All Hands-On ML notebooks can be run directly in the browser on several cloud platforms without installing any software. This page covers each option in detail so you can choose the one that suits your workflow. Google Colab is the recommended starting point for most readers.Documentation Index
Fetch the complete documentation index at: https://mintlify.com/ageron/handson-ml3/llms.txt
Use this file to discover all available pages before exploring further.
- Google Colab (recommended)
- Kaggle
- Binder
- Deepnote
Google Colab is the platform the author recommends. It provides a hosted Jupyter environment on Google’s infrastructure, with optional free access to GPU and TPU hardware accelerators. All you need is a Google account.Open the notebooks:
https://colab.research.google.com/github/ageron/handson-ml3/blob/main/Clicking that link opens Colab’s notebook browser for the repository. Select any chapter notebook to open it. Colab installs the required dependencies in the background when you run the first cell.When to choose Colab:
- You want to get started immediately with no setup
- You need free GPU access for the deep learning chapters (chapters 10–19)
- You are already working in the Google ecosystem (Drive, GCP)
- You want the most tested and supported cloud option for this repository
Platform comparison
| Feature | Colab | Kaggle | Binder | Deepnote |
|---|---|---|---|---|
| Free tier | Yes | Yes | Yes | Yes (limited) |
| GPU access | Yes (free) | Yes (30 h/week) | No | Yes (paid) |
| Session persistence | No (temporary) | Yes | No (temporary) | Yes |
| Account required | Kaggle | No | Deepnote | |
| Tested with this repo | Yes | Partial | Partial | Partial |
| Startup time | Fast | Fast | Slow | Fast |
Tips for cloud usage
The first cell of each notebook typically installs or checks for required packages. Run it even if Colab or Kaggle appear to have the libraries installed — the cell handles version pinning.
- Save your work regularly. On temporary platforms (Colab, Binder), download your notebook or sync to Google Drive after making changes.
- GPU runtime in Colab. To enable a GPU, go to Runtime > Change runtime type and select T4 GPU. Free GPU access is subject to availability and usage limits.
- Data downloads. Notebooks that download datasets (e.g., chapters 2, 15) will re-download each session on temporary platforms. This is expected behavior.
Viewing notebooks without running code
If you only want to read the notebooks without executing any cells, you can view a static render:- nbviewer — renders notebooks with correct math equation formatting
- GitHub’s built-in viewer — works for smaller notebooks, but may fail to load large ones and does not always render math correctly
Ready to install locally?
Local installation
Full walkthrough for Anaconda, the homl3 environment, and GPU setup
Docker setup
Run notebooks in an isolated container with docker-compose