Install Tensorflow on Ubuntu


TensorFlow is a versatile platform for developing and deploying Machine Learning models. This guide will walk you through installing and setting up TensorFlow on your Ubuntu system and configure it for NVIDIA acceleration. For this you will need:

  1. Computer with NVIDIA graphics
  2. Ubuntu 22.04 installed
  3. NVIDIA graphics drivers installed (these are already pre-installed on Entroware NVIDIA systems and do not need to be reinstalled)

Installing the CUDA toolkit

Running the commands below will install the CUDA toolkit. The most up-to-date commands can be found at

Fetching and installing the CUDA toolkit

Fetch and add the apt pin file

1user@machine:~$ wget
1user@machine:~$ sudo mv /etc/apt/preferences.d/cuda-repository-pin-600

Download and install the CUDA repository package and add the keys

1user@machine:~$ wget
1user@machine:~$ sudo dpkg -i cuda-repo-ubuntu2204-12-4-local_12.4.1-550.54.15-1_amd64.deb
1user@machine:~$ sudo cp /var/cuda-repo-ubuntu2204-12-4-local/cuda-*-keyring.gpg /usr/share/keyrings/

Update the apt package list

1user@machine:~$ sudo apt-get update

Install the CUDA Toolkit package

1user@machine:~$ sudo apt-get -y install cuda-toolkit-12-4

Export variables for the CUDA toolkit

1user@machine:~$ CUDA_VER=$(ls /usr/local/ | grep cuda | tail -n1)
1user@machine:~$ echo "export PATH=\$PATH:/usr/local/${CUDA_VER}/bin" >> ~/.bashrc
1user@machine:~$ echo "export LD_LIBRARY_PATH=\$LD_LIBRARY_PATH:/usr/local/${CUDA_VER}/lib64" >> ~/.bashrc
1user@machine:~$ source ~/.bashrc

Installing and testing TensorFlow

Installing Python3 and the TensorFlow framework

1user@machine:~$ sudo apt-get install python3-venv
1user@machine:~$ python3 -m venv venv
1user@machine:~$ source venv/bin/activate
1user@machine:~$ pip install tensorflow[and-cuda]
1user@machine:~$ export CUDNN_PATH=$(dirname $(python -c "import nvidia.cudnn; print(nvidia.cudnn.__file__)"))
1user@machine:~$ export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:${CUDNN_PATH}/lib

Testing GPU detection

You can now use TensorFlow to test whether your NVIDIA GPU is being detected properly

1user@machine:~$ python3 -c "import tensorflow as tf;

If the GPU is being detected correctly, the output should look similar to the following:

1[PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')]

If the GPU is not being detected correctly, the output will be similar to:

1Skipping registering GPU devices...


You should now have a functioning install of the TensorFlow framework, complete with CUDA acceleration to turbocharge your Machine Learning models. More information, guides, and tools can be found on the TensorFlow website: