Module 2 · Datasets & Training

Lesson 3The Data Side of Fine-Tuning

6 topics18 min readTopics 12–17

Welcome to Module 2: Datasets & Training — where you stop understanding models and start shaping them. Everything in this module hangs off one picture: the three stages a model passes through, and which kind of dataset feeds each one.

The three-stage pipeline: raw documents build the base model in pretraining, instruction pairs teach it to answer in SFT, and preference pairs polish its judgment in alignment.
The three-stage pipeline: raw documents build the base model in pretraining, instruction pairs teach it to answer in SFT, and preference pairs polish its judgment in alignment.

You already met this pipeline conceptually in Lesson 2. Here's the reframe that makes this module matter: when you fine-tune, you are personally running stage 2 (and optionally stage 3) of this pipeline — just on your own data, at a scale of thousands of examples instead of trillions of tokens. Same mechanism, your steering wheel.

In this lesson

  1. 12Fine-Tuning Basicswhen to reach for it (and when not to)Fine-tuning = taking an already-trained model and continuing training on a small, focused dataset, so the knobs shift slightly toward your…3 min
  2. 13SFT Datasetsthe raw materialAn SFT dataset (Supervised Fine-Tuning dataset) is a collection of demonstrations: here's an input, here's exactly the output we want.…3 min
  3. 14Instruction Tuningwhy the transformation worksInstruction tuning is the specific, most important flavor of SFT: training on (instruction → response) pairs so a base model stops…3 min
  4. 15Dataset Formattingwhere fine-tunes silently dieThe least glamorous topic in this module, and the cause of the majority of "my fine-tuned model is broken" posts. Attend closely. Storage…3 min
  5. 16Preference Datasetsteaching "better," not just "good"SFT has a structural blind spot: it only ever shows positive examples. But much of what makes a model good is relative — this answer is…3 min
  6. 17Synthetic Datasetsmanufacturing your training dataEverything so far assumed you have examples. Usually you have fifty, and need five thousand. The modern answer: use a strong LLM to…5 min

Start with Topic 12: Fine-Tuning Basics, or browse all 90 topics.