Submit, monitor, resume, and report Slurm, RunPod, or local experiment jobs.
npx skills add https://github.com/wangdepin/raca-codex --skill run-job-wangdepin
Please help me install this Agent Skill. Skill: run-job Source: https://github.com/wangdepin/raca-codex/tree/main/.claude/skills/run-job Command: npx skills add https://github.com/wangdepin/raca-codex --skill run-job-wangdepin
This Skill removes the friction of running HPC or cloud experiment jobs by orchestrating script generation, submission, monitoring, artifact validation, and resumable execution so results are delivered reliably without manual babysitting.
Ask your agent to run the canary or submit the experiment to your target cluster with the provided job spec and then stream progress updates until the final validated results and dashboard link are produced.
High-intent search queries and answers about installing and using this skill.
To automate Slurm job submission, this workflow generates sbatch scripts, submits them to the cluster, and monitors status with periodic log-based health checks. It manages the complete experiment lifecycle from pre-flight validation to final artifact reporting.
Yes, you can resume a RunPod training job from a checkpoint after an OOM crash. The workflow enforces deterministic checkpointing and supports resuming interrupted execution automatically, ensuring experiments continue safely from their last valid state.
Yes, this job orchestration workflow supports incremental Hugging Face uploads during training. It manages an incremental artifact pipeline that uploads partial outputs progressively, validates artifacts, and synchronizes a dashboard with the latest results.
Pre-flight validation for experiment execution is a safety check performed before job submission. It verifies references, cluster configuration, connectivity, and resource fit to prevent silent parameter changes and ensure the compute environment is correctly configured.
To run SFT and RL training jobs locally without manual babysitting, use a template-driven job orchestration workflow. It automates script generation, monitors log health for truncation or NaN issues, and handles checkpoint-resume automatically.
Limitations of template-driven scripts for experiment execution include the strict requirement for pre-flight validation of references and cluster config. The workflow prevents silent parameter changes without user confirmation, requiring explicit setup for safe resumable execution.
Hello-QMValidate HPC cluster configurations via SSH before submitting VASP workflows.
ksopylaUpload model checkpoints and generate model cards on the Hugging Face Hub.
Zayne-spragueAutomate SLURM job execution from sbatch generation to artifact upload.
TianyuDuDiscover Slurm cluster configurations and validate job submissions via sbatch, salloc, and srun.
