How do I specify nvidia runtime from docker-compose.yml?

Question

I am able to run a tensorflow container w/ access to the GPU from the command line w/ the following command

$ sudo docker run --runtime=nvidia --rm gcr.io/tensorflow/tensorflow:latest-gpu

I would like to be able to run this container from docker-compose. Is it possible to specify the --runtime flag from docker-compose.yml?

score 45 · Answer 1 · answered Aug 28 '18 at 18:04

45

Currently (Aug 2018), NVIDIA container runtime for Docker (nvidia-docker2) supports Docker Compose.

Yes, use Compose format 2.3 and add runtime: nvidia to your GPU service. Docker Compose must be version 1.19.0 or higher.

Example docker-compose.yml:

version: '2.3'

services:
  nvsmi:
    image: ubuntu:16.04
    runtime: nvidia
    environment:
      - NVIDIA_VISIBLE_DEVICES=all
    command: nvidia-smi

More example from NVIDIA blog uses Docker Compose to show how to launch multiple GPU containers with the NVIDIA Container Runtime.

answered Aug 28 '18 at 18:04

cedrickchee

832
8
12

5

This option is possible only for docker-compose v2 for now. For more info, please look at the corresponding pull request here https://github.com/docker/compose/pull/5405 – hgasimov Dec 30 '18 at 09:33
This option is more elegant as it allows for better packaging using docker-compose – Blaze Jan 22 '19 at 08:13
hm can different containers have different runtimes in one docker-compose file? – kodlan Feb 01 '21 at 20:59

score 32 · Accepted Answer · answered Nov 26 '17 at 11:58

32

You should edit /etc/docker/daemon.json, adding the first level key "default-runtime": "nvidia", restart docker daemon (ex. "sudo service docker restart") and then all containers on that host will run with nvidia runtime.

More info on daemon.json here

answered Nov 26 '17 at 11:58

Gabriel Miretti aka gmiretti

1,429
13
16

score 5 · Answer 3 · answered Jul 21 '18 at 15:00

Or better: using systemd and assuming the path is /usr/libexec/oci/hooks.d/nvidia

Configure

mkdir -p /etc/systemd/system/docker.service.d/
cat > /etc/systemd/system/docker.service.d/nvidia-containers.conf <<EOF
[Service]
ExecStart=
ExecStart=/usr/bin/dockerd -D --add-runtime nvidia=/usr/libexec/oci/hooks.d/nvidia --default-runtime=nvidia
EOF

Restart

systemctl daemon-reload
systemctl restart docker

Demo

Don't need to specify --runtime=nvidia since we set default-runtime=nvidia in the configuration step.

docker run --rm gcr.io/tensorflow/tensorflow:latest-gpu

Solution Inspired from my tutorial about KATA runtime.

How do I specify nvidia runtime from docker-compose.yml?

3 Answers3

Configure

Restart

Demo