datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
ara-ebooks-heavy-8.3M
BabyLM Arabic - Modified Corpora
Five alternative ~8.3M-word compositions of the BabyLM Arabic corpus
(BabyLM-community/babylm-ara), built to address that dataset's over-reliance on
OpenSubtitles (42%) and Habibi songs (37%) - together almost 80% of the original corpus -
by substituting in Arabic e-book content instead. See
jumelet2025babybabellm for the original dataset
and the wider BabyBabelLM project.
All five keep a fixed ~20% base of the original corpus's most… See the full description on the dataset page: https://huggingface.co/datasets/kabaros/ara-ebooks-heavy-8.3M.ara-ebooks-movies5-8.3M
BabyLM Arabic - Modified Corpora
Five alternative ~8.3M-word compositions of the BabyLM Arabic corpus
(BabyLM-community/babylm-ara), built to address that dataset's over-reliance on
OpenSubtitles (42%) and Habibi songs (37%) - together almost 80% of the original corpus -
by substituting in Arabic e-book content instead. See
jumelet2025babybabellm for the original dataset
and the wider BabyBabelLM project.
All five keep a fixed ~20% base of the original corpus's most… See the full description on the dataset page: https://huggingface.co/datasets/kabaros/ara-ebooks-movies5-8.3M.ara-ebooks-habibi5-8.3M
BabyLM Arabic - Modified Corpora
Five alternative ~8.3M-word compositions of the BabyLM Arabic corpus
(BabyLM-community/babylm-ara), built to address that dataset's over-reliance on
OpenSubtitles (42%) and Habibi songs (37%) - together almost 80% of the original corpus -
by substituting in Arabic e-book content instead. See
jumelet2025babybabellm for the original dataset
and the wider BabyBabelLM project.
All five keep a fixed ~20% base of the original corpus's most… See the full description on the dataset page: https://huggingface.co/datasets/kabaros/ara-ebooks-habibi5-8.3M.gutenberg_selected_ebooks
Gutenberg selected ebooks dataset
This dataset is a collection of passages from ebooks handpicked from the Gutenberg Project.
These writings are:
Alice's Adventures in Wonderland
Pride and Prejudice
Romeo and Juliet
The Adventures of Sherlock Holmes
The Odyssey
Winnie-the-Pooh
Source
The texts of the passages were derived from a larger Gutenberg-based set: sedthh/gutenberg_english, which was sourced directly from the project's site.
Metadata
Each passage… See the full description on the dataset page: https://huggingface.co/datasets/enelpol/gutenberg_selected_ebooks.ebooks-multilingual