知识卡片
训练推理与微调分工
模型可视为由数值参数决定输入输出映射的计算组件。训练用数据调整参数,微调在已有模型上继续适配,推理用训练所得参数处理新输入;提示则是在一次使用时提供任务和上下文。训练目标与方法不止书中输入输出对示例,模型参数规模也不是所有模型都达万亿。
这张卡片看不懂?
这张卡片的反馈已满 10 条,暂时无法继续提交。
已有反馈 · 0 / 10 条
还没有人提交反馈。
在整书笔记中的位置
原书出处 · 1 处
- Chapter 15. AI Inference and Serving · 查看原书依据
inference is the process of evaluating the model on user inputs where the output is unknown. Inference is typically implemented as a RESTful API, which you can call with the user’s input. Many cloud providers provide inference as a service. Such solutions can make it significantly easier to
参考来源
- 位置:《Designing Distributed Systems, 2nd Edition》第15章《Chapter 15. AI Inference and Serving》(来源项目 book-sources;源文件:extracted/Designing Distributed Systems, 2nd Edition/_epub-src/OEBPS/ch15.html)
- 结论依据:原文训练、微调、推理与提示定义提供关系,限制其简化训练与规模描述。
- 原始内容:inference is the process of evaluating the model on user inputs where the output is unknown. Inference is typically implemented as a RESTful API, which you can call with the user’s input. Many cloud providers provide inference as a service. Such solutions can make it significantly easier to