Dataset · lmms-lab
Video question-answering dataset for video-LLM training
lmms-lab/LLaVA-Video-178K
Video and text pairs for training and evaluating video question-answering models.
What's inside
The dataset contains video and English text data for visual question-answering and video-text-to-text tasks. The card specifies 178,510 caption records; the collection was compiled from five main video-language data sources and was used to train LLaVA-Video. Permitted for academic research and education only.
What you can do with it
- Training LLaVA-Video and similar video-LLMs
- Evaluating models on video question answering
- Video-text-to-text research
- Tasks
- Visual question answering, Video text → text
- Languages
- English
- Size
- 1M–10M rows
- Downloads this month
- 106.1K
- Likes
- 215
Models trained on it
- HuggingFaceTB/SmolVLM2-500M-Video-InstructImage text → text 1.7M
- HuggingFaceTB/SmolVLM2-2.2B-InstructImage text → text 194.6K
- HuggingFaceTB/SmolVLM2-256M-Video-InstructImage text → text 44.7K
- lmms-lab/LLaVA-Video-7B-Qwen2Video text → text 12.3K
- DAMO-NLP-SG/VideoLLaMA3-7BVideo text → text 5.5K
- DAMO-NLP-SG/VideoLLaMA3-2BVideo text → text 2.6K
- mradermacher/SWIM-7B-i1-GGUF 2.6K
- HuggingFaceTB/SmolVLM2-2.2B-BaseImage text → text 2.4K
How to load it
from datasets import load_dataset
ds = load_dataset("lmms-lab/LLaVA-Video-178K")