perf-hybrid-context-parallel

Official

Optimize hierarchical context parallelism in Megatron-Bridge.

AuthorNVIDIA-NeMo
Version1.0.0
Installs0

System Documentation

What problem does it solve?

This Skill provides guidance and configuration instructions for enabling hierarchical context parallelism in Megatron-Bridge, helping users scale model training efficiently.

Core Features & Use Cases

  • Configuration Guidance: Explains how to set context_parallel_size and hierarchical_context_parallel_sizes.
  • Validation & Constraints: Details the validation rules for hierarchical parallelism parameters and available code anchors.
  • Use Case: Use this Skill when scaling Megatron models beyond KV heads or troubleshooting OOM issues related to context parallelism; for example, deploying a large model with layered parallel splits.

Quick Start

Refer to the provided documentation and code snippets to activate hierarchical context parallelism in your Megatron-Bridge setup.

Dependency Matrix

Required Modules

None required

Components

scriptsreferences

💻 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: perf-hybrid-context-parallel
Download link: https://github.com/NVIDIA-NeMo/Megatron-Bridge/archive/main.zip#perf-hybrid-context-parallel

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 471,000+ vetted skills library on demand.