知识卡片
特征存储及其 API 能力
特征存储是专门存放已抽取特征(及标签)值的系统,使模型训练乃至推理能快速、一致地读取。多数训练/服务系统都有类似组件,只是未必叫这个名字;它在大型、集中管理、多模型共享特征时最有价值。作者强调其最重要的特性是 API,至少要能:存放特征定义、存放特征值本身、高效提供特征数据(不能让昂贵的 CPU 或加速器空等 I/O)、与元数据系统协调写入。按数据形态分两类:结构化、可按列分解且常按时间排序的 Columns 适合列式存储;本质无序、非结构化的 Blobs 适合按字节高效存放。
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
- Feature and Training Data · 查看原书依据
Feature store A feature store is a storage system designed to store extracted feature (and label) values so that they can be quickly and consistently read during training a model and even during inference.
参考来源
- 位置:《Reliable Machine Learning》第04章《Feature and Training Data》(源文件:extracted/Reliable Machine Learning/_epub-src/OEBPS/ch04.xhtml)
- 结论依据:原文明确阐述该主题,并给出定义、实践或反例。
- 原始内容:Feature store A feature store is a storage system designed to store extracted feature (and label) values so that they can be quickly and consistently read during training a model and even during inference.