sim2real-operate

To monitor Sim2Real VLM-to-RL pipelines on Kubernetes GPU clusters, manage job submission, track lifecycle status, and review reward trend reports generated by the orchestrator to verify active training runs.

To troubleshoot a stalled robotic simulation training run, perform preflight checks on S3 storage, verify cluster storage secrets, and review registry access to resolve orchestration failures.

Run preflight validation for Isaac sim and VLM trainer components by executing a health check on the current cluster configuration to verify S3 storage, registry access, and cluster readiness before job execution.

Yes, you need SkyPilot runbooks and S3-compatible object storage to operate Sim2Real workflows because they maintain pipeline state and artifact consistency during complex robotic simulation job execution.

Multi-stage job orchestration fails during cold-start on Kubernetes GPU infrastructure when S3 storage secrets, registry access, or cluster readiness checks fail before submitting complex VLM and trainer components.