skills.rest
yo-steven
yo-stevenCOMMUNITY
Explore all skills by this author →
GitHub

ad-model-onboard

Translate HuggingFace models into prefill-only AutoDeploy custom models with validation.

Updated May 23, 2026
Search Tags:#huggingface#model registry#model onboarding#tensorrt-llm#autoredeploy#torch export#equivalence testing
One-click install
npx skills add https://github.com/yo-steven/skills-exploration-20260522 --skill ad-model-onboard-yo-steven
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ad-model-onboard
Source: https://github.com/yo-steven/skills-exploration-20260522/tree/main/skills/TensorRT-LLM/ad-model-onboard
Command: npx skills add https://github.com/yo-steven/skills-exploration-20260522 --skill ad-model-onboard-yo-steven

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Onboarding a new HuggingFace model into TensorRT-LLM AutoDeploy is error-prone and often fails late, wasting time on export and deployment surprises.

Core Features & Use Cases

  • Translate to a prefill-only AutoDeploy custom model using a minimal, audit-friendly modeling file designed for torch.export.
  • Validate correctness with hierarchical equivalence tests that compare custom ops against HuggingFace reference behavior.
  • Register models and run end-to-end AutoDeploy using standalone registry YAML so the cookbook and deployment paths stay aligned.

Quick Start

Use ad-model-onboard to create a prefill-only AutoDeploy custom model, then verify it with hierarchical equivalence tests and a full end-to-end AutoDeploy run using the standalone registry config YAML.

Frequently Asked Questions about ad-model-onboard

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

FAQPage Schema
How do I onboard a HuggingFace model into TensorRT-LLM AutoDeploy?▼

Translating a HuggingFace model into TensorRT-LLM AutoDeploy requires creating a prefill-only custom model with deterministic canonical ops, updating the standalone registry YAML, and running end-to-end build verification.

What tests are needed to validate a HuggingFace model for AutoDeploy onboarding?▼

Validating a HuggingFace model for AutoDeploy onboarding requires hierarchical equivalence tests comparing custom ops against reference behavior, plus torch.export tests ensuring forward input contract compliance with input_ids and position_ids.

Why does exporting a HuggingFace model to TensorRT-LLM often fail late?▼

Exporting HuggingFace models to TensorRT-LLM fails late due to unexpected export and deployment issues, which AutoDeploy onboarding mitigates by applying robust input and memory guardrails alongside deterministic canonical ops.

Does TensorRT-LLM AutoDeploy require position_ids for HuggingFace model translation?▼

Yes, TensorRT-LLM AutoDeploy requires forward input contract compliance, meaning both input_ids and non-null position_ids must be provided when translating a HuggingFace model into a prefill-only custom model.

When should I use a prefill-only custom model for TensorRT-LLM AutoDeploy?▼

A prefill-only custom model is needed when onboarding new HuggingFace architectures or missing model families into TensorRT-LLM AutoDeploy, ensuring the modeling file remains minimal and audit-friendly for torch.export.

What is the best way to keep cookbook and deployment paths aligned during model onboarding?▼

The best way to align cookbook and deployment paths is using a standalone registry YAML configuration, ensuring both paths run the same end-to-end AutoDeploy verification for the newly onboarded HuggingFace model.

Related Skills

Explore More Software Engineering Skills→
yo-stevenyo-steven

ad-add-fusion-transformation

Add TensorRT-LLM AutoDeploy fusion transformation passes for graph pattern rewriting.

Community
Advanced
ancolemanancoleman
503

model-serving

Deploy LLM and ML models for production inference with vLLM, TensorRT-LLM, Ollama, and BentoML.

Community
Advanced
KristopherlbKristopherlb

manifest-cookbook-authoring

Create validated Shinobi manifest cookbook patterns with YAML examples and deployment documentation.

Community
Intermediate
olibartfastolibartfast

deepstream-import-vision-model

Automate HuggingFace or NVIDIA NGC object detection model integration into NVIDIA DeepStream pipelines.

Community
Advanced