← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8742.b · 生成模型:GPT-6

容器性能应分阶段测量

理解等级:中级

比较宿主与容器时应固定工作、解释器和硬件,分别记录数据读取、预处理和训练。原书小样本测到的约1.7%不是普遍开销,I/O挂载、虚拟化和安全设置可能改变结果,先测再决定优化。

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

      Measure before optimizing. A benchmark analysis ran a computationally intensive deep learning task both inside and outside of containers and found no noticeable differences in performance. Another benchmark analysis comparing the performance impact of running various deep learning models (e.g., InceptionV3, VGG16) in containers and on host found that performance differences are close to zero.

    参考来源

    - 位置:《Effective Machine Learning Teams (David Tan, Ada Leung, David Colls)》第03章《依赖管理的原则与工具》(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/ch03.html) - 结论依据:摘录中关于容器性能应分阶段测量的定义、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:Measure before optimizing. A benchmark analysis ran a computationally intensive deep learning task both inside and outside of containers and found no noticeable differences in performance. Another benchmark analysis comparing the performance impact of running various deep learning models (e.g., InceptionV3, VGG16) in containers and on host found that performance differences are close to zero.