[ANNOUNCE] Kubeflow Trainer v2.2.0 is released!

[Andrey Velichkevich]

Hi Folks,

We are excited to share general availability of Kubeflow Trainer v2.2 release!

This release marks a major milestone for Cloud Native AI – expanding framework support, improving observability, and pushing deeper into HPC workloads.

Major highlights include:

• 📊 Training Progress & Metrics (no more “stuck” jobs)
  
• ⚡️ First-class JAX and XGBoost support
  
• 🧠 Flux Runtime for HPC + MPI
  
• ⏱️ Better control over TrainJob Lifecycle

Full details are available in the LinkedIn announcement and the blog post.

We have an exciting roadmap for distributed AI on Kubernetes – with upcoming work on Workload Aware Scheduling, multi-node NVLink, DRA for TrainJobs, and intelligent GPU configuration.

🙌 Huge thanks to all contributors who made this release possible.

Regards,

Andrey