[Release] Training operator 1.7.0 release

[johnugeorge]

This is the tracking issue for Training operator 1.7.0 release. Following features are planned in this release

Core

• Merge Kubeflow/common to training operator  #1714
• Kubernetes versions to be supported in the next training operator release #1803
• Support koordinator gang scheduler plugin #1746
• Implement suspend semantics #1859
• Set correct ENV for PytorchJob to support torchrun #1840
• Make scheduler-plugins the default gang scheduler. #1747

Documentation

• GPU benchmark image does not exist #1672
• Outdated GPU monitoring info in README #1805

Please add any if I missed anything

/cc @kubeflow/wg-training-leads @tenzen-y @andreyvelich

[tenzen-y]

@johnugeorge Thank you for creating this issue!

I'd like to support queueing Jobs by adding suspend semantics.

#1519

[tenzen-y]

Also, it is worth mentioning #1768.

[johnugeorge]

v1.7.0-rc.0 release tag is cut for training operator from 1.7 release branch

Training operator SDK 1.7.0rc0 has been released.

[github-actions]

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.

[github-actions]

This issue has been automatically closed because it has not had recent activity. Please comment "/reopen" to reopen it.