Custom visualization in Vertex AI (KubeflowV2DagRunner)

[jeongukjae]

If the feature is related to a specific library below, please raise an issue in
the respective repo directly:

TensorFlow Data Validation Repo

TensorFlow Model Analysis Repo

TensorFlow Transform Repo

TensorFlow Serving Repo

System information

• TFX Version (you are using): 1.9.x
• Environment in which you plan to use the feature (e.g., Local
  (Linux/MacOS/Windows), Interactive Notebook, Google Cloud, etc..): Vertex AI
• Are you willing to contribute it (Yes/No): If possible

Describe the feature and the current behavior/state.

As we can check in this link here, Vertex AI supports the markdown and html visualization, but TFX doesn't.

I think it can be supported via adding artifact types, but I'm not sure. (KFP HTML/Markdown Artifact types:
https://github.com/kubeflow/pipelines/blob/834d966033ebb8284981e86b831c2bb086ef30ba/sdk/python/kfp/components/types/artifact_types.py#L472-L503)

Will this change the current API? How?

Who will benefit with this feature?

The people who want to visualize custom component output in Vertex AI.

Do you have a workaround or are completely blocked by this?

Name of your Organization (Optional)

Any Other info.

[singhniraj08]

@jeongukjae,

We have a similar ongoing issue in which mlpipeline-metrics from custom Python function are not displayed in Kubeflow UI. Since team is already working on this issue, I will present this feature request for Vertex AI too and update this issue once we have an update. Thank you for the feature request!

@1025KB, Please take a look into this feature request for custom visualization in Vertex AI.

[jeongukjae]

Is there any ETA for this issue?

[jeongukjae]

Found the related commit: 30b2155