Docker: Error response from daemon: Unknown runtime specified nvidia

353 Views Asked by At

This is a common problem but I still cannot resolve the issue. I'm trying to run a docker app called FakeFinder and I'm using the command sudo docker-compose up -d --build as directed by the setup instructions to run it, but run into the above error when I do. I'm using the Deep learning AMI (ubuntu 18.04) Version 52.0 which comes with nvidia-docker and all the drivers set up. Running sudo docker info | grep -i runtime doesn't show nvidia under the runtimes regardless of what I try.

My /etc/docker/daemon.json is reasonable and looks as follows:

{

    "runtimes": {
        "nvidia": {
            "path": "nvidia-container-runtime",
            "runtimeArgs": []
        }
    }
}

Changing path to "usr/bin/nvidia-container-runtime" does not help.

I've tried to restart dockerd with

sudo pkill -SIGHUP dockerd

I've tried restarting daemon with

sudo systemctl daemon-reload
sudo systemctl restart docker

I've tried all three options seen here to add the runtime: https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/user-guide.html

0

There are 0 best solutions below