Hi Everyone,
I'm happy to announce the release of Kubeflow Trainer v2.3.0 🎉
Install with a single command:
helm install kubeflow-trainer oci://ghcr.io/kubeflow/charts/kubeflow-trainer --namespace kubeflow-system --create-namespace --version 2.3.0 --set runtimes.defaultEnabled=true
⚠️ Breaking Changes
1. CRDs have moved into the Helm chart template directory. If you are running v2.0, v2.1, or v2.2, you must upgrade to v2.3 before upgrading to any later version. See the migration instruction
2. Runtime Finalizers have been removed
Highlights
✅ Trainer now has its own docs site at with the new users guides: https://trainer.kubeflow.org
✅ Multi-cluster orchestration with managedBy and MultiKueue
✅ Runtime Snapshot mechanism to decouple the lifecycle of TrainingRuntimes from TrainJobs
✅ Ability to inject PyTorch environment variables to initContainers for pre-flight checks
✅ Enhanced MPI orchestration – the launcher now waits for worker readiness before starting
✅ Proposal for OptimizationJob to build a Kubernetes-native API for hyperparameter tuning as the next iteration of the Katib project
Check the full release notes here: https://github.com/kubeflow/trainer/releases/tag/v2.3.0
Huge thanks to everyone who contributed to this release!
If you have any questions, please reach out in #kubeflow-trainer Slack channel or join the Trainer community calls.