知识卡片
流水线必须提供质量反馈
只构建和搬运模型的流水线无法自动说明改动质量。把软件测试、数据检查、训练冒烟、模型指标和行为测试接到相应阶段,失败时阻断发布并给出可诊断信息。拥有CI工具不等于已经持续集成;还要看团队是否频繁合并,以及检查覆盖了哪些风险。
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
原书出处 · 1 处
- · 查看原书依据
CI/CD pipelines without tests is a contradiction in terms—how can we continuously integrate (CI) code to the main branch if we don’t have automated tests to check for errors? Yet it’s common to see teams with this smell—possibly because ML engineers know how to set up CI pipelines and data scientists know how to train and evaluate models, but not all teams have worked out
参考来源
- 位置:《Effective Machine Learning Teams (David Tan, Ada Leung, David Colls)》第09章《MLOps与持续交付》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Effective Machine Learning Teams (David Tan, Ada Leung, David Colls)/_epub-src/OEBPS/ch09.html)
- 结论依据:摘录中关于流水线必须提供质量反馈的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。
- 原始内容:CI/CD pipelines without tests is a contradiction in terms—how can we continuously integrate (CI) code to the main branch if we don’t have automated tests to check for errors? Yet it’s common to see teams with this smell—possibly because ML engineers know how to set up CI pipelines and data scientists know how to train and evaluate models, but not all teams have worked out