skills.rest
NVIDIA
NVIDIAOFFICIAL
Explore all skills by this author →
WebsiteGitHub

exec-env-check

Detects GPU availability, Docker support, and Slurm access to determine the execution scenario.

14.5k|2.7k|Updated Aug 16, 2023
Search Tags:#docker#environment check#cluster#Infrastructure#nvidia-smi#gpu detection#Slurm
One-click install
npx skills add https://github.com/NVIDIA/TensorRT-LLM --skill exec-env-check
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: exec-env-check
Source: https://github.com/NVIDIA/TensorRT-LLM/tree/main/.claude/skills/exec-env-check
Command: npx skills add https://github.com/NVIDIA/TensorRT-LLM --skill exec-env-check

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Before launching a GPU workload, you need to know whether the current machine can run it locally, inside Docker, or through a Slurm cluster. This Skill probes the environment and returns a structured execution scenario with GPU counts and hardware details, removing guesswork from workload placement.

Core Features & Use Cases

  • Local GPU Detection: Runs nvidia-smi with a timeout to count available GPUs and identify the device type (e.g., B200, H100, A100).
  • Docker vs Direct Execution: Checks whether the Docker CLI and daemon are usable to distinguish satisfied, local, docker from satisfied, local, direct scenarios.
  • Slurm Cluster Resolution: On login nodes without local GPUs, delegates hostname-based cluster identification to the internal-env-info skill to obtain GPU type, GPUs per node, and default storage paths.
  • Use Case: An orchestrator needs to schedule a 4-GPU TensorRT-LLM benchmark. It calls this Skill with required_devices: 4 and receives a scenario such as satisfied, slurm, local along with cluster name and per-node GPU count to route the job correctly.

Quick Start

Check whether this machine can run a workload requiring 4 GPUs and report the execution scenario, GPU count, and device type.

Frequently Asked Questions about exec-env-check

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I check GPU availability before running a workload?▼

Run nvidia-smi with a timeout to query GPU names and memory, then count the returned lines as available GPUs. This Skill automates that check and compares the count against your required_devices input to decide the execution scenario.

How to detect if Docker is usable for GPU jobs?▼

Verify both that the Docker CLI exists and that the daemon responds to docker info within a short timeout. If both succeed the scenario is satisfied, local, docker; otherwise it falls back to satisfied, local, direct when GPUs are present.

Can I detect Slurm cluster GPU type from a login node?▼

Yes, by capturing the hostname and delegating cluster identification to the internal-env-info skill, which maps hostname patterns to cluster names, GPU types, and GPUs per node. No srun allocation on compute nodes is needed.

What happens if the internal-env-info skill is not installed?▼

The scenario falls back to not_satisfied with null values for device type, cluster name, and storage paths. This is treated as a normal outcome, not an error, and no compute-node probing via srun is attempted.

Why does nvidia-smi hang and how is it handled?▼

nvidia-smi can hang on systems with driver issues or inaccessible GPUs. The Skill wraps the call in a 5-second timeout and treats failure or empty output as zero available GPUs, then proceeds to Slurm detection.

Related Skills

Explore More Software Engineering Skills→
dongzhuoyaodongzhuoyao

single-gpu-dryrun

Run small single-GPU probe jobs to validate ML experiments before production.

Community
Advanced
kang-jaehyunkang-jaehyun

submit

Submit machine-learning training jobs to a SLURM cluster with resource requirements.

Community
Advanced
X-School-AcademyX-School-Academy
19

terminal-forward-remote-to-local

Forward a local port to a remote host via SSH tunneling.

Official
Intermediate
qcmuuqcmuu

lambda-labs-gpu-cloud

Provision Lambda Labs GPU instances and Slurm clusters for ML workloads.

Community
Intermediate