ICS 35.240.50 CCS L 60 团体标准 T/CES 103—2022 电力人工智能边端侧模型技术规范 Technical specification for electric power artificial intelligence edge model 2022-06-22 发布 2022-06-24 实施 中国电工技术学会 发布 全国团体标准信息平台 T/CES 103—2022 I 目 次 前言 .................................................................................................................................................................... Ⅱ 1 范围 ............................................................................................................................... ................................. 1 2 规范性引用文件 ............................................................................................................................... ............. 1 3 术语和定义 ............................................................................................................................... ..................... 1 4 缩略语 ............................................................................................................................... ............................. 5 5 通用要求 ............................................................................................................................... ......................... 5 5.1 应用场景 ............................................................................................................................... .................. 6 5.2 硬件架构适配 ............................................................................................................................... .......... 6 5.3 模型超参转换 ............................................................................................................................... .......... 6 5.4 边端推理效率 ............................................................................................................................... .......... 6 6 通用算子要求 ............................................................................................................................... ................. 6 7 模型封装要求 ............................................................................................................................... ................. 6 7.1 模型初始化服务 ............................................................................................................................... ...... 7 7.2 数据管理服务 ............................................................................................................................... .......... 7 7.3 容器化模型打包 ............................................................................................................................... ...... 8 8 模型评价指标及方法 ............................................................................................................................... ..... 9 8.1 模型评价体系 ............................................................................................................................... .......... 9 8.2 安全性评价 ............................................................................................................................... .............. 9 8.3 鲁棒性评价 ............................................................................................................................... .............. 9 8.4 模型推理效率评价 ............................................................................................................................... .. 9 8.5 评价指标 ............................................................................................................................... ................ 11 8.6 评价方法 ............................................................................................................................... ................ 11 8.7 模型性能等级判定 ............................................................................................................................... 13 附录 A（资料性附录） 常见框架算子对照表 ............................................................................................ 17 附录 B（资料性附录） 模型评价指标计算 ..