Skip to content

Dataset · utter-project

Multilingual Web Text Corpus for EuroLLM-22B Training

utter-project/EuroWeb-2512

A massive corpus of filtered multilingual web data ranging from 1 billion to 10 billion records for language model pretraining.

What's inside

Includes multilingual text and tabular web data from various sources. The dataset size ranges from 1 billion to 10 billion records. Materials underwent standard cleaning and classification using the utter-project/EuroFilter-v1 filter for the EuroLLM-22B model.

What you can do with it

  • Pretraining and fine-tuning multilingual LLMs
  • Training web text quality classifiers
  • Researching multilingual corpus filtering
Size
1B–10B rows
Downloads this month
20.9K
Likes
6

How to load it

from datasets import load_dataset

ds = load_dataset("utter-project/EuroWeb-2512")