train_3.5M_CN

该仓库提供约350万条中文指令数据，由BELLE项目生成，适用于文本到文本生成任务。数据以对话形式组织，包含多轮和单轮对话，每个样本包括数据ID和对话内容。此外，该仓库还提供了一个带有指令类别字段的增强版本。该数据集仅限研究目的使用，禁止商业用途及其他危害社会的应用。对于因使用该数据集造成的任何损害或纠纷，该仓库不承担任何责任。

This repository provides approximately 3.5 million Chinese instruction datasets generated by the BELLE project, which are suitable for text-to-text generation tasks. The data is organized in the form of dialogues, including both multi-turn and single-turn conversations. Each sample consists of a data ID and dialogue content. In addition, this repository also provides an enhanced version with an instruction category field. This dataset is strictly limited to research purposes, and commercial use and other socially harmful applications are prohibited. The repository shall not be liable for any damages or disputes caused by the use of this dataset.