Dataset · mvp-lab
LLaVA-OneVision-1.5 multimodal instructions
mvp-lab/LLaVA-OneVision-1.5-Instruct-Data
22M image-text-to-text instructions for training and fine-tuning large multimodal models.
What's inside
Collected and merged during the development of LLaVA-OneVision-1.5. Features 22M image-text-to-text instructions with image and text modalities in English, designed for training large multimodal models.
What you can do with it
- SFT for LLaVA-OneVision-1.5
- Fine-tuning open VLMs
- Image-text-to-text instruction evaluation
- Tasks
- Image text → text
- Languages
- English
- Size
- 10M–100M rows
- Downloads this month
- 53.5K
- Likes
- 91
Models trained on it
- kyutai/CASA-Qwen2_5-VL-3BImage text → text 12.4K
- kyutai/CASA-Helium1-VL-2BImage text → text 205
- kyutai/Helium1-VL-2BImage text → text 68
- kyutai/CASA-Helium1-VL-2B-SharedImage text → text 40
- kyutai/CASA-Qwen2_5-VL-3B-SharedImage text → text 36
- KU-MIIL/CRIT-VL-8BImage text → text 26
- KU-MIIL/CRIT-VL-38BImage text → text 8
- finalpandas/CASA-Helium1-VL-2BImage text → text 7
How to load it
from datasets import load_dataset
ds = load_dataset("mvp-lab/LLaVA-OneVision-1.5-Instruct-Data")