Files
nexus/wiki/concepts/Fine-Tuning.md

41 lines
1.5 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
---
title: "Fine-Tuning"
type: concept
tags: [AI, ML, fine-tuning, foundation-model, customization]
sources: [public-cloud-learning-sessions-opentext-ai-use-cases-20241126-160106-meeting-rec]
last_updated: 2026-05-12
---
## Aliases
- Fine-tuning
- Model Fine-tuning
- 模型的微调
## Definition
Fine-Tuning微调是在预训练基础模型之上使用特定领域或任务的数据进一步训练模型使其适应特定业务场景。与 RAG 不同,微调直接修改模型权重,而非在推理时注入外部知识。
## Key Facts
- 属于三大数据整合方法之一RAG / Fine-tuning / Continued Pre-training
- 与 RAG 的核心区别RAG 保留原始模型权重通过检索增强回答Fine-tuning 修改模型权重,改变模型本身
- 适用场景:特定领域术语、风格、任务类型的深度适配
- 成本:需要额外的训练资源和时间
- AWS Amazon Bedrock 支持 Fine-tuning 基础模型
## Comparison with RAG
| 维度 | Fine-Tuning | RAG |
|------|-------------|-----|
| 修改模型权重 | 是 | 否 |
| 推理延迟 | 无额外延迟 | 有检索开销 |
| 外部知识库 | 不依赖 | 依赖 |
| 适用场景 | 风格/任务适配 | 知识密集型问答 |
| 成本 | 训练成本高 | 索引/检索成本 |
## Related Concepts
- [[Foundation-Models]]:微调作用于基础模型
- [[RAG]]:另一种数据整合方法,与微调互补
- [[Amazon-Bedrock]]:提供 Fine-tuning 能力
## Sources
- [[public-cloud-learning-sessions-opentext-ai-use-cases-20241126-160106-meeting-rec]]