Understanding RAG Part IX: Fine-Tuning LLMs for RAG