Dataset · HuggingFaceM4
Documents and questions for Document VQA
HuggingFaceM4/Docmatix
Massive English dataset for Document VQA used in Idefics3 fine-tuning.
What's inside
Docmatix is a massive dataset for Document Visual Question Answering (Document VQA) used to fine-tune the Idefics3 vision-language model. The data consists of document images and text question-answer pairs in English, with a volume between 1M and 10M examples.
What you can do with it
- Fine-tune VLMs for document and scan-based question answering
- Train LoRA for Idefics3 on Document VQA
- Evaluate OCR/VQA systems on documents
- Tasks
- Visual question answering
- Languages
- English
- Size
- 1M–10M rows
- Downloads this month
- 26.8K
- Likes
- 319
Models trained on it
- HuggingFaceTB/SmolVLM2-500M-Video-InstructImage text → text 1.7M
- HuggingFaceTB/SmolVLM-500M-InstructImage text → text 258K
- HuggingFaceTB/SmolVLM-256M-InstructImage text → text 229.3K
- HuggingFaceTB/SmolVLM2-2.2B-InstructImage text → text 194.6K
- HuggingFaceTB/SmolVLM2-256M-Video-InstructImage text → text 44.7K
- HuggingFaceM4/Idefics3-8B-Llama3Image text → text 37.6K
- HuggingFaceTB/SmolVLM-InstructImage text → text 16.8K
- DAMO-NLP-SG/VideoLLaMA3-7BVideo text → text 5.5K
How to load it
from datasets import load_dataset
ds = load_dataset("HuggingFaceM4/Docmatix")