知识卡片
模型即代码与回滚
发布模型与发布代码风险等价:模型也是代码,新模型可能冲垮 serving、甚至影响训练(如迁移学习)。发布原则——慢发布(先小比例用户/服务器,双向限制损害与信心)、发布而非重构(一次尽量只改一点,便于定位问题)、数据层隔离(新旧代码路径的日志格式须隔离,否则诊断极难)。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 发布
原书出处 · 1 处
- Introduction · 查看原书依据
Models as code Remember that models are code every bit as much as your training system binaries, serving path, and data processing code are. Deploying a new model can most definitely crash your serving system and ruin your online recommendations.
参考来源
- 位置:《Reliable Machine Learning》第01章《Introduction》(源文件:extracted/Reliable Machine Learning/_epub-src/OEBPS/ch01.xhtml)
- 结论依据:原文将模型与代码并列,提出慢发布、发布而非重构、数据层隔离三条发布准则。
- 原始内容:Models as code Remember that models are code every bit as much as your training system binaries, serving path, and data processing code are. Deploying a new model can most definitely crash your serving system and ruin your online recommendations.