diffusion-chat-mixture-1024
收藏资源简介:
ChatMix_2M是一个大规模对话数据集,包含200万条训练样本和5000条验证样本。数据集采用结构化格式,每条样本包含三个主要字段:1) messages字段,这是一个消息列表,每条消息包含role(角色)和content(内容)两个子字段,表明数据以多轮对话形式组织;2) source字段,记录数据的来源信息;3) license字段,提供数据的许可信息。数据集总大小约6.6GB,下载大小约7.06GB。从数据结构推断,该数据集适用于对话系统开发、聊天机器人训练、语言模型对话能力微调等自然语言处理任务。数据以角色-内容对的形式组织,暗示其可能用于训练能够理解不同对话角色(如用户、助手等)的AI模型。
ChatMix_2M is a large-scale dialogue dataset consisting of 2 million training samples and 5,000 validation samples. It follows a structured format, where each sample includes three core fields: 1) The `messages` field, a list of messages where each message contains two sub-fields: `role` (speaker role) and `content` (message content), indicating the dataset is organized in a multi-turn dialogue structure; 2) The `source` field, which records the provenance information of the dataset; 3) The `license` field, which provides the licensing details of the dataset. The total size of the dataset is approximately 6.6 GB, with a download size of around 7.06 GB. Based on its data structure, this dataset is applicable to natural language processing tasks including dialogue system development, chatbot training, and fine-tuning of language models for dialogue capabilities. Organized as role-content pairs, the dataset is intended for training AI models that can comprehend various dialogue roles such as users and assistants.
数据集概述:Diffusion Chat Mixture 1024
基本信息
- 数据集名称:Diffusion Chat Mixture 1024
- 来源平台:Hugging Face Datasets
- 配置名称:chatmix_2m
- 许可证:未在 README 中指定,需参考原始数据集的 license 字段(字段值为字符串,具体内容未提供)
数据规模
- 总下载大小:约 7.06 GB(7,056,258,555 字节)
- 总数据集大小:约 6.60 GB(6,602,669,585 字节)
- 训练集:2,000,000 条样本,大小为 6.59 GB(6,590,364,035 字节)
- 验证集:5,000 条样本,大小为 12.31 MB(12,305,550 字节)
数据结构
数据集包含三个字段:
- messages:对话消息列表,每条消息包含两个字段:
role:字符串类型,表示消息的角色(如 user 或 assistant)content:字符串类型,表示消息的具体内容
- source:字符串类型,表示数据来源
- license:字符串类型,表示数据使用的许可证信息
数据划分
- train:训练集,包含 2,000,000 条样本
- validation:验证集,包含 5,000 条样本
数据文件
- 训练集文件路径:
chatmix_2m/train-* - 验证集文件路径:
chatmix_2m/validation-*




