知识卡片
训练环境围绕数据与隔离选择
原书为避免各环境重复且不一致的全量训练,建议在可获得所需数据的环境训练,再提升同一模型制品。可取原则是控制训练版本与制品流转,不能把“只能在生产训练”当普遍规则;受控训练环境同样可获得合适数据,须按隔离、权限和运行影响设计。
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
run full model training in only one environment where you have access to the best data—in most cases, it’s production. You can still run training smoke tests on a small sample of data (as described in Chapter 6) in the other lower environments before running full training in production.
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:摘录中关于训练环境围绕数据与隔离选择的定义、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:run full model training in only one environment where you have access to the best data—in most cases, it’s production. You can still run training smoke tests on a small sample of data (as described in Chapter 6) in the other lower environments before running full training in production.