Fine-tuning
A big model knows a lot, but it does not know your task. Fine-tuning trains the model a little more on your own examples, so it learns your style or your narrow job. It is powerful, but it should be your last step, not your first.
New words
- Fine-tuning
- Training a ready model a little more on your own examples.
- Weights
- The numbers inside a model. Training changes them.
- LoRA
- A way to fine-tune by training only a small set of extra numbers.
- Training set
- The examples the model learns from.
- Validation set
- Examples kept aside to check the model on new data.
- Overfitting
- The model remembers its examples but fails on new ones.
- Few-shot prompt
- A prompt that includes a few examples of the task.
What you will learn
- The tailor and the ready-made shirt
- When is a prompt not enough?
- What changes: the weights or the input?
- The data you need
- Full fine-tuning and LoRA
- Train, validate, and avoid overfitting
- Try it: should I fine-tune?
- Cost, upkeep and the decision order
- In a real system, and interview tip
- Key takeaways
10 quiz questions at the end.
The full lesson, with interactive demos and the quiz, opens after you sign in with Google. It is free.