FineWeb-Edu-Ar

FineWeb-Edu-Ar is a machine-translated Arabic version of the FineWeb-Edu dataset designed to support the development of Arabic small language models (SLMs).

Dataset Details:

• Languages: Arabic, English (paired)
• Size: 202 billion tokens
• License: CC-BY-NC-4.0
• Source: Machine-translated from the deduplicated version of Hugging Face’s FineWeb-Edu dataset
• Translation model: facebook/nllb-200-distilled-600M

Application:

FineWeb-Edu-Ar is suitable for pre-training Arabic language models, especially those focused on general knowledge and common-sense reasoning. Researchers and developers can utilize this dataset to enhance the performance of Arabic SLMs across various NLP tasks.

Python Usage:

To load and utilize the FineWeb-Edu-Ar dataset in Python, you can use the datasets package:

from datasets import load_dataset

dataset = load_dataset("kaust-generative-ai/fineweb-edu-ar", "ar", streaming=True)

print(next(iter(dataset["train"])))

Citation:

If you use FineWeb-Edu-Ar in your research, please cite the following paper:

https://arxiv.org/abs/2411.06402

@techreport{alrashed2024finewebeduar,
  author      = {Sultan Alrashed and Dmitrii Khizbullin and David R. Pugh},
  title       = {{FineWeb-Edu-Ar: Machine-translated Corpus to Support Arabic Small Language Models}},
  institution = {{King Abdullah University of Science and Technology (KAUST)}},
  year        = {2024},
  number      = {arXiv:2411.06402},
  url         = {https://arxiv.org/abs/2411.06402}
}