Skills
.Work. You
Rest

debug-distributed

Official

Diagnose and fix distributed training issues fast

Software Engineering#debugging#troubleshooting#PyTorch#diagnostics#distributed-training#NCCL#FSDP
AuthorinclusionAI
Version1.0.0
Installs0

System Documentation

What problem does it solve?

This skill guides engineers to diagnose and fix distributed training issues in AReaL, including hangs, inconsistent results, OOM, and NCCL/communication errors that can occur across ranks.

Core Features & Use Cases

  • Minimal reproduction workflows: Create small, reproducible examples to isolate the failing operation.
  • Rank-aware diagnostics: Verify group membership, tensor shapes, and synchronization across all ranks.
  • Environment & tooling guidance: Provide environment variable tips and debugging tools to collect actionable signals.

Quick Start

Diagnose and fix distributed training issues in AReaL by following the guided workflow and applying the recommended checks to reproduce and resolve failures.

Dependency Matrix

Required Modules

None required

Components

Standard package

💻 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: debug-distributed
Download link: https://github.com/inclusionAI/AReaL/archive/main.zip#debug-distributed

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.