Skip to main content

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.

This guide walks through the complete process of running the Hands-On ML notebooks on your own machine. The setup uses Anaconda (or Miniconda) to create an isolated homl3 conda environment with Python 3.10 and every library the notebooks require. If you just want to try a few notebooks without installing anything, see the cloud options instead.

​
Installation walkthrough

1

Install git

Open a terminal and type git to check whether it is already installed. If not, download it from git-scm.com.On Windows, after installation you should use the Anaconda Shell (not the standard Command Prompt) for all subsequent steps.
2

Install Anaconda or Miniconda

Download and install Anaconda or the lighter Miniconda. Either works — Anaconda bundles many scientific packages, while Miniconda is minimal and only installs what is needed.
  • macOS / Linux: Accept the conda init prompt during installation so that conda is available in every new terminal. Restart your terminal after installation.
  • Windows: Do not add Anaconda to the system PATH when the installer asks. Instead, open Anaconda Shell from the Start Menu whenever you want to use conda.
After installation, update the conda packaging tool:
conda update -n base -c defaults conda
3

Clone the repository

Navigate to the directory where you want to store the project, then clone the repository:
cd $HOME
git clone https://github.com/ageron/handson-ml3.git
cd handson-ml3
If you prefer not to use git, download main.zip, unzip it, rename the resulting directory to handson-ml3, and move it to your development directory.
4

Create the homl3 conda environment

From inside the handson-ml3 directory, create the environment using the included environment.yml file. This installs Python 3.10 along with Scikit-Learn 1.3, TensorFlow 2.14, Keras, NumPy 1.26, Pandas 2.1, Matplotlib 3.8, and all other dependencies:
conda env create -f environment.yml
This step takes several minutes. When it completes, activate the new environment:
conda activate homl3
The environment name is homl3 by default. If you want a different name, pass -n your_name to the conda env create command. You will then need to adjust the conda activate command accordingly.
5

Register the Jupyter kernel

Register the homl3 environment as a Jupyter kernel named python3. The notebooks default to a kernel named python3, so using this name means you will not need to change the kernel manually every time you open a notebook:
python3 -m ipykernel install --user --name=python3
6

Launch Jupyter

Start Jupyter Notebook:
jupyter notebook
Jupyter will open in your browser and show the contents of the handson-ml3 directory. Click index.ipynb to see the table of contents and navigate to any chapter. If your browser does not open automatically, visit localhost:8888.To stop Jupyter, press Ctrl-C in the terminal where you started it.

​
Starting Jupyter in future sessions

Each time you want to work on the notebooks, open a terminal and run: 
cd $HOME/handson-ml3
conda activate homl3
jupyter notebook

​
GPU setup (NVIDIA)

If you have a TensorFlow-compatible NVIDIA GPU (Compute Capability 3.5 or higher), you can enable GPU acceleration for the deep learning chapters.
1

Install the NVIDIA GPU driver

Download and install the latest driver for your GPU from nvidia.com.
2

Install CUDA and cuDNN

When you install TensorFlow from the Anaconda channel (which the environment.yml does), CUDA and cuDNN are installed automatically alongside TensorFlow. You do not need to install them manually.If you are not using Anaconda, follow the manual GPU installation instructions at tensorflow.org/install/gpu.
3

Verify GPU access

After completing the main installation and launching Jupyter, open any notebook and run:
import tensorflow as tf
tf.config.list_physical_devices("GPU")
If TensorFlow can see your GPU, a list containing your device(s) will be printed.
GPU support in Docker requires additional setup. See the Docker guide for details.

​
Updating the project

The notebooks are updated regularly to fix issues and add library support. To update your local copy: 
cd $HOME/handson-ml3
git pull
If git pull fails because you modified a notebook, commit your changes to a branch first: 
git checkout -b my_branch
git add -u
git commit -m "describe your changes here"
git checkout main
git pull
After pulling, update the libraries by recreating the conda environment: 
conda update -c defaults -n base conda
conda activate base
conda env remove -n homl3
conda env create -f environment.yml
conda activate homl3
jupyter notebook

​
Conda vs pip

The recommended approach is to use conda with environment.yml, which handles all library versions and binary dependencies (including GPU support) in a tested, consistent way. If you prefer not to use Anaconda, you can install the libraries with pip using requirements.txt: 
pip install -r requirements.txt
TensorFlow is much easier to install via Anaconda, especially on Windows or when using a GPU. The pip path is not recommended unless you are comfortable managing binary dependencies manually.

​
Library versions installed

The environment.yml specifies exact versions for reproducibility. Key packages include:
PackageVersion
Python3.10
Scikit-Learn1.3
TensorFlow2.14
NumPy1.26
Pandas2.1
Matplotlib3.8
JupyterLab4.0
Keras Tuner1.4.6
Transformers4.35

Build docs developers (and LLMs) love

Get started for freeTalk to us