Map: SFT, LoRA, QLoRA, and DPO

• Distinguish prompt engineering, RAG, SFT, LoRA, QLoRA, and DPO.
• Choose the right technique based on cost, data, and goal.
• Avoid training a model when you only need to retrieve information.

Quick decision

• Prompt: you want to change format, tone, or simple instructions.
• RAG: you want it to use updatable or private documents.
• SFT: you want it to learn question-answer patterns from your domain.
• LoRA: you want to train few parameters and save a lightweight adapter.
• QLoRA: you want LoRA using quantization to reduce memory.
• DPO: you want to align preferences by comparing good and bad responses.

Decision card

objetivo: "respond to support emails with brand tone"
datos_disponibles:
  ejemplos_buenos: 800
  ejemplos_malos: 120
  documentos_actualizables: true
mejor_opcion:
  - RAG for policies that change
  - SFT/LoRA for tone and format
no_entrenar_para:
  - memorizing prices
  - storing private contracts
  - replacing permissions