Self-paced course examples

• Optimize for learners opening and running a section independently. Each hands-on section keeps its own code/ directory beside its notebook, including its own supporting Python modules.
• Repeating model, data, or training code across sections is intentional. Do not consolidate examples into a parent directory or make one section import another section's implementation.
• Use job.py, client.py, and model.py for the main example files. Use prepare_data.py for data preparation helpers and download_data.py for scripts devoted to downloading data. Keep each section's implementation focused on its own lesson; include additional variants only when that lesson compares them.
• Notebook commands and source displays must use that section's own code. Verify the example runs and exports from its local directory.
• Training and export commands needed by later cells must be executable notebook cells using the selected kernel's Python. Keep a compact exit-status check so a failed command stops execution before later cells can read stale output. Validate notebooks from a fresh workspace without running those commands separately beforehand.
• Keep learner code focused on the APIs taught by each lesson. Put checkpoint assertions and result-integrity checks in validation, and use the printed run summary in introductory lessons. Avoid repetitive file-existence and expected-result assertions in teaching cells; retain the compact command exit-status check. Teach file inspection where it is the lesson's subject.
• Avoid chapter and section numbers in headings and navigation labels so lessons can be reordered.