Skip to content

Dataset · tabtoyou

Korean LLaVA instruction dataset for multimodal LLMs

tabtoyou/KoLLaVA-Instruct-150k

150,000 dialogues and detailed image descriptions in Korean for SFT of vision-language models.

What's inside

This dataset is a Korean translation of the LLaVA Instruct 150K dataset performed using DeepL. It contains between 100,000 and 1 million text examples featuring dialogues between users and assistants about objects in images, their quantities and locations, as well as detailed scene descriptions.

What you can do with it

  • Supervised fine-tuning of vision-language models for image question answering in Korean
  • Training assistants to generate detailed image descriptions in Korean
Tasks
Visual question answering, Question answering
Languages
Korean
Size
100K–1M rows
Downloads this month
62
Likes
23
Access
on request on Hugging Face

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("tabtoyou/KoLLaVA-Instruct-150k")