Tao Run On Slurm

Bridges TAO training jobs to on-prem SLURM clusters over SSH. You give it credentials (SLURM_USER, SLURM_HOSTNAME, SSH key), point it at Lustre datasets, and it handles the sbatch submission, Pyxis container execution, and status polling. Built for DGX clusters with shared storage. The preflight checks are thorough, including a heredoc trick to install NGC credentials without leaking your key into shell history. Multi-node PyTorch jobs get the right WORLD_SIZE and rendezvous variables automatically. Default wall time is 4 hours, not 12, which matches the packaged partition list. If you're running TAO workloads on managed GPU infrastructure instead of local docker, this is the connector.

skills.sh installs: 1.6K
GitHub stars: 3.9K
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add nvidia/skills --skill tao-run-on-slurm --agent claude-code

Installs into .claude/skills of the current project.

Files

SLURM

Standalone install? If this session was not initialized by the TAO skill bank plugin, run the tao-setup skill first (host preflight, credentials, cross-skill discovery).

Remote GPU compute platform for clusters managed by SLURM. Jobs are submitted from the launch host to a login node over SSH, staged on a shared filesystem, submitted with sbatch, and executed with srun container support.

When to use

Use SLURM when the user has access to a managed GPU cluster, shared Lustre storage, and scheduler-owned GPU allocation. Do not use SLURM for local files that exist only on the agent machine; data and outputs must be reachable from the cluster.