What problem does it solve?
Manually managing Slurm benchmark jobs across remote HPC clusters leads to inconsistent workflows, missed job status updates, un-synced code errors, and wasted SSH round-trips that consume valuable model tokens.
Core Features & Use Cases
- Cluster-first SSH execution: All job operations run directly on the remote cluster host to avoid local environment mismatches.
- Automated file syncing: Syncs local code changes to the cluster via git, rsync, or scp before job submission to ensure jobs run the latest version.
- Token-efficient polling: Uses planned, combined polling for job status, logs, and output files to minimize unnecessary SSH calls.
Use case: ML researchers and HPC engineers running large-scale training benchmarks can use this skill to automate the full job lifecycle from submission to troubleshooting without manual intervention.
Quick Start
Use the remote-slurm-bootstrap skill to submit your benchmark training job to the remote Slurm cluster, sync your latest local code changes first, and monitor the job with optimized polling until it completes.
