merge-ov2

Official

Merge ViT and LLM into LlavaOnevision2 checkpoint.

AuthorEvolvingLMMs-Lab
Version1.0.0
Installs0

System Documentation

What problem does it solve?

Merges standalone ViT encoder and LLM into a unified LlavaOnevision2 HuggingFace checkpoint, preserving weights and enabling end-to-end parity checks.

Core Features & Use Cases

  • Remaps and merges visual encoder weights into a single cross-modality checkpoint.
  • Provides end-to-end validation comparing merged weights and inference outputs against originals.
  • Supports optional verification steps and guidance for Megatron-style conversions.

Quick Start

Run the merge_ov2 tool with your ViT encoder, LLM, and processor checkpoints to generate a unified LlavaOnevision2 checkpoint.

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: merge-ov2
Download link: https://github.com/EvolvingLMMs-Lab/LLaVA-OneVision-2/archive/main.zip#merge-ov2

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.