Technical Documentation for Clean Installation of Ubuntu, CUDA, cuDNN, and PyTorch on WSL2

2025/3/23

This guide provides step-by-step instructions for setting up a clean installation of Ubuntu on Windows Subsystem for Linux 2 (WSL2), followed by installing CUDA, cuDNN, and PyTorch for machine learning development.

Prerequisites

Windows 10 version 2004 or higher (Build 19041 or higher) or Windows 11
A system with an NVIDIA GPU compatible with CUDA 12.4
Administrative privileges on your Windows machine
At least 20GB of free disk space

Step 1: Remove Old or Unused Linux Distributions

Open Windows Terminal, PowerShell, or Command Prompt as Administrator.
List all installed WSL distributions:
```
wsl --list -v
```
Unregister any unwanted distributions:
```
wsl --unregister DistroName
```
Replace DistroName with the name of the distribution you want to remove (e.g., Ubuntu-20.04).

Step 2: Install Ubuntu(Specifically Ubuntu 24.04)

Ensure WSL2 is enabled and set as the default WSL version:
```
wsl --set-default-version 2
```
Install Ubuntu using WSL:
```
wsl --install -d Ubuntu-24.04
```
Once the installation is complete, you will be prompted to reboot your system. Once the system has been rebooted, open the ubuntu terminal and set your username and password.

Step 3: Update and Upgrade Ubuntu

Open the Ubuntu terminal.
Run the following commands to update and upgrade the system:
```
sudo apt update && sudo apt upgrade -y
```

Install essential development tools:

sudo apt install -y build-essential git wget curl

Step 4: Install CUDA (Version 12.4)

Visit the CUDA 12.4.1 download page:
CUDA 12.4.1 Download Archive

Follow the installation instructions provided on the installer page. Just before <sudo apt-get -y install cuda-toolkit-12-4>, make sure to follow this step:

# Open a new file for storing the sources list
sudo nano /etc/apt/sources.list.d/ubuntu.sources

# Paste the following at the end of the file:
Types: deb
URIs: http://old-releases.ubuntu.com/ubuntu/
Suites: lunar
Components: universe
Signed-By: /usr/share/keyrings/ubuntu-archive-keyring.gpg

Save the file using Ctrl+O, press Enter, and exit using Ctrl+X.

Update package lists and install CUDA:

sudo apt update
sudo apt-get -y install cuda-toolkit-12-4

Configure environment variables by editing the .bashrc file:
```
nano ~/.bashrc
```

Add the following lines at the end of the file:

# Add local bin to PATH
if [ -d $HOME/.local/bin ]; then
    export PATH=$HOME/.local/bin:$PATH
fi

# CUDA configuration
export CUDA_HOME=/usr/local/cuda-12.4

if [ -d $CUDA_HOME/bin ]; then
    export PATH=$CUDA_HOME/bin:$PATH
fi

if [ -d $CUDA_HOME/lib64 ]; then
    export LD_LIBRARY_PATH=$CUDA_HOME/lib64:$LD_LIBRARY_PATH
fi

Save the file using Ctrl+O, press Enter, and exit using Ctrl+X.
Apply the changes:
```
source ~/.bashrc
```
Verify the CUDA installation:
```
nvcc --version
nvidia-smi
```

Step 5: Install cuDNN

Visit the cuDNN download page: cuDNN Downloads
Follow the installation instructions provided on the page.
Ensure compatibility with the installed CUDA version (12.4).

Step 6: Install PyTorch

Install Python and pip if not already installed:

sudo apt install -y python3 python3-pip python3-venv

Create a virtual environment for PyTorch:
```
python3 -m venv pytorch_env
```
Activate the virtual environment:
```
source pytorch_env/bin/activate
```

Install PyTorch with CUDA support:

pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu124

Verify the PyTorch installation and CUDA compatibility:

python3 -c "import torch; print('PyTorch version:', torch.__version__); print('CUDA available:', torch.cuda.is_available()); print('CUDA version:', torch.version.cuda if torch.cuda.is_available() else 'Not available')"

Deactivate the virtual environment when done:
```
deactivate
```

Common Troubleshooting

CUDA Not Found

If CUDA is not found after installation, verify that:

The environment variables are correctly set in your .bashrc file
You’ve sourced the .bashrc file after modifications
The CUDA installation path matches the path in your environment variables

PyTorch CUDA Issues

If PyTorch doesn’t detect CUDA, try:

Reinstalling PyTorch with the specific CUDA version you installed
Checking compatibility between your PyTorch and CUDA versions
Verifying that both CUDA and cuDNN are correctly installed

WSL2 GPU Access Issues

If WSL2 cannot access your GPU, ensure that:

You have the latest NVIDIA drivers installed on Windows
WSL2 GPU support is enabled in your system
You’ve restarted your computer after driver installations

Conclusion

You now have a clean installation of Ubuntu on WSL2 with CUDA 12.4, cuDNN, and PyTorch properly configured. This setup provides a robust environment for machine learning and deep learning development.

For more information, consult the official documentation:

LOADING