Syncs tracked ML experiments with actual SLURM job states and metrics from logs.
npx skills add https://github.com/kang-jaehyun/Mac-the-researcher --skill sync-kang-jaehyun
Please help me install this Agent Skill. Skill: sync Source: https://github.com/kang-jaehyun/Mac-the-researcher/tree/main/.claude/skills/sync Command: npx skills add https://github.com/kang-jaehyun/Mac-the-researcher --skill sync-kang-jaehyun
Synchronize local experiment tracking with actual SLURM job states to ensure dashboards and records reflect real progress.
Run the sync to refresh active experiments against SLURM and update local tracking.
High-intent search queries and answers about installing and using this skill.
To sync local experiment tracking with SLURM job states, this Skill queries SLURM job_info, maps states like PENDING to queued and RUNNING to running, and updates local records to reflect real progress across multi-cluster setups.
Yes, you can extract metrics automatically. For completed SLURM experiments, the Skill tails logs to parse values like best_metric and val_loss, then updates local tracking records with the extracted results.
To track failed SLURM jobs, the Skill maps FAILED, CANCELLED, and TIMEOUT states to a failed status, captures error messages from logs, and appends them to local records for troubleshooting.
The best way to monitor active experiments across multi-cluster SLURM setups is to filter local records for non-terminal states, query SLURM for job_info, parse the runtime states, and generate a concise summary of all progress.
Yes, this SLURM synchronization method works without external dependencies. It loads tracked experiments using MCP, queries SLURM directly for job_info, parses the states, and updates local records independently.
Your local dashboard shows outdated progress because local records are not synchronized with actual SLURM job states. Running a sync refreshes active experiments against SLURM to ensure dashboards reflect real-time progress.
RobertFrenkenReport experiment, SLURM job, and pipeline status per dataset.
hongmaple0820Log and visualize ML training metrics via CLI and Python API.
GiorgioRicciardielloRun parameter sweeps over computational scripts and generate summary reports.
umich-foreseerDiagnose Slurm job failures by analyzing sacct, scontrol, and logs.
