mcore-onboard-gb200-1node-tests

Generate 1-node GB200 model_config.yaml variants and update MR test recipes.

Updated Apr 16, 2026
One-click install
npx skills add https://github.com/sayalinvidia/sayali-skills-test --skill mcore-onboard-gb200-1node-tests
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: mcore-onboard-gb200-1node-tests
Source: https://github.com/sayalinvidia/sayali-skills-test/tree/main/skills/Megatron-Core/mcore-onboard-gb200-1node-tests
Command: npx skills add https://github.com/sayalinvidia/sayali-skills-test --skill mcore-onboard-gb200-1node-tests

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Onboard 1-node GB200 GitHub MR tests from existing 2-node tests.

Core Features & Use Cases

  • Generate 1-node (4-GPU) variants of existing 2-node GB200 tests for the Megatron-Core suite.
  • Create corresponding _1node model_config.yaml directories and update test recipes to reflect 1-node configurations.
  • Align scope and branching for MR tests with the mr-github workflow.

Quick Start

Copy existing 2-node GB200 test configs and adapt them to 1-node _1node variants.

Frequently Asked Questions about mcore-onboard-gb200-1node-tests

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

FAQPage Schema
How do I convert 2-node GB200 tests to 1-node configurations for Megatron-Core?

To convert 2-node GB200 tests to 1-node variants, generate _1node model_config.yaml directories and adapt tensor and parallelism settings for a 4-GPU setup. You also update test recipes and align branching scope with the mr-github workflow.

What is a 1-node GB200 test variant in Megatron-Core?

A 1-node GB200 test variant is a 4-GPU configuration derived from existing 2-node Megatron-Core test suites. It involves creating _1node directories and adjusting parallelism settings to validate changes within a single-node environment.

How do I wire Megatron-Core MR tests with the mr-github scope?

You wire Megatron-Core MR tests with the mr-github scope by updating the test recipe scope during the 1-node onboarding process. This ensures the 1-node GB200 configurations align with GitHub MR branching workflows.

Does onboarding 1-node GB200 tests require changing tensor parallelism settings?

Yes, onboarding 1-node GB200 tests requires adapting tensor and parallelism settings. When generating the _1node model_config.yaml from existing 2-node tests, configurations must be adjusted to support the 4-GPU single-node architecture.

What is the best way to create 1-node test directories for Megatron-Core GB200 suites?

The best way to create 1-node test directories is to copy existing 2-node GB200 test configs and generate dedicated _1node directories. You then populate them with adapted model_config.yaml files and update the corresponding test recipes and mr-github scopes.