start-run

Robot learning training runs are initiated by selecting Stage 1, Stage 2, Stage 3, or rollout processes. The system manages execution across these multiple stages while supporting both local environments and cluster job submission.

Robot learning jobs can be submitted to a cluster environment. The system supports cluster job submission and implements intelligent dual-submission and cancellation features specifically for LRZ cluster job optimization.

Automated documentation for major robot learning runs is generated by the skill during initiation. It automatically creates dated configurations and documentation for major runs, streamlining the tracking of complex training and rollout processes.

Yes, the tool supports local execution for robot learning rollouts. It is environment agnostic, allowing users to seamlessly initiate and manage both rollout processes and training stages locally or on a cluster.

Dual-submission for cluster job optimization is an advanced feature that intelligently submits and cancels jobs on the LRZ cluster. This mechanism streamlines the management of complex robot learning training runs.