Tool Versions¶
The versions of ML & Data tools included with each version of deployKF.
Kubeflow Ecosystem¶
This section lists the versions of tools in the Kubeflow Ecosystem which are included in each version of deployKF.
Kubeflow Pipelines¶
The Kubeflow Pipelines component versions:
| deployKF | 0.1.0 - 0.1.4 | 0.1.5+ |
|---|---|---|
| Kubeflow Pipelines | 2.0.0-alpha.7 | 2.1.0-deploykf.0 |
| Argo Workflows | 3.3, 3.4 | 3.3, 3.4 |
Warning, 2.0.0-alpha.7 is NOT Kubeflow Pipelines v2
Confusingly, the 2.0.0-alpha.7 version is actually part of the KFP v1 line, so you must use 1.X.X SDK versions with it, for example: kfp==1.8.22.
What does the -deploykf.X version suffix mean?
Due to upstream issues with Kubeflow Pipelines, we maintain a fork in the deployKF/kubeflow-pipelines repository. From an end-user perspective, the fork is functionally identical to upstream, and works with the official kfp Python SDK.
Can I bring my own version of Argo Workflows?
By default, deployKF will install the correct version of Argo Workflows for the version of Kubeflow Pipelines you are using. Right now, its a little complex to bring your own Argo Workflows version.
See deployKF/deployKF#116 to join the discussion.
Kubeflow Notebooks¶
The Kubeflow Notebooks component versions:
| deployKF | 0.1.0+ |
|---|---|
| Notebooks Web App | 1.7.0 |
| Notebooks Controller | 1.7.0 |
| PodDefaults Webhook | 1.7.0 |
Kubeflow Katib¶
The Kubeflow Katib component versions:
| deployKF | 0.1.0+ |
|---|---|
| Kubeflow Katib | 0.15.0 |
Kubeflow Training Operator¶
The Kubeflow Training Operator component versions:
| deployKF | 0.1.0+ |
|---|---|
| Kubeflow Training Operator | 1.6.0 |
Kubeflow Volumes¶
The Kubeflow Volumes component versions:
| deployKF | 0.1.0+ |
|---|---|
| Volumes Web App | 1.7.0 |
Kubeflow TensorBoards¶
The Kubeflow TensorBoards component versions:
| deployKF | 0.1.0+ |
|---|---|
| TensorBoards Web App | 1.7.0 |
| TensorBoards Controller | 1.7.0 |
Created: 2024-05-10