Launching Jupyter Notebooks

Overview

Questions

• How do you launch a Jupyter notebook from OnDemand?
• How do you configure resources for your notebook session?
• How do you use notebook cells, run code, and write documentation?
• How do you install additional Python packages?

Objectives

• Launch a Jupyter notebook session from OnDemand
• Configure CPU, memory, and time for your session
• Create and execute code and markdown cells
• Upload data and install packages in the notebook environment

Launching Jupyter

---

1. Click Interactive Apps in the left sidebar
2. Select Jupyter Notebook (or Jupyter Lab)
3. Configure your resources:
   • CPU cores: 1–4 for most work
   • Memory: 4–16 GB depending on dataset size
   • Time: 1–8 hours
   • GPU (optional): Select if you need accelerated computing
4. Click Launch
5. Wait for allocation (typically 1–5 minutes)
6. Click Connect to Jupyter when the session is ready

The launch form is where you choose a partition and ask for resources – the same decisions as an sbatch script, as a web form.

Callout

First Launch Takes Time

When you request Jupyter, expect to wait 1–5 minutes for resource allocation, especially during busy hours. The system is requesting compute nodes for your session. Subsequent launches typically start faster.

Using Jupyter Notebooks

---

Creating a Notebook

Click New then Python 3 to create a new notebook. Each notebook is a .ipynb file saved in your working directory.

Cell Types and Execution

• Code cells: Write Python code and press Shift+Enter to execute
• Markdown cells: Write formatted text, headings, and documentation
• Output: Results, tables, and plots appear directly below each cell

A common first cell:

PYTHON

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
print("Libraries loaded successfully!")

Uploading Data

1. Use the Jupyter file browser on the left to navigate to your data
2. Or click the Upload button within Jupyter
3. Read files in your notebook:

PYTHON

data = pd.read_csv('/bigdata/lab/<labname>/data.csv')
print(data.head())

Use absolute paths (like /bigdata/lab/<labname>/) to avoid confusion about your working directory.

Installing Packages

Install additional packages directly from a notebook cell:

PYTHON

!pip install scikit-learn
!pip install seaborn

Or open a terminal in Jupyter (New > Terminal) for more complex installations.

Session Management

---

• Sessions expire after idle time or when the requested time limit is reached
• Save your work frequently with Ctrl+S
• Download important notebooks before your session expires
• Check remaining time in the OnDemand Active Jobs view

Use batch jobs (sbatch) for computations longer than 8 hours instead of interactive Jupyter sessions.

Challenge

Challenge: Create Your First Notebook

1. Launch a Jupyter notebook from OnDemand with 1 core, 4 GB memory, and 1 hour

2. Create a new Python 3 notebook

3. In cell 1, import numpy, pandas, and matplotlib

4. In cell 2, create a sample dataset:

   PYTHON

   data = np.random.normal(loc=50, scale=10, size=500)
   df = pd.DataFrame({'values': data})
   print(df.describe())

5. In cell 3, create a histogram:

   PYTHON

   plt.hist(df['values'], bins=25, edgecolor='black')
   plt.xlabel('Value')
   plt.ylabel('Frequency')
   plt.title('Sample Distribution')
   plt.show()

6. Save your notebook

Show me the solution

1. Jupyter launches after 1–5 minutes of resource allocation
2. A new notebook opens with an empty code cell
3. All three libraries import without errors
4. df.describe() prints count, mean, std, min, max, and quartiles
5. A histogram appears inline showing a normal distribution centered around 50
6. The notebook saves as an .ipynb file in your working directory

If your session takes a long time to start, the cluster may be busy – try reducing your resource request or launching during off-peak hours.

Key Points

• Launch Jupyter from Interactive Apps with configurable CPU, memory, and time
• Expect 1–5 minutes for initial resource allocation
• Use code cells for Python and markdown cells for documentation
• Install packages with !pip install directly in notebook cells
• Save work frequently and download notebooks before sessions expire
• Use batch jobs for computations longer than 8 hours