Open MichaelClifford opened 3 years ago
We should be clear about the MLOps workflow we want to show, the current workflow is pretty simplistic. Seldon seems to be preferred by ODH/RHODS. How do we automate deployment and metrics feedback from a Seldon-based deployment?
/kind feature /triage needs-information
We now have the seldon operator deployed in the moc zero cluster and can start to update the demo to use seldon.
There are still a couple of problems with it, but it is usable and we have confirmed that we can deploy models and query inference endpoints.
We have opened issues on opf/support and seldon upstream to address them.
related #176
/assign @oindrillac
Issues go stale after 90d of inactivity.
Mark the issue as fresh with /remove-lifecycle stale
.
Stale issues rot after an additional 30d of inactivity and eventually close.
If this issue is safe to close now please do so with /close
.
/lifecycle stale
/lifecycle frozen
The current tutorial relies on a custom flask app and some
oc
commands. We should update the tutorial to use Seldon to remove the user's need for interacting with the cluster directly.