Running GPU Executors
This document is applicable to the following:
This document outlines how to run GPU (graphics processing unit) machine executors using CircleCI Server.
Prerequisites
Configure the vm-service in the Replicated management console to start a GPU-enabled instance.
Overview
Run the following commands on any Nvidia GPU-enabled instance. The following example uses CUDA 8.0, but you can use any CUDA runtime version supported by your GPU instance.
wget https://developer.nvidia.com/compute/cuda/8.0/prod/local_installers/cuda-repo-ubuntu1404-8-0-local_8.0.44-1_amd64-deb
sudo apt-get update
export OS_RELEASE=$(uname -r)
sudo apt-get install -y linux-image-extra-$OS_RELEASE linux-headers-$OS_RELEASE linux-image-$OS_RELEASE
sudo dpkg -i cuda-repo-ubuntu1404-8-0-local_8.0.44-1_amd64-deb
sudo apt-get update
sudo apt-get --yes --force-yes install cuda
nvidia-sminvidia-smi is only required for testing purposes. After you install the CUDA driver in Step 7 you should be good to go!
Adding GPU Steps to an AMI
To avoid start up time associated with the above steps, they may be included in an AMI by following the instructions in the Configuring VM Service documentation.