugging Face

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("BAAI/SVIT")