oieieio/smollm-corpus
收藏Hugging Face2025-08-19 更新2025-08-30 收录
下载链接：
https://hf-mirror.com/datasets/oieieio/smollm-corpus
下载链接
链接失效反馈官方服务：
资源简介：
SmolLM-Corpus是一个经过精心挑选的高质量教育和合成数据集，旨在用于训练小型语言模型。Cosmopedia v2是最大的合成预训练数据集，包含超过3900万本教科书、博客文章和故事。Python-Edu子集包含评分达到4或更高的Python文件。FineWeb-Edu-Dedup是一个去重后的教育网页数据集，包含了2200亿个高质量教育网页的标记。
SmolLM-Corpus is a curated collection of high-quality educational and synthetic data designed for training small language models. Cosmopedia v2 is the largest synthetic pre-training dataset, consisting of over 39 million textbooks, blog posts, and stories. The Python-Edu subset contains Python files scored 4 or higher by the educational code model. FineWeb-Edu-Dedup is a deduplicated dataset of educational web pages containing 220 billion tokens of high-quality educational content.
提供机构：
oieieio
