Development Setup#

This guide covers setting up your development environment for ACE workflows, including installing the NVIDIA driver, Docker, the NVIDIA Container Toolkit, and logging into NVIDIA’s container registry using your NGC API key.

Prerequisites#

This guide assumes you are using a Linux system with an NVIDIA GPU. The instructions have been tested on Ubuntu 22.04, but may also work on other Linux versions and distributions.

Install NVIDIA driver#

Run the following command to check if the NVIDIA driver is installed:

nvidia-smi

If the command is not found, or if the Driver Version is lower than 535, install the NVIDIA driver:

sudo apt update
sudo apt install nvidia-driver-535

Reboot your machine, and run nvidia-smi again to verify that the driver is correctly setup.

Install NGC and set API key#

This documentation uses Docker images from NVIDIA’s container registry. If not done already, create an NGC account and retrieve your API key.

Run the following command to check if you have already installed the ngc CLI:

ngc config set

If the command is not found, install the NGC CLI using your API key, and run ngc config set again.

Install UCS Tools#

You will need UCS Tools to build UCS applications. Run the following command to check whether they are already installed:

ucf_app_builder_cli -h

If the command does not exist, follow the steps in Install UCS Tools.

Then, sync the all the repos:

ucf_app_builder_cli registry repo sync

Install Docker#

Run the following command to check if Docker is already installed:

docker -v

If the command is not found, install docker.

Login to the nvcr.io container registry#

Ensure that your docker installation is logged in to the nvcr.io container registry:

docker login nvcr.io -u \$oauthtoken

When prompted for a password, enter your NGC API key. If the key is recognized, the command outputs Login Succeeded.

Install NVIDIA container toolkit#

Run the following command to check if NVIDIA Container Toolkit is installed:

sudo dpkg -l | grep nvidia-container-toolkit

If the command returns nothing:

To verify that NVIDIA Container Toolkit has been correctly configured and installed, run the following:

docker run --rm --gpus all nvidia/cuda:11.0.3-base-ubuntu20.04 nvidia-smi

Note

Note: you may have to run the command with sudo, if you have not configured Docker to run as non-root user.

This command should output information about the NVIDIA driver version, GPU and more.

On this page