知识卡片
模型对象存储要记录不可变版本
对象键提供模型制品入口,覆盖同一个键会使不同时间的加载取得不同内容。保留摘要、编码器、标签及兼容依赖,加载前验证来源;pickle/joblib还应只从可信制品读取,不能把公开可下载当安全保证。
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
Now that you know the basics of microservices and containers, it’s time to dive in and actually work with the prediction microservice. You need to train a model and save (dump) it to a joblib pickle file so that it can be reused again on other machines. A cloud service for object storage is a good choice for storing large files like pickle files. This
参考来源
- 位置:《Natural Language Processing in Action, Second Edition》附录E《附录E:容器化微服务部署》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Natural Language Processing in Action, Second Edition/_epub-src/OEBPS/Text/appendix-e.html)
- 结论依据:摘录中关于模型对象存储要记录不可变版本的定义、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:Now that you know the basics of microservices and containers, it’s time to dive in and actually work with the prediction microservice. You need to train a model and save (dump) it to a joblib pickle file so that it can be reused again on other machines. A cloud service for object storage is a good choice for storing large files like pickle files. This