adding-model-support

Guide engineers through five phases to add HF model family support to Megatron-Bridge.

2.8k|332|Updated Feb 25, 2026

Search Tags:#llm #mapping #provider #vlm #megatron-bridge #model-bridge #hf-config

Ready to run

npx skills add https://github.com/NVIDIA/skills --skill adding-model-support

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: adding-model-support
Source: https://github.com/NVIDIA/skills/tree/main/skills/Megatron-Bridge/adding-model-support
Command: npx skills add https://github.com/NVIDIA/skills --skill adding-model-support

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This guide helps engineers extend Megatron-Bridge to support new HF model families by providing a repeatable blueprint for bridges, providers, mappings, tests, and docs.

Core Features & Use Cases

Structured extension workflow: Phase-based guidance for discovery, bridge/provider/mappings, recipe scaffolding, tests, and documentation.
Model-agnostic patterns: Applicable to both LLMs and VLMs with recommended strategies for private HF configs, top-level vs sub-config mappings, and optional VLM providers.
Improved reliability: Ensures consistent mappings, provider hooks, and test coverage to minimize integration risk.

Quick Start

Follow Phase 1–5 to add support for a new HF model family by implementing the bridge, mapping, provider (if needed), recipes, tests, and docs.

Frequently Asked Questions about adding-model-support

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

FAQPage Schema

How do I add support for a new Hugging Face model family in Megatron-Bridge?▼

To add support for a new HF model family in Megatron-Bridge, follow the structured phase-based workflow covering discovery, bridge wiring, mapping, provider creation, recipe scaffolding, tests, and documentation.

Does the Megatron-Bridge extension pattern work for both LLMs and VLMs?▼

Yes, the Megatron-Bridge extension pattern is model-agnostic and applicable to both LLMs and VLMs, providing recommended strategies for top-level versus sub-config mappings and optional VLM providers.

What is the recommended workflow for creating bridges and providers for new models?▼

The recommended workflow is a five-phase process: Phase 1 discovery, Phase 2 bridge and provider wiring, Phase 3 recipe scaffolding, Phase 4 tests, and Phase 5 docs, ensuring consistent mappings and validation checks.

Can I use private HF configs when mapping new models in Megatron-Bridge?▼

Yes, you can use private HF configs when mapping new models. The bridge extension pattern includes specific guidance on handling local mappings and private configurations for new model families.

What steps are required to scaffold recipes and tests for a new bridge mapping?▼

Scaffolding recipes and tests requires following Phase 3 and Phase 4 of the workflow, which involve creating recipe structures and implementing test coverage to minimize integration risk and validate hooks.

When should I implement provider hooks for a new model bridge?▼

You should implement provider hooks during Phase 2 bridge and provider wiring if the new model requires custom integration logic, ensuring reliable mappings and proper validation checks throughout the process.