metadata icon indicating copy to clipboard operation
metadata copied to clipboard

Allow to pass workspace/run name or id in initialization

Open Jeffwan opened this issue 6 years ago • 1 comments

/kind feature

I am trying to understand the best practice here.

  1. If I have training job runs once every day, should I create one Run for it and create a Execution every day?

If that's true, I think I only need to create one Workspace and one Run. Consider the SDK init, seems I have to get object and pass as parameter. Could we pass id or name as parameter and it would be easier to use.

  1. If I have a Kubeflow pipeline, is run here equivalent to pipeline concept there?

Since it has more steps and components, what's the best practice?

Have to pass workspace and run object to create instance.

test_run = metadata.Run(
    workspace=workspace,
    name="run-" + datetime.utcnow().isoformat("T") ,
    description="a run in workspace",
)

exec = metadata.Execution(
    name = "execution" + datetime.utcnow().isoformat("T") ,
    workspace=test_workspace,
    run=test_run,
    description="execution example",
)
print("An execution is create with id %s" % exec.id)

Jeffwan avatar Aug 07 '19 05:08 Jeffwan

/kind question /area engprod /priority p2

jtfogarty avatar Jan 14 '20 21:01 jtfogarty