知识卡片

普通读书笔记卡 · 8760.c · 生成模型:GPT-6

非功能预算也可设为门禁

理解等级:中级

模型大小、并发延迟和训练耗时可按部署和反馈预算设阈值,及早发现资源回退。测量需固定硬件、负载和统计口径;偶发慢一次不必是代码退化,预算也不代替模型效果。

这张卡片看不懂?

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

      Training duration tests The model training pipeline must complete within a specified duration. This can help teams detect and prevent the gradual lengthening of model training cycles from, say, two hours to three or four hours. This test helps teams detect performance degradations as soon as they are introduced—e.g., as part of a pull request—which makes it easier to debug and identify the change that

    参考来源

    - 位置:《Effective Machine Learning Teams (David Tan, Ada Leung, David Colls)》第06章《模型自动化测试》(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/ch06.html) - 结论依据:摘录中关于非功能预算也可设为门禁的定义、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:Training duration tests The model training pipeline must complete within a specified duration. This can help teams detect and prevent the gradual lengthening of model training cycles from, say, two hours to three or four hours. This test helps teams detect performance degradations as soon as they are introduced—e.g., as part of a pull request—which makes it easier to debug and identify the change that