skills.rest
NVIDIA
NVIDIAOFFICIAL
Explore all skills by this author →
WebsiteGitHub

cicd

Determine Megatron-LM CI test scope and trigger internal pipelines from PR labels.

17.4k|4.3k|Updated Mar 21, 2019
Search Tags:#ci/cd#github actions#test scope#log investigation#gitlab pipeline#pull request labels#artifact debugging
One-click install
npx skills add https://github.com/NVIDIA/Megatron-LM --skill cicd-nvidia
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: cicd
Source: https://github.com/NVIDIA/Megatron-LM/tree/main/skills/cicd
Command: npx skills add https://github.com/NVIDIA/Megatron-LM --skill cicd-nvidia

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires gh, python, git, and includes references (resource) components.

What problem does it solve?

It helps you understand why Megatron-LM CI is running (or failing) and what PR label or internal trigger to use to get the right test coverage, without wasting time guessing the pipeline structure or where logs live.

Core Features & Use Cases

  • CI pipeline structure mapping: Explains how the main workflow fans out into pre-flight steps, container builds, unit tests, integration tests, and the final gate.
  • PR test-scope decisioning: Provides a label-based ruleset to choose lightweight vs functional scopes and the correct number of repetitions.
  • Internal CI triggering: Documents how to push your branch to the internal GitLab remote and start a pipeline via script, including dry-run usage.
  • Failure investigation workflow: Shows how to locate the PR from a CI branch, list runs, view logs for failures, and download per-rank log artifacts for root-cause analysis.

Quick Start

Use the cicd skill to diagnose a failing GitHub Actions run by identifying the CI branch’s PR number, selecting the correct test-scope label to rerun, and downloading the per-rank log artifacts to pinpoint the failing stage.

Frequently Asked Questions about cicd

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

FAQPage Schema
How do I determine the correct test scope and repeat counts for a Megatron-LM pull request?▼

To determine test scope, apply specific pull request labels to route CI pipelines to the correct lightweight or functional test scopes and set the required repetition counts.

How do I trigger an internal GitLab CI pipeline from a GitHub pull request branch?▼

Trigger internal GitLab CI by pushing your branch to the internal remote and executing the tools/trigger_internal_ci.py script, which supports dry-run usage to validate pipeline configuration.

What is the best way to investigate a failing GitHub Actions run for Megatron-LM?▼

Investigate a failing GitHub Actions run by identifying the CI branch PR number, listing runs with gh commands, and downloading per-rank log artifacts to pinpoint the failing stage.

How does the Megatron-LM CI pipeline structure fan out into different test stages?▼

The Megatron-LM CI pipeline fans out from main workflows into pre-flight steps, container builds, unit tests, integration tests, and a final gate to validate code changes comprehensively.

Do I need gh and git installed to debug per-rank log artifacts from CI failures?▼

Yes, you need gh and git installed, as the failure investigation workflow relies on gh run commands to list runs, view logs, and download per-rank log artifacts for root-cause analysis.

Related Skills

Explore More Software Engineering Skills→
EdanStarfireEdanStarfire
15

github-authenticator

Authenticate GitHub CLI credentials and troubleshoot authentication failures via gh commands.

Community
Intermediate
jMertajMerta
130

ci-fix

Diagnose and patch failing GitHub Actions CI workflows using the GitHub CLI.

Community
Intermediate
josuenavarroaguadojosuenavarroaguado

ci-cd-and-automation

Automate GitHub Actions CI/CD pipelines with quality gates and preview deployments.

Community
Advanced
apacheapache
79

magpie-flaky-test-triage

Classify GitHub Actions CI jobs as flaky, broken, or clean from run data.

Official
Intermediate