Deep Learning (DL) · Advanced level

Advanced level · Module 5: Responsible Deep Learning · Lesson 3 of 4

Compute Cost and Environmental Impact

Weigh the financial and environmental costs of training and running deep models.

In this lesson you will

  • Estimate compute cost and energy use
  • Choose efficient approaches that meet the need

Key concepts

training vs inference costenergy useefficient modelsproportionality

The lesson

Training large models uses a lot of electricity and money; running them for many users adds up too. Costs are higher in settings that rely on generators or expensive cloud services paid in foreign currency.

Prefer efficient choices: fine-tune small pre-trained models, use transfer learning, compress models and avoid unnecessary experiments.

Ask whether a deep model is needed at all — a simpler model or rule may meet the need at a fraction of the cost.

Step by step

  1. Estimate the GPU hours used in your projects.
  2. Estimate energy and cost.
  3. Identify the most wasteful steps.
  4. Propose efficiency changes.

Practice activity

Write a compute and cost note for the capstone plan.

Expected result: An estimate with concrete efficiency measures.

Check your understanding

  1. Name two ways to reduce deep learning compute cost.

    Show answer

    For example fine-tuning small pre-trained models and compressing models.

  2. Why consider simpler models?

    Show answer

    They may meet the need at much lower cost.

Taught live by an instructor — online or in person. Free course lessons are for self-study and do not lead to a certificate — see registration options.