[Release 1.1] training operator release

[Jeffwan]

Opening this issue to track releasing training operators for Kubeflow 1.1 and would like to add area OWNERS for @johnugeorge @terrytangyuan @Jeffwan

Seems we want to these works done before 1.1 release. We may not finish all of them but let's have a try.

• Migration operator implementation to kubeflow/common fashion. (TF, PyTorch in pending)
  • Migrate controller implementation to kubeflow/common fashion training-operator#1171 @ChanYiLin
  • Migrate code implementation to kubeflow/common fashion pytorch-operator#282
• Graduate Operator to v1
  • MXNet Operator Graduate mxnet-operator to v1 mxnet-operator#66
  • XGBoost Operator Graduate XGBoost Operator to v1 xgboost-operator#44
  • MPI Operator Graduate MPI Operator to v1 mpi-operator#217

[issue-label-bot]

Issue-Label Bot is automatically applying the labels:

Label	Probability
area/operator	0.86
kind/feature	0.79

Please mark this comment with 👍 or 👎 to give our bot feedback!
Links: app homepage, dashboard and code for this bot.

[kf-label-bot-dev]

Issue-Label Bot is automatically applying the labels:

Label	Probability
area/operator	0.86
feature	0.79

Please mark this comment with 👍 or 👎 to give our bot feedback!
Links: app homepage, dashboard and code for this bot.

[terrytangyuan]

@Jeffwan Shouldn't we go through some sort of review process before the graduation of operators though? I don't think we would have sufficient time to finish that before 1.1 release.

[jlewi]

• the apps cd pipelines have been fixed
• application owners please review PRs to update your apps and ensure they get merged
  • https://github.com/kubeflow/manifests/pulls/kubeflow-bot

/cc @terrytangyuan @Jeffwan

[Jeffwan]

@terrytangyuan

Agree. I think once these operator meet done criterion, then they can be graduated. It would be better for us to connect owners and have some discussion. If we can not catch 1.1, I think it's fine.

[jlewi]

@terrytangyuan @Jeffwan any update?

[terrytangyuan]

I think we should be ready regarding manifests (@Jeffwan correct me if I am wrong). Although There seems to be more pending auto PRs. @Jeffwan Could you check whether they need to be merged to master and cherry-picked to 1.1 release branch?

[Jeffwan]

1.1 and 1.2 release are done. We may not use exact same version as Kubeflow in the future. WG operators will have their own branch or tags. We just need to make sure we bring the right version to kubeflow release.