Fine-Tuning & Model Adaptation

Whether to fine-tune or retrieve: when a model needs new behaviour rather than new facts, and how to adapt one through instruction tuning, data preparation, training and evaluation.

Requirement: Optional
For whom: ML engineers and anyone weighing fine-tuning against retrieval
Notebooks: 13
Before this track: Knowledge Graphs & GraphRAG
By: Sourav Banerjee

Objectives

Decide whether a problem needs fine-tuning or retrieval, and argue it from a measured comparison
Prepare instruction data, train a small model and evaluate it against a retrieval baseline

Fine-tuning changes how a model behaves, not what it knows, and picking it for the wrong reason costs weeks.

T14.1
Fine-Tuning & Model Adaptation
9 lessons
9 lessons

Exercise: Fine-tune a small model on refund and billing questions and score it against the RAG baseline
Done when: You can answer every module's Interview questions without notes

Objectives

Fine-Tuning & Model Adaptation1 module

Fine-Tuning & Model Adaptation