Vision Language Models

This directory contains documentation for Vision Language Models (VLMs) supported by Megatron Bridge. These models combine vision and language capabilities for multimodal AI applications.

Available Models

Megatron Bridge supports the following VLM families:

Model	Documentation	Description
Gemma 3 VL	gemma3-vl.md	Google Gemma 3 Vision Language model
Nemotron Nano V2 VL	nemotron-nano-v2-vl.md	NVIDIA Nemotron Nano V2 Vision Language model
Qwen2.5 VL	qwen2.5-vl.md	Alibaba Cloud Qwen2.5 Vision Language model
Qwen3 VL	qwen3-vl.md	Alibaba Cloud Qwen3 Vision Language model

Quick Navigation

I want to

🔍 Find a specific VLM model → Browse the model list above or use the index page

🔄 Convert models between formats → Each model page includes conversion examples for Hugging Face ↔ Megatron Bridge

🚀 Get started with training → See Training Documentation for training guides

📚 Understand VLM architecture → Each model page documents vision-language architecture features

🔧 Add support for a new VLM → Refer to Adding New Models

Related Documentation

• Models Overview - Return to main models documentation

• Large Language Models - LLM model documentation

• Training Documentation - Training and customization guides

• Bridge Guide - Working with Hugging Face models

• Adding New Models - Extending model support

Vision Language Model Features

VLMs typically support:

• Image Understanding - Processing and understanding visual inputs

• Multimodal Fusion - Combining vision and language representations

• Vision-Language Tasks - Image captioning, visual question answering, and more

• Cross-Modal Learning - Learning relationships between visual and textual data

---

Ready to explore? Choose a model from the list above or return to the main documentation.