Skillversitas

Working with Jupyter Online

You do not need your own machine to run notebooks: hosted Jupyter services give you kernels in the browser with preinstalled data science stacks. They are ideal for quick analysis, teaching, and trying languages you have not installed locally.

Google Colab

# colab.research.google.com
# new notebook -> Python 3
# code cells run on Google's VM; runtime -> change runtime type selects CPU/GPU/TPU
  • Free tier: CPU VM with a few hours of continuous runtime; GPUs available with limits.
  • Drive integration: mount your Google Drive and read/write files from cells.
  • Share: link with "view" or "edit" permissions; notebooks live in your Drive.

Classic Hosted Options

  • JupyterHub-based services (various university and cloud offerings): one-click kernels, often with preinstalled environments.
  • Cloud notebooks (AWS SageMaker, Azure Notebooks-style services, GCP AI Notebooks): kernels tied to your cloud account with storage and GPU options.
  • Kaggle Kernels: notebooks with attached datasets and free GPU time, good for competition-style work.

Using Your Own Notebooks Online

  1. Upload a .ipynb (or link a public GitHub repo) to the hosted service.
  2. Install missing packages in a setup cell: !pip install mypackage.
  3. Data: upload files to the service's file area, or fetch from a URL in the first cell.
  4. Export the result back to GitHub or download the .ipynb to keep the record.

Caveats to Plan Around

  • Sessions are ephemeral: anything not saved to attached storage (Drive, S3) or committed to a repo is lost when the kernel dies.
  • Runtime limits (idle timeouts, hour caps) mean long jobs need checkpoints: save intermediate artifacts to storage every N steps.
  • Secrets: never paste API keys into cells that will be shared; use the service's secrets UI where available.

Summary

Work in hosted Jupyter environments - Google Colab, JupyterHub services, or cloud notebooks - to get kernels, preinstalled stacks, and GPUs without local setup. Upload your .ipynb, install extra packages in a setup cell, attach storage for data, and save results back to a repo or cloud storage before the session ends.