知识卡片
RAG与微调可以组合
请求先检索租户资料增强提示,再选择对应微调模型或适配器,能同时利用外部知识和行为适配。二者职责不同,组合不是必需步骤,应根据所需体验评估;不能把微调当作实时知识库,也不能把RAG当作训练替代词。
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
- GenAI and Multi-Tenancy · 查看原书依据
request is submitted, your solution will need to augment the prompt, using RAG to inject tenant-specific context into the request (step 1). Then, it will need to determine which logical model it will invoke to apply tenant-specific fine-tuning (step 2). This will send the augmented prompt to the appropriate per-tenant fine-tuned model for processing, ideally yielding a more tenant-targeted response.
参考来源
- 位置:《Building Multi-Tenant SaaS Architectures》第16章《GenAI and Multi-Tenancy》(来源项目 a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Building Multi-Tenant SaaS Architectures/_epub-src/OEBPS/ch16.html)
- 结论依据:原文具体说明请求先检索租户资料增强提示,再选择对应微调模型或适配器,能同时利用外部知识和行为适配;本卡据此归纳机制及适用边界。
- 原始内容:request is submitted, your solution will need to augment the prompt, using RAG to inject tenant-specific context into the request (step 1). Then, it will need to determine which logical model it will invoke to apply tenant-specific fine-tuning (step 2). This will send the augmented prompt to the appropriate per-tenant fine-tuned model for processing, ideally yielding a more tenant-targeted response.