Query-of-CC/Retrieve-Pile
收藏Hugging Face2025-02-07 更新2025-02-15 收录
下载链接：
https://hf-mirror.com/datasets/Query-of-CC/Retrieve-Pile
下载链接
链接失效反馈官方服务：
资源简介：
Retrieve-Pile是一个基于Retrieve-from-CC方法构建的高质量知识数据集，包含数学推理数据以及生物学、物理学等多个领域的丰富知识语料。该数据集总大小为735GB，共有188B tokens，使用Llama2 tokenizer进行编码。
Retrieve-Pile is a high-quality knowledge dataset constructed using the Retrieve-from-CC method, containing mathematical reasoning data as well as rich knowledge corpora from fields such as biology, physics, etc. The dataset is 735GB in size with a total of 188B tokens, encoded using the Llama2 tokenizer.
提供机构：
Query-of-CC
