Overview

Jupyter Notebooks allow you to run interactive Python sessions with embedded code, visualizations, and documentation. On SeaWulf, you can launch Jupyter through the Open OnDemand (OOD) web interface, making it easy to work from a browser without needing to SSH into the cluster.

Getting Started

1. Navigate to the SeaWulf Open OnDemand portal: https://sn-ood.seawulf.stonybrook.edu/
2. Log in using your NetID and DUO authentication.
3. From the dashboard, select Interactive Apps and choose Jupyter Notebook or JupyterLab.
4. Select the compute queue and number of cores you need. For heavy workloads, choose a larger compute node.
5. Click Launch. Once the session starts, your notebook interface will open in the browser.

Best Practices

• Use the appropriate node type based on your computation needs (e.g., 28-core Haswell, 40-core Skylake, 96-core Milan, or 96-core Xeon Max HBM).
• Save notebooks frequently to your home or project directory to avoid losing work.
• Install Python packages in a Conda environment within your notebook session for reproducibility.
• Close the session when finished to free up cluster resources.
• Monitor resource usage if running long computations to ensure you do not exceed allocated resources.

Accessing Files

You can access your home, scratch, and project directories directly from Jupyter. Use relative paths to your workspace or mount directories as needed. This ensures your notebooks can read and write data efficiently.