Currently, if a docker task has the appropriate labels, we try to run the docker task with the runtime='nvidia' parameter. As of Docker 19.03, there is docker built-in support for the --gpus flag (more generally, for a device request). As soon as docker-py supports this (see issue docker/docker-py#2395 and PRs docker/docker-py#2419, docker/docker-py#2471), we should switch to using the appropriate method. Until then, there are extra steps to use nvidia dockers.
Currently, if a docker task has the appropriate labels, we try to run the docker task with the
runtime='nvidia'parameter. As of Docker 19.03, there is docker built-in support for the--gpusflag (more generally, for a device request). As soon as docker-py supports this (see issue docker/docker-py#2395 and PRs docker/docker-py#2419, docker/docker-py#2471), we should switch to using the appropriate method. Until then, there are extra steps to use nvidia dockers.