Skills
.Work. You
Rest

llava-onevision2-consistency

Official

Validate HF vs Megatron consistency end-to-end

Software Engineering#consistency#pytest#validation#model-checking#hf#megatron#mcore
AuthorEvolvingLMMs-Lab
Version1.0.0
Installs0

System Documentation

What problem does it solve?

This skill helps teams verify behavioral consistency between HuggingFace and Megatron/MCore checkpoints for LLaVA-OneVision2, ensuring reliable cross-framework performance.

Core Features & Use Cases

  • Cross-checks HF and Megatron/MCore across TP and PP configurations using a targeted pytest suite and a legacy script workflow.
  • Guides users through HF→mcore conversion when needed and provides clear test coverage including weight, encoder, and end-to-end LLM checks.
  • Use case: validate a newly released HF checkpoint against an existing Megatron/MCore deployment to ensure identical downstream behavior.

Quick Start

Run the consistency tests inside the container using run_consistency_tests.sh.

Dependency Matrix

Required Modules

None required

Components

Standard package

💻 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: llava-onevision2-consistency
Download link: https://github.com/EvolvingLMMs-Lab/LLaVA-OneVision-2/archive/main.zip#llava-onevision2-consistency

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.