skills.rest
dongzhuoyao
dongzhuoyaoCOMMUNITY
Explore all skills by this author →
Website𝕏@vtaohuGitHub

single-gpu-dryrun

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

Updated Aug 23, 2026
Search Tags:#dry-run#slurm#ml-experiments#lumi#parameter-counting#gpu-probe#production-gating
One-click install
npx skills add https://github.com/dongzhuoyao/deepresearch --skill single-gpu-dryrun
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: single-gpu-dryrun
Source: https://github.com/dongzhuoyao/deepresearch/tree/main/.codex/skills/single-gpu-dryrun
Command: npx skills add https://github.com/dongzhuoyao/deepresearch --skill single-gpu-dryrun

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Running full ML production experiments without prior validation often leads to wasted GPU hours, failed runs due to missing dependencies or configuration errors, and unexpected inode consumption on shared HPC clusters. This Skill eliminates that risk by enabling small, real single-GPU probe runs that catch issues early.

Core Features & Use Cases

  • LUMI dev-g Pre-Production Probe: Runs tiny, real experiments on LUMI's dev-g partition to validate environment setup, model construction, and basic training steps before launching expensive full runs.
  • Accurate Parameter Counting: Tallies total, trainable, and non-trainable model parameters using real tensor operations, avoiding unreliable estimates from config files or research papers.
  • Production Gating Validation: Includes mandatory inode consumption risk checks and minimal evaluation path testing (e.g., FID for generative models) to ensure probes are valid gates for full production launches.
  • Use Case: If you are preparing a large-scale image generation experiment on LUMI, use this Skill to run a 1-step dry run first to catch missing dependencies, config errors, or inode shortages before wasting hundreds of GPU hours on a failed full run.

Quick Start

Use the single-gpu-dryrun skill to run a validation probe for your LUMI ML experiment before launching the full production run.

Frequently Asked Questions about single-gpu-dryrun

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

FAQPage Schema
How do I validate ML experiments before launching expensive full production runs?▼

You validate ML experiments by executing small single-GPU dry runs that catch environment, configuration, and resource issues early. This probe approach prevents wasted GPU hours and failed runs before expensive full production runs on HPC clusters.

How do I count trainable parameters accurately in a model instead of relying on config estimates?▼

To count parameters accurately, execute a real single-GPU probe job that tallies total, trainable, and non-trainable parameters using actual tensor operations. This avoids unreliable estimates from config files or research papers.

Can I test Slurm job configurations and inode consumption risks on LUMI dev-g before full scale-up?▼

Yes, you can test Slurm job configurations on LUMI dev-g by running tiny real experiments that include mandatory inode consumption risk checks. This single-GPU sandbox environment validates setup before full production launches.

What is production gating validation for ML experiments and when do I need it?▼

Production gating validation is the process of testing minimal evaluation paths, like FID for generative models, and checking inode consumption risks to ensure a probe run is a valid gate. You need it before authorizing expensive full production runs.

Why do my large-scale ML experiments fail on HPC clusters without a dry run?▼

Experiments fail without a dry run due to missing dependencies, configuration errors, and unexpected inode consumption on shared HPC clusters. Running a small probe first catches these issues before wasting hundreds of GPU hours.

Related Skills

Explore More Software Engineering Skills→
dongzhuoyaodongzhuoyao
10

lumi-supercomputer

Configure Slurm and ROCm environments for PyTorch jobs on LUMI.

Community
Advanced
dongzhuoyaodongzhuoyao
10

slurm-gpu-training

Submit GPU training jobs to Slurm clusters with preflight validation and monitoring.

Community
Advanced
TianyuDuTianyuDu

slurm-hpc-operator

Discover Slurm cluster configurations and validate job submissions via sbatch, salloc, and srun.

Community
Advanced
ehoyos007ehoyos007

startup-hook-skill

Generate and validate SessionStart hooks that install dependencies for Claude Code web sessions.

Community
Intermediate