toolcaller-train-mix
收藏资源简介:
该数据集是一个结构化的对话数据集,包含1,786个对话样本,分为训练集(1,608个样本)和验证集(178个样本)。每个样本包含两个核心字段:1) messages字段,为列表结构,每个列表元素包含role(角色标识,如用户、助手等)和content(对话内容)两个子字段,表明数据以多轮对话形式组织;2) source字段,为字符串类型,可能用于标识数据来源或类别。数据集总大小约13.8MB。基于其数据结构,该数据集适用于对话系统训练、指令遵循模型微调、对话生成等自然语言处理任务。
This is a structured dialogue dataset comprising 1,786 dialogue samples, which are split into a training set (1,608 samples) and a validation set (178 samples). Each sample contains two core fields: 1) The `messages` field, which is a list structure. Each element in the list includes two sub-fields: `role` (role identifier, such as user, assistant, etc.) and `content` (dialogue content), indicating that the data is organized in multi-turn dialogue format; 2) The `source` field, which is a string-type field that can be used to identify the data source or category. The total size of the dataset is approximately 13.8 MB. Based on its data structure, this dataset is suitable for natural language processing tasks such as dialogue system training, instruction-following model fine-tuning, and dialogue generation.
数据集概述:toolcaller-train-mix
该数据集用于工具调用相关任务,包含训练集和验证集两部分。
数据集结构
-
特征(Features):
messages:包含对话消息列表,每条消息包含:role(字符串):消息角色(如用户、助手等)content(字符串):消息内容
source(字符串):数据来源
-
数据划分(Splits):
- 训练集(train):1,608 条样本,大小约 12.45 MB
- 验证集(validation):178 条样本,大小约 1.36 MB
数据集规模
- 总下载大小:约 13.79 MB
- 总数据集大小:约 13.81 MB
配置文件
- 默认配置(
default)包含:- 训练集:
data/train-* - 验证集:
data/validation-*
- 训练集:




