Delete or update outdated deployer component for TFServing

[jlewi]

There is a pipeline component to deploy models for TFServing here https://github.com/kubeflow/pipelines/tree/ba64bd3866023c36c93cf2b46a066a1b008bb6c4/components/kubeflow/deployer

But it looks very outdated

• It is using ksonnet and the old TFServing ksonnet component which has long since been deleted

I'm classifying this as a P1 bug as I suspect this component doesn't work since its so outdated and the ksonnet package has likely been deleted.

/cc @jinchihe @animeshsingh

[jinchihe]

Agree, we can use KFServing to do last mile for model serving, that's common and easy using.

[gaoning777]

This component is still working since the sample test covers this kubeflow-serving. Now we have seldon samples for the serving. We could potentially remove this one after adding the seldon serving in the core samples. @jinchihe Would you mind adding the sample test for the seldon samples? /assign @jinchihe /unassign @gaoning777

[jinchihe]

@gaoning777 Why not using KFServing to deploy serving in the sample? Thanks.

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[Bobgy]

/lifecycle frozen

/cc @numerology @chensun I just found this issue and looks like we should delete the tf serving sample.

[numerology]

Sounds good. Should we use new KF serving instead?

[Bobgy]

@jinchihe @animeshsingh will you be interested in contributing a kfserving sample?

From my understanding, that requires a kfserving install to properly run, so we need to run it in full Kubeflow, is that right?