skills.rest
sayalinvidia
sayalinvidiaCOMMUNITY
Explore all skills by this author →
GitHub

mcore-build-and-dependency

Set up containerized Megatron-LM environments with CUDA toolchain and uv dependency locking.

Updated Apr 16, 2026
Search Tags:#uv#dependencies#container#cuda#dev-ops#megatron-lm
One-click install
npx skills add https://github.com/sayalinvidia/sayali-skills-test --skill mcore-build-and-dependency
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: mcore-build-and-dependency
Source: https://github.com/sayalinvidia/sayali-skills-test/tree/main/skills/Megatron-Core/mcore-build-and-dependency
Command: npx skills add https://github.com/sayalinvidia/sayali-skills-test --skill mcore-build-and-dependency

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Megatron-LM development and container setup often require aligning CUDA toolchains, PyTorch builds, and pre-compiled extensions across hosts. This skill provides a repeatable, container-based workflow to acquire the correct environment, manage dependencies with uv, and lock exact versions to ensure reproducibility.

Core Features & Use Cases

  • Containerized development environment for Megatron-LM with correct CUDA toolkit and pre-compiled extensions.
  • Dependency management inside the container using uv, including adding, syncing, and locking dependencies.
  • Separate dev vs lts workflows to guarantee stability and reproducibility across teams and CI pipelines.

Quick Start

Launch a containerized Megatron-LM workspace and run the recommended uv commands to set up the environment.

Frequently Asked Questions about mcore-build-and-dependency

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

FAQPage Schema
How do I set up a Megatron-LM development environment with the correct CUDA toolchain?▼

This skill provides a container-based workflow to acquire the correct environment, aligning the CUDA toolchain, PyTorch builds, and pre-compiled extensions for Megatron-LM to ensure reproducible builds across hosts.

How do I manage Python dependencies inside a CUDA container using uv?▼

You can manage dependencies inside a CUDA container using uv to add, update, sync, and lock exact package versions, generating a uv.lock file that guarantees reproducibility across team and CI environments.

What is the difference between dev and lts container variants for reproducible builds?▼

Dev and lts container variants provide separate workflows for Megatron-LM, where dev supports active development and lts guarantees long-term stability and reproducibility across teams and CI pipelines.

Does uv work with pre-compiled extensions in a Megatron-LM container?▼

Yes, uv works within the containerized Megatron-LM environment to manage dependencies alongside pre-compiled extensions, locking exact versions via uv.lock to maintain consistent CUDA and PyTorch builds.

Why do I need version pinning for Megatron-LM containerized CI workflows?▼

Version pinning is needed because Megatron-LM requires aligning CUDA toolchains, PyTorch builds, and pre-compiled extensions across hosts, and locking exact versions with uv guarantees reproducible containerized CI workflows.

Related Skills

Explore More Software Engineering Skills→
jayll1303jayll1303
17

docker-gpu-setup

Build GPU-enabled Docker containers with NGC base images and uv.

Community
Advanced
robandersonrobanderson

uv-python-manager

Manage Python projects, dependencies, and virtual environments with UV.

Community
Advanced
zangjeicyzangjeicy

slime-rl-training

Train LLMs with GRPO and PPO using Megatron-LM and SGLang.

Community
Advanced
while-basicwhile-basic

enact/hello-python

Print a personalized greeting with timestamp and Python version.

Community
Basic