Deep Learning (DL) · Advanced level

Advanced level · Module 4: Deploying Deep Learning Models · Lesson 1 of 4

Exporting Models

Convert trained PyTorch models into portable formats for other runtimes.

In this lesson you will

  • Export models with TorchScript and ONNX
  • Verify exported models give the same results

Key concepts

TorchScriptONNXruntimeexport verification

The lesson

Training code is not deployment code. Exporting converts a model into a format that runs without the original Python training setup: TorchScript for PyTorch runtimes, ONNX for many runtimes and devices.

Export with an example input and check that the exported model gives the same outputs as the original on test images, within a tiny tolerance.

Record the export format, version and input requirements in the model's documentation.

Step by step

  1. Export the model to ONNX.
  2. Run it with ONNX Runtime.
  3. Compare outputs with the PyTorch model.
  4. Document the input format.

Practice activity

Export the project model and verify it.

Expected result: An exported model whose outputs match the original.

Check your understanding

  1. Why export models for deployment?

    Show answer

    To run them without the training code and on other runtimes and devices.

  2. How do you verify an export?

    Show answer

    Compare its outputs with the original model on test inputs.

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.