iterative / terraform-provider-iterative

☁️ Terraform plugin for machine learning workloads: spot instance recovery & auto-termination | AWS, GCP, Azure, Kubernetes
https://registry.terraform.io/providers/iterative/iterative/latest/docs
Apache License 2.0
290 stars 27 forks source link

nvidia-docker2 might not be needed #674

Closed DavidGOrtega closed 1 year ago

DavidGOrtega commented 2 years ago

With the release of Docker 19.03, usage of nvidia-docker2 packages is deprecated since NVIDIA GPUs are now natively supported as devices in the Docker runtime.

0x2b3bfa0 commented 2 years ago

Note that with the release of Docker 19.03, usage of nvidia-docker2 packages are deprecated since NVIDIA GPUs are now natively supported as devices in the Docker runtime.

Docker 19.03 introduced the --gpus option (https://github.com/docker/cli/pull/1714) to replace the old --runtime=nvidia option.

0x2b3bfa0 commented 2 years ago

Replace nvidia-docker2 with nvidia-container-toolkit here:

https://github.com/iterative/terraform-provider-iterative/blob/1e888fbb77b1be0d640ac95dd0df40168b066b20/environment/setup.sh#L40

aliabbasjaffri commented 1 year ago

Hello @DavidGOrtega! I would like to take over this issue. Can you please assign it to me?

0x2b3bfa0 commented 1 year ago

Yes, sure!

aliabbasjaffri commented 1 year ago

Are there any other changes required apart from the one mentioend in the comment above? I tried to look for instances where the old --runtime was being used; either in code or readme, but none found.

0x2b3bfa0 commented 1 year ago

I think that no other changes are required

0x2b3bfa0 commented 1 year ago

After seeing that https://github.com/NVIDIA/nvidia-docker/commit/88a2fda83cb3cf42221931dbd0c08aa4370d4fe5 points to a documentation page explicitly requiring the installation of the allegedly deprecated nvidia-docker2 package, the best thing we can do is ~question their sanity~ close this issue.