fineweb-edu HuggingFaceFW

๐Ÿ“š FineWeb-Edu 1.3 trillion tokens of the finest educational data the ๐ŸŒ web has to offer Paper: https://arxiv.org/abs/2406.17557 What is it? ๐Ÿ“š FineWeb-Edu dataset consists of 1.3T tokens and 5.4T tokens (FineWeb-Edu-score-2) of educational web pages filtered from ๐Ÿท FineWeb dataset. This is the 1.3 trillion version. To enhance FineWeb's quality, we developed an educational quality classifier using annotations generated by LLama3-70B-Instruct. Weโ€ฆ See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceFW/fineweb-edu.

Type
dataset
License
odc-by
Language
en
Downloads
403,783
Likes
1,258
Access
public
Files
0

README

--- license: odc-by task_categories: - text-generation language: - en pretty_name: FineWeb-Edu size_categories: - n>1T configs: - config_name: default data_files: - split: train path: data/*/* features: - name: text dtype: string - name: id dtype: string - name: dump dtype: string - name: url dtypeโ€ฆ

ๆŸฅ็œ‹ๅฎŒๆ•ด้กต้ข ยท ๆŸฅ็œ‹ๅŽŸๆ–‡