Skip to content

Tao run on docker vs Tao run on kubernetes

Tao run on docker is an open-source agent skill for developer tools. Tao run on kubernetes is an open-source agent skill for developer tools.

  • Tao run on docker is best for launch a single-node tao training job in a docker container on a local gpu.
  • Tao run on kubernetes is best for submit a tao training run to our eks cluster on four gpus.
What it is
Tao run on docker is an open-source agent skill for developer tools.
Tao run on kubernetes is an open-source agent skill for developer tools.
Best for
Launch a single-node TAO training job in a Docker container on a local GPU
Submit a TAO training run to our EKS cluster on four GPUs
Kind
Skill
Skill
Pricing
Open source
Open source
Platforms
Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, Windsurf, Goose
Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, Windsurf, Goose
GitHub stars
4K
4K
Last commit
Nov 2026
Nov 2026
License
Apache-2.0
Apache-2.0
Open source
Good for
Launch a single-node TAO training job in a Docker container on a local GPU; Run a TAO job on a remote GPU server reached over SSH; Check the status and logs of a running Docker-based training container
Submit a TAO training run to our EKS cluster on four GPUs; Check whether a running Kubernetes training job has finished or failed; Pull the last 200 log lines from a TAO job on the cluster

What Tao run on docker does

  • Four-verb job lifecycle

    Submit, check status, read logs, and cancel a containerized job using standard docker commands.

  • Local and remote daemons

    Targets a local Docker daemon or a remote GPU host reached through DOCKER_HOST over SSH.

  • GPU selection guidance

    Explains selecting GPUs by device id on shared hosts so jobs are not all placed on GPU 0.

  • Safe writable mounts

    Runs containers as the submitting host user with group access and writable cache paths.

  • NGC and secret handling

    Covers nvcr.io login and passes credentials through environment variables rather than inline values.

  • Troubleshooting reference

    Lists common errors such as missing GPU runtime, shared memory exhaustion, and full disks with fixes.

What Tao run on kubernetes does

  • GPU capacity gate

    Checks allocatable nvidia.com/gpu across nodes and stops the submit before a job that cannot fit is applied.

  • Single and multi-node jobs

    Renders a single-pod Job for one node or an Indexed Job with a headless Service for distributed training.

  • Job status and logs

    Maps Kubernetes job and pod signals to job states and fetches recent container logs.

  • Flexible storage mounts

    Supports mounting existing PVC or NFS volumes, or staging data from and uploading results to S3.

  • Job cancellation and cleanup

    Deletes the Job with foreground cascade, removes the credential Secret and marks the job record canceled.

  • Local cluster testing

    Documents running minikube or kind to exercise the job lifecycle without cluster quota.

Tao run on docker vs Tao run on kubernetes: questions

Is Tao run on docker or Tao run on kubernetes cheaper?

Both are open source. Compare current plans on each official site.

What is the difference between Tao run on docker and Tao run on kubernetes?

Tao run on docker is an open-source agent skill for developer tools. Tao run on kubernetes is an open-source agent skill for developer tools.

Keep comparing