uukuguy/MindSpeed-Infinity-Instruct-Gen

Infinity Instruct数据集是基于Infinity Instruct项目构建的，旨在为MindSpeed-LLM提供多轮对话微调格式的高质量指令数据集。它包含数百万条指令，分为基础数据集和聊天数据集，支持英语和中文等多种语言。数据集通过从开源数据集中选择和迭代指令构建而成，可用于构建自定义指令数据集，并提供了自我构建的指令标注系统和奖励模型。

The Infinity Instruct dataset is built upon the Infinity Instruct project, aiming to provide a high-quality instruction dataset for multi-round dialogue fine-tuning with MindSpeed-LLM. It contains millions of instructions, divided into foundational and chat datasets, supporting multiple languages including English and Chinese. The dataset is constructed by selecting and iterating instructions from open-source datasets, and it can be used to build custom instruction datasets with a self-built instruction labeling system and reward model.