感谢您发送咨询！我们的团队成员将很快与您联系。
感谢您发送预订！我们的团队成员将很快与您联系。
课程大纲
Introduction to Retrieval-Augmented Generation (RAG)
- What is RAG and why it matters for enterprise AI
- Components of a RAG system: retriever, generator, document store
- Comparison with standalone LLMs and vector search
Setting Up a RAG Pipeline
- Installing and configuring Haystack or similar frameworks
- Document ingestion and preprocessing
- Connecting retrievers to vector databases (e.g., FAISS, Pinecone)
Fine-Tuning the Retriever
- Training dense retrievers using domain-specific data
- Using sentence transformers and contrastive learning
- Evaluating retriever quality with top-k accuracy
Fine-Tuning the Generator
- Selecting base models (e.g., BART, T5, FLAN-T5)
- Instruction tuning vs. supervised fine-tuning
- LoRA and PEFT methods for efficient updates
Evaluation and Optimization
- Metrics for evaluating RAG performance (e.g., BLEU, EM, F1)
- Latency, retrieval quality, and hallucination reduction
- Experiment tracking and iterative improvement
Deployment and Real-World Integration
- Deploying RAG in internal search engines and chatbots
- Security, data access, and governance considerations
- Integration with APIs, dashboards, or knowledge portals
Case Studies and Best Practices
- Enterprise use cases in finance, healthcare, and legal
- Managing domain drift and knowledge base updates
- Future directions in retrieval-augmented LLM systems
Summary and Next Steps
要求
- 理解自然语言处理（NLP）概念
- 具备基于transformer的语言模型经验
- 熟悉Python及基本机器学习工作流程
受众
- NLP工程师
- 知识管理团队
14 小时
