Exams / AIF-C01 / Training and fine-tuning / Question 2
AIF-C01 practice question: Training and fine-tuning (2 of 3)
Applications of Foundation Models A free question from our original AIF-C01 practice exam. Try it, then open the answer to see why each option is right or wrong.
A retailer uses a very large foundation model to classify customer emails, and the quality is excellent. However, responses are too slow and each request costs too much at peak volume. The retailer wants a model that keeps similar quality for this one task but is faster and cheaper. Which approach fits BEST?
Choose one answer.
- AA. Continued pre-training of the large model on all company emails
- BB. Raising the large model's temperature to shorten its answers
- CC. Pre-training a new model from scratch on public email datasets
- DD. Model distillation, using the large model as teacher for a smaller student
Show the answer and explanation
Answer: D
- A. Wrong: This makes the same large model more expert, but it stays just as large, slow and costly to run.
- B. Wrong: Temperature affects randomness, not the speed or cost of running the model.
- C. Wrong: Training from scratch is the most expensive option and is unlikely to match the large model's quality.
- D. Right: Correct. Distillation fine-tunes a smaller, faster, cheaper model on the larger model's responses so it performs well on the specific use case.
Explanation
Amazon Bedrock Model Distillation transfers knowledge from a larger teacher model to a smaller student model by generating responses from the teacher and fine-tuning the student on them. The result is a smaller, faster, more cost-effective model whose performance is improved for that use case. Validate it against your own evaluation data.
Study
Keep practising
Answer questions with a score at the end, or practise just this topic with the explanation after each one. No sign-up needed.