nvidia-resiliency-ext

Community

Enhance NVIDIA application resiliency.

Authorpedestrianlove
Version1.0.0
Installs0

System Documentation

What problem does it solve?

This Skill provides tools and utilities to enhance the resiliency of NVIDIA applications, particularly in distributed computing environments, by offering fault tolerance mechanisms and diagnostic capabilities.

Core Features & Use Cases

  • Fault Injection: Simulate various failure scenarios (e.g., rank failures, hangs) to test application robustness.
  • Checkpointing Utilities: Asynchronous and robust checkpointing mechanisms for saving and resuming training states.
  • Monitoring and Diagnostics: Tools for monitoring process health, capturing logs, and analyzing execution traces.
  • Use Case: During a large-scale distributed training job, simulate a GPU failure to ensure the training process can gracefully recover and resume from the last checkpoint without significant data loss.

Quick Start

Use the nvidia-resiliency-ext skill to inject a simulated GPU error after a 60-second delay.

Dependency Matrix

Required Modules

None required

Components

scriptsreferencesassets

💻 Claude Code Installation

Recommended: Let Claude install automatically. Simply copy and paste the text below to Claude Code.

Please help me install this Skill:
Name: nvidia-resiliency-ext
Download link: https://github.com/pedestrianlove/skills/archive/main.zip#nvidia-resiliency-ext

Please download this .zip file, extract it, and install it in the .claude/skills/ directory.
View Source Repository

Agent Skills Search Helper

Install a tiny helper to your Agent, search and equip skill from 223,000+ vetted skills library on demand.