4 Tutorial Jupyter Notebooks

The notebooks have been tested in VSCode. This settings file ensures notebooks start with the project root as the working directory.

Warning

If using an editor other than VSCode, ensure the Jupyter kernel’s root directory matches the project root.

4.1 Pipeline Verification Notebooks

4.1.1 Inspect Expert Output

Notebook: notebooks/inspect_expert_output.ipynb

Run the expert agent, produce data samples, and verify correct operation.

4.1.2 Load Pre-trained Model and Run Offline Inference

Notebook: notebooks/carla_offline_inference.ipynb

Load model checkpoints, visualize random data samples, and run offline inference on collected data.

4.2 Understanding Data Format

Notebook: notebooks/data_format.ipynb

Explore the dataloader output structure and learn how to adapt it for custom models.

4.3 Debug Closed-Loop Evaluation

Notebook: notebooks/inspect_sensor_agent_io.ipynb

Interactively debug model inputs and outputs during closed-loop evaluation to diagnose issues.