LLMTeamAkiyama/cleand_microsoft_rStar-Coder

这是一个英文问答数据集，包含269,863个数据条目，每个条目的平均令牌数为11674，最大令牌数为31,184，总令牌数为3,150,447,484。数据集以JSONL格式存储，并经过了一系列的加工处理，包括使用合成SFT、基于令牌数的过滤、种子问题数量和生成数量的限制、去除不完整的think标签、令牌化处理以及重复内容的移除。

This is an English question-answering dataset containing 269,863 entries, with an average of 11,674 tokens per entry, a maximum of 31,184 tokens, and a total of 3,150,447,484 tokens. The dataset is stored in JSONL format and has undergone a series of processing steps, including the use of synthetic SFT, filtering based on token count, limits on the number of seed questions and generations, removal of incomplete think tags, tokenization processing, and elimination of duplicate content.