run-experiment

To run ML experiments on a remote GPU server, this skill automates environment detection, code synchronization, and remote execution via SSH. It deploys your scripts using screen sessions and captures logs with tee for reliable monitoring across multiple machines.

Yes, you can track remote GPU training runs with W&B through optional experiment tracking integration. This feature allows data scientists to monitor metrics and logs in real-time while the skill handles the automated deployment and execution across local or remote machines.

Automating environment setup and GPU scheduling is handled through pre-flight GPU checks and resource binding before execution. This ensures your machine learning training jobs have the necessary GPU resources allocated and dependencies configured before deployment begins.

To safely resume a machine learning training job after a connection drop, this skill implements safe resumable operation across hosts using screen sessions. This allows your deployed training process to continue running independently on the remote GPU server despite network interruptions.

No, you do not need manual configuration to synchronize code before deploying to multiple GPU machines. The skill automatically handles code synchronization during the deployment phase, ensuring your latest experiment scripts are correctly pushed to each target host.