Verified:GitHub

mcore-bump-base-image

Update PyTorch base image pins in GitHub and GitLab CI configurations.

Updated Apr 16, 2026

Search Tags:#docker #ci-cd #pytorch #ci #pinning #megatron-lm #nvcr

One-click install

npx skills add https://github.com/sayalinvidia/sayali-skills-test --skill mcore-bump-base-image

Or copy as Structured Prompt for Agent▼

Please help me install this Agent Skill.
Skill: mcore-bump-base-image
Source: https://github.com/sayalinvidia/sayali-skills-test/tree/main/skills/Megatron-Core/mcore-bump-base-image
Command: npx skills add https://github.com/sayalinvidia/sayali-skills-test --skill mcore-bump-base-image

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Bump the NVIDIA PyTorch base image across Megatron-LM CI so GitHub and GitLab pipelines stay in sync, preventing mismatched pins and flaky tests.

Core Features & Use Cases

Update both CI pins to the new nvcr.io/nvidia/pytorch:<YY.MM>-py3 tag in GitHub and GitLab configurations.
Validate synchronization with quick sanity checks and run the functional test loop, refreshing golden values when drift is detected.
Use case: when updating base images after a new PyTorch tag release to ensure CI remains green across both pipelines.

Quick Start

Update the PyTorch base image to the target tag in docker/.ngc_version.dev and the corresponding GitLab pins, then run functional tests.

Frequently Asked Questions about mcore-bump-base-image

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

FAQPage Schema

How do I update the PyTorch base image in Megatron-LM CI pins?▼

Update the PyTorch base image by changing the tag in docker/.ngc_version.dev and .gitlab/stages/01.build.yml to nvcr.io/nvidia/pytorch:<YY.MM>-py3. This synchronizes the base image across both GitHub and GitLab pipelines.

Why do my Megatron-LM CI tests fail after updating the NVCR PyTorch image?▼

CI tests fail after updating the NVCR PyTorch image when golden values drift. Running the functional test loop and refreshing golden values when drift is detected resolves the flaky tests and keeps pipelines green.

Does this CI image bump workflow support both GitHub and GitLab pipelines?▼

Yes, the CI image bump workflow updates the PyTorch base image in both GitHub and GitLab pins. It modifies docker/.ngc_version.dev and GitLab stage configurations to ensure both platforms reference the same nvcr.io base image.

What is the best way to sync CI base image pins across dev and lts branches?▼

The best way to sync CI base image pins is to apply the nvcr.io/nvidia/pytorch:<YY.MM>-py3 tag update across dev and potential lts pins. This prevents mismatched pins and flaky tests across both GitHub and GitLab pipelines.

When do I need to refresh golden values in Megatron-LM CI?▼

You need to refresh golden values in Megatron-LM CI when drift is detected during the functional test loop after bumping the PyTorch base image. This ensures the tests match the updated nvcr.io base image behavior.