Archive verified bulk output directories into tar archives to free inodes.
npx skills add https://github.com/dongzhuoyao/deepresearch --skill sue-cleanup
Please help me install this Agent Skill. Skill: sue-cleanup Source: https://github.com/dongzhuoyao/deepresearch/tree/main/.codex/skills/sue-cleanup Command: npx skills add https://github.com/dongzhuoyao/deepresearch --skill sue-cleanup
This Skill solves the critical issue of inode exhaustion on shared parallel filesystems (such as LUMI Lustre and Snellius GPFS) caused by tens of thousands of small output files generated during ML/HPC scale-up experiments, which can block new runs and disrupt team workflows.
Invoke the sue-cleanup skill after a dryrun or fullrun completes to archive high-inode bulk output directories from your sandbox-scoped experiment workspace, ensuring all benchmark reports, ledgers, and checkpoints are preserved for downstream processing.
High-intent search queries and answers about installing and using this skill.
To free up inodes on shared parallel filesystems, you can archive high-inode bulk output directories into tar files. This skill safely compresses already-summarized experiment outputs while preserving critical artifacts like checkpoints and logs on platforms like LUMI and Snellius.
The best way to archive experiment outputs without losing critical artifacts is to use a sandbox-scoped archiving workflow that verifies all metrics are extracted first. It strictly preserves reports, progress CSVs, logs, checkpoints, and readiness stamps during the tar compression process.
Yes, this inode cleanup workflow supports both tmux and Slurm execution models for long-running remote archive operations. It operates exclusively on scoped single-workspace output trees across sandbox backends like LUMI, Snellius, Brev, AutoDL, and RunPod.
No, you cannot use this archiving process across multiple sandbox workspaces simultaneously. It operates exclusively on a single selected sandbox backend's workspace output tree to avoid unintended project-wide or multi-workspace modifications during the inode conservation process.
You should run the sandbox cleanup workflow immediately after a dryrun or fullrun completes. This ensures all benchmark reports, ledgers, and checkpoints are fully generated and preserved for downstream processing before high-inode bulk output directories are archived.
Archiving bulk output directories helps with inode exhaustion on Lustre or GPFS filesystems because it consolidates tens of thousands of small output files generated during ML experiments into a single tar file, freeing up massive amounts of inodes while maintaining data integrity.
OneWave-AIAggregate buyer-intent signals from job postings, funding rounds, and leadership changes.
mzqefOrchestrate multi-day reinforcement learning training campaigns with automated monitoring and checkpoint management.
jackspaceAutomate Codex CLI code analysis, refactoring, and editing with sandbox controls.
emilneuraz-aiApply defensive patterns to Bash scripts with strict mode and error trapping.
