Before we begin lets cover the prerequisites and what to expect during the installation process.
Ubuntu 18.04 or 20.04 OS, a terminal window and latest Chrome web browser.
A terminal window.
The latest Chrome web browser.
Python version 3.6.9 or later pip.
Mobile or desktop workstation with NVIDIA GPU.
16 GB of GPU memory recommended.
8 GB of GPU memory minimum.
You will use command-line tools such as Git, pip, nvidia-smi, NGC, kaggle, aws-cli in your favorite shell, usually bash. During the initial Workbench install, you will be prompted to enter sudo password for the pip3 installation. Workbench automatically clones the NVIDIA Data Science Stack to facilitate NVIDIA Linux driver installation.
The current NVIDIA driver installed on the system will be upgraded to support the most up-to-date NGC Deep Learning framework (TensorFlow, PyTorch) containers. NVIDIA updates these containers once a month.
When Workbench updates, older versions of the containers will still be available in docker images reporting. After Workbench is installed and running, look for an NVIDIA icon in the upper right-hand side of the Ubuntu toolbar near the networking icon. To start Workbench, click on the NVIDIA icon, and a pulldown menu of data science tasks will appear.