llava-onevision2-consistency

Name: llava-onevision2-consistency
Availability: InStock
Author: LMMs-Lab

Compare HuggingFace and Megatron/MCore checkpoints for LLaVA-OneVision2 behavioral consistency.

1.2k|78|Updated Sep 16, 2025

Search Tags:#consistency #pytest #validation #model-checking #hf #megatron #mcore

Ready to run

npx skills add https://github.com/EvolvingLMMs-Lab/LLaVA-OneVision-2 --skill llava-onevision2-consistency

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: llava-onevision2-consistency
Source: https://github.com/EvolvingLMMs-Lab/LLaVA-OneVision-2/tree/main/.opencode/skills/llava-onevision2-consistency
Command: npx skills add https://github.com/EvolvingLMMs-Lab/LLaVA-OneVision-2 --skill llava-onevision2-consistency

SYSTEM DOCUMENTATION & REQUIREMENTS

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.

Frequently Asked Questions about llava-onevision2-consistency

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

FAQPage Schema

How do I validate HuggingFace and Megatron checkpoint consistency for LLaVA-OneVision2?▼

To validate HuggingFace and Megatron checkpoint consistency for LLaVA-OneVision2, run the targeted pytest suite and legacy script workflow via run_consistency_tests.sh to verify behavioral parity across TP and PP configurations.

What environment variables do I need to run HF vs Megatron consistency tests?▼

Running HF vs Megatron consistency tests requires configuring HF_MODEL_PATH, MCORE_CHECKPOINT_PATH, CONSISTENCY_TEST_TP, and CONSISTENCY_TEST_PP in your pytest environment to execute cross-framework validation.

What does cross-framework checkpoint consistency testing cover for LLaVA models?▼

Cross-framework checkpoint consistency testing for LLaVA models covers weight, encoder, and end-to-end LLM checks to ensure reliable behavioral parity between HuggingFace and Megatron/MCore deployments.

Can I use a pre-converted MCore artifact for Megatron consistency validation?▼

You can use a pre-converted MCore artifact for Megatron consistency validation, or the skill guides you through HF to mcore conversion when needed to verify identical downstream behavior.

Does HuggingFace to Megatron consistency validation work across different TP and PP configurations?▼

HuggingFace to Megatron consistency validation works across different TP and PP configurations, enabling reproducible cross-checks in varied hardware setups using both the fast pytest suite and legacy scripts.