Dolci-Think-SFT-Olmo-Hybrid
收藏资源简介:
Dolci Think SFT Olmo Hybrid 是一个面向人工智能训练的大规模对话数据集,采用 Open Data Commons Attribution License (ODC-By) 许可,明确限定用于研究和教育用途。数据集包含 2,932,239 条训练样本,总大小约 86.5GB。每条数据记录包含以下核心字段:1) 'messages' 字段(结构化列表形式),内含对话内容('content')、角色标识('role')、函数调用信息('function_calls')和函数定义('functions');2) 唯一标识符('id');3) 数据来源标识('source')。该数据集适用于对话系统训练、函数调用理解等自然语言处理任务,其混合式数据结构支持多轮对话和功能性指令的联合建模。
Dolci Think SFT Olmo Hybrid is a large-scale conversational dataset tailored for artificial intelligence training, licensed under the Open Data Commons Attribution License (ODC-By), which explicitly restricts its usage to research and educational purposes. The dataset comprises 2,932,239 training samples with a total size of approximately 86.5 GB. Each data record includes the following core fields: 1) the 'messages' field (in the form of a structured list), which contains conversational content ('content'), role identifier ('role'), function call information ('function_calls'), and function definitions ('functions'); 2) a unique identifier ('id'); 3) a data source identifier ('source'). This dataset is applicable to natural language processing tasks such as conversational system training and function call understanding, and its hybrid data structure supports joint modeling of multi-turn conversations and functional instructions.
Dolci-Think-SFT-Olmo-Hybrid 数据集概述
数据集基本信息
- 数据集名称: Dolci-Think-SFT-Olmo-Hybrid
- 发布者: allenai
- 许可证: Open Data Commons Attribution License v1.0 (ODC-By)
- 预期用途: 研究与教育用途
- 责任使用指南: https://allenai.org/responsible-use
数据规模与结构
- 总下载大小: 38,822,134,509 字节
- 数据集大小: 86,495,338,231 字节
- 数据分割: 仅包含训练集(train)
- 训练集样本数量: 2,932,239 条
- 训练集数据大小: 86,495,338,231 字节
数据特征
数据集中每个样本包含以下特征:
- id: 字符串类型,样本唯一标识符
- source: 字符串类型,数据来源标识
- messages: 列表类型,包含多条消息,每条消息具有以下字段:
- content: 字符串类型,消息内容
- role: 字符串类型,消息角色
- function_calls: 字符串类型,函数调用信息
- functions: 字符串类型,函数定义信息
数据文件
- 训练集文件路径模式:
data/train-*



