Fine-Tuning Small Language Models

Decide when to fine-tune a small model vs. use a frontier model, choose the right technique (LoRA, distillation, quantization), and design a hybrid routing architecture for production

4 sections ยท 11 lessons

Course outline

Why Small Models Matter in Production

  1. The cost and latency case
  2. Data privacy as a forcing function
  3. When small beats frontier

Fine-Tuning Techniques

  1. LoRA: fine-tune in an hour
  2. Distillation: learning from a larger model
  3. Quantization: shrinking model size
  4. RAG vs. fine-tuning: when to use each

Limitations and Failure Modes

  1. Where small models still fail
  2. Catastrophic forgetting

Production Architecture

  1. The 90/10 routing pattern
  2. Speculative decoding

Start learning with Wondering