If you're stepping into the world of machine learning and AI development, getting your system properly configured is one of the first hurdles you'll face. One essential component for GPU-accelerated deep learning is cuDNN (CUDA Deep Neural Network library), a GPU-accelerated library developed by NVIDIA for deep neural networks. In this beginners guide to install cuDNN on Ubuntu, we’ll break down the process into clear steps using insights from the Vultr documentation.

Why You Need cuDNN

cuDNN provides highly optimized implementations for standard operations like convolutions, pooling, and activation functions that are fundamental to deep learning. Installing cuDNN on Ubuntu ensures that frameworks like TensorFlow and PyTorch can fully leverage your GPU for faster training and inference.

Ubuntu 22.04 is a preferred choice for developers due to its stability and wide support for development tools, including NVIDIA drivers and CUDA libraries. However, installing cuDNN correctly can seem complex if you're new to system configuration—this guide simplifies it.

Prerequisites

Before installing cuDNN, ensure the following:

You're using Ubuntu 22.04.





You have a compatible NVIDIA GPU.





NVIDIA drivers and the CUDA Toolkit are already installed.





If CUDA isn’t installed yet, make sure to install it before proceeding with cuDNN, as cuDNN depends on it.

Step-by-Step: Install cuDNN on Ubuntu

1. Download cuDNN from NVIDIA

First, visit the NVIDIA cuDNN download page and sign in to access the downloads. Make sure to choose the version of cuDNN that matches your installed CUDA version.

For Ubuntu 22.04, download the following .deb files:

libcudnn8_*.deb





libcudnn8-dev_*.deb





libcudnn8-samples_*.deb





2. Install cuDNN Packages

Once you have downloaded the .deb files, open a terminal and navigate to the download directory. Run the following commands:

sudo dpkg -i libcudnn8_*.deb

sudo dpkg -i libcudnn8-dev_*.deb

sudo dpkg -i libcudnn8-samples_*.deb



These commands install the cuDNN runtime, development files, and example projects.

3. Verify Installation

You can verify that cuDNN was installed successfully by compiling and running the included samples:

cp -r /usr/src/cudnn_samples_v8/ ~/

cd ~/cudnn_samples_v8/mnistCUDNN

make

./mnistCUDNN



If the sample runs and outputs results without errors, cuDNN is correctly installed and working.



Final Thoughts

Learning how to install cuDNN on Ubuntu is a valuable step for any beginner interested in deep learning or AI development. With the help of Vultr’s easy-to-follow guide, you can complete the setup in just a few minutes. Proper cuDNN installation ensures your machine learning frameworks can take full advantage of your GPU, significantly speeding up model training.

Bookmark this guide and return to it whenever you need to configure a new development environment or troubleshoot GPU setup issues.