coderforge-preview-unified-100000

该数据集包含100,000个训练样本，总大小约20.57GB（压缩下载大小5.06GB）。每个样本包含三个主要字段：1) 'conversations'字段为对话列表，每条对话包含'content'(文本内容，字符串类型)和'role'(角色标识，字符串类型)两个子字段；2) 'instance_id'字段为实例唯一标识符(字符串类型)；3) 'source'字段表示数据来源(字符串类型)。数据集仅提供训练集划分，未提供关于具体应用场景或收集背景的文本描述。This dataset contains 100,000 training samples, with a total size of approximately 20.57 GB (compressed download size: 5.06 GB). Each sample includes three main fields: 1) The 'conversations' field is a list of dialogues, where each dialogue contains two sub-fields: 'content' (text content, string type) and 'role' (role identifier, string type); 2) The 'instance_id' field is the unique identifier of the instance (string type); 3) The 'source' field represents the data source (string type). Only the training set split is provided for this dataset, and no textual descriptions regarding specific application scenarios or data collection background are available.