End-to-end examples

Downloading data, training a model, and uploading predictions

This example demonstrates how to use the ChallengeClient to download data, train a model, and upload predictions.

Submission automation

Scheduling a Kaggle notebook

If you're just starting out, we recommend using Kaggle Notebooks to schedule your submissions.

1. Settings (⚙) → Schedule a notebook run → On
2. Choose Frequency (daily / weekly / monthly), Start date, Runs ≤ 10 → Save
3. A clock icon appears; each run writes a new Version with full logs & outputs
4. Limits: CPU-only • ≤ 9 h per run • 1 private / 5 public schedules active
5. Pause or delete the job anytime from the same Settings card

_{Need GPUs? Trigger notebook commits with the Kaggle API from cron/GitHub Actions.}

Scheduling a Google Colab (Vertex AI) notebook

1. Create a Google Cloud account if you don't have one already
2. Go to Google Colab Notebooks in Vertex AI
3. Set up a schedule:
4. Open your notebook in Colab
5. Click Runtime → Manage sessions
6. Select Recurring and configure your schedule
7. Set frequency (daily/weekly/monthly) and duration
8. Click Save
9. Authentication options:
10. Use service account keys stored securely
11. Set up environment variables in the Vertex AI console
12. Use Google Cloud's Secret Manager for API keys

_{Note: Scheduled Colab notebooks run on Google Cloud and may incur charges based on your usage.}