Stas Bekman

Arctic Long Sequence Training: Scalable And Efficient Training For Multi-Million Token Sequences

arXiv 2025

OBELICS: An Open Web-Scale Filtered Dataset of Interleaved Image-Text Documents

obelics-an-open-web-scale-filtered-dataset-of

What Language Model to Train if You Have One Million GPU Hours?

arXiv 2022

Datasets: A Community Library for Natural Language Processing

EMNLP (ACL) 2021 11

Victor Sanh

Alexander M. Rush

Lucile Saulnier

Teven Le Scao

Thomas Wang

Abhishek Thakur

Albert Villanova del Moral

Amanpreet Singh

Angelina McMillan-Major

Anton Lozhkov