Skip to content

Dataset · Infi-MM

InfiMM-WebMath-40B multimodal math corpus

Infi-MM/InfiMM-WebMath-40B

Large multimodal dataset of math problems with text and images in English and Chinese.

What's inside

Designed for mathematical reasoning, the dataset contains texts and images extracted from web documents. According to metadata, the size is 10M to 100M examples; languages are English and Chinese. The card indicates that the data is open-source and created for pre-training multimodal models.

What you can do with it

  • Pre-training multimodal models for mathematics
  • Fine-tuning image-text-to-text models on image-based tasks
  • Testing models on math problems with diagrams and formulas
Tasks
Image text → text
Languages
English, Chinese
Size
10M–100M rows
Downloads this month
2.3K
Likes
69

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("Infi-MM/InfiMM-WebMath-40B")