Submarine for K8s supports distributed TensorFlow and PyTorch.
Submarine can run on K8s >= 1.14, supports features like GPU isolation.
We have validated Submarine on the following versions:
| K8s Version | Support? |
|---|---|
| 1.13.x (or earlier) | X |
| 1.14.x | √ |
| 1.15.x | √ |
| 1.16.x | √ |
| 1.17.x | To be verified |
| 1.17.x | To be verified |
Submarine can be deployed on any K8s environment if version matches. If you don't have a running K8s, you can set up a K8s using Docker Desktop, MiniKube, or kind, Kubernetes-in-Docker.
From our experiences, Docker Desktop is an easier choice.
After you have an up-and-running K8s, you can follow Submarine Helm Charts Guide to deploy Submarine services on K8s cluster in minutes.
Please refer to Notebook guide if you want to use jupyter notebook in Submarine.
You can now submit Submarine model training on the UI, please refer to Submit Experiment from Submarine UI
You can also follow the guidance to run Submarine model training using Python SDK, please refer to Submit Experiment using Python SDK
Alternatively, we support use REST API to submit, list, delete experiments (model training)