tudor-iustin22/myriad-1s
收藏资源简介:
MYRIAD-1S是一个开源的全栈代码数据集,由Tudor Iustin发布。它旨在支持代码生成、Web开发、监督微调、指令调优、数据集研究和评估工作流,适用于具备代码能力的AI系统。该数据集包含16,000个全栈代码样本,总计约102M个令牌,专注于现实软件构建任务,包括Web应用程序、产品界面、仪表板、浏览器游戏、开发工具和响应式全栈体验。数据集以Parquet格式发布,包含三列:query_ID(唯一标识符)、input(用户指令或任务提示)和output(生成的响应,包括实现、解释和代码内容)。它强调完整的实现模式,而非短代码片段,覆盖现代前端和Web应用框架(如Next.js、React、TypeScript、Tailwind CSS等),并注重高质量UI/UX、边缘案例和验证行为。数据集使用Creative Commons Attribution 4.0 International许可证,要求在使用时进行署名。
MYRIAD-1S is an open source full-stack code dataset released by Tudor Iustin. It is designed to support code generation, web development, supervised fine-tuning, instruction tuning, dataset research, and evaluation workflows for code-capable AI systems. The dataset contains 16,000 full-stack code samples totaling approximately 102M tokens. It focuses on realistic software-building tasks, including web applications, product interfaces, dashboards, browser games, developer tools, and responsive full-stack experiences. The dataset is released in Parquet format with three columns: query_ID (unique sample identifier), input (user-facing coding instruction or task prompt), and output (generated response containing the implementation, explanation, and code content). It emphasizes complete implementation patterns rather than short code snippets, covering modern frontend and web application frameworks (e.g., Next.js, React, TypeScript, Tailwind CSS), and prioritizes high-quality UI/UX, practical edge cases, and validation behavior. The dataset is licensed under Creative Commons Attribution 4.0 International, requiring attribution when used.





