SB
收藏资源简介:
该数据集是一个大规模训练集,包含9,498,083个样本,总数据量约为378.6 GB。数据以结构化格式组织,核心特征为messages,这是一个列表类型字段,其中每个元素包含三个字符串字段:content(内容)、role(角色)和reasoning_content(推理内容)。此外,每个样本还包含一个source(来源)字符串字段,可能用于标识数据出处。数据集仅提供训练分割,下载大小约为378.4 GB。从字段名称推断,数据可能涉及多轮对话或指令遵循任务,且包含推理过程内容,适用于语言模型训练、推理能力增强或对话生成等场景。
This dataset is a large-scale training set containing 9,498,083 samples with a total data volume of approximately 378.6 GB. The data is organized in a structured format, with the core feature being messages, which is a list-type field where each element contains three string fields: content, role, and reasoning_content. Additionally, each sample includes a source string field, possibly used to identify the data origin. The dataset only provides a training split, with a download size of approximately 378.4 GB. Based on the field names, the data may involve multi-turn dialogue or instruction-following tasks and includes reasoning content, making it suitable for scenarios such as language model training, reasoning enhancement, or dialogue generation.
数据集概述
- 数据集名称:SB
- 数据集地址:https://huggingface.co/datasets/M1keR/SB
- 配置:default
- 数据文件:训练集数据位于
data/train-*路径下
数据集特征
数据集包含以下特征:
- messages:列表类型,包含三个字段:
content(字符串):消息内容role(字符串):角色标识reasoning_content(字符串):推理内容
- source(字符串):数据来源
数据集划分与规模
- 训练集:
- 样本数:9,498,083
- 数据大小:378,598,590,603 字节(约352.5 GB)
- 总下载大小:378,430,017,167 字节(约352.4 GB)
- 总数据集大小:378,598,590,603 字节(约352.5 GB)




