NVIDIA / NVIDIA/TensorRT-LLM

Add a AD specific config and trtllm common config via in model registry

Open
#13,697 3 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

AutoDeploy Community want to contribute
Dominant language
Python
Stars
14.7k
Forks
2.8k
Avg merge
2d 23h
Merged PRs (30d)
489

Description

Set all user exposed and/or trtllm common knobs in <model>.yaml
AD specific configs for the model in <model>_ad_default.yaml (hidden from the user and trtllm-serve command)

If the backend is autodeploy, it loads <model>_ad_default.yaml  and then <model>.yaml
If trtllm pytorch (let's say the model works in both workflows), it just loads <model>.yaml

For the user, the trtllm-serve command and args/config yaml stay the same between AD and pytorch flow except for adding --backend _autodeploy  for AD.

model registry should support this easily

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by tracing the model registry's config-loading entry point and the .yaml and _ad_default.yaml paths named in the issue. Check how trtllm-serve selects the autodeploy and PyTorch flows, then verify that shared user-facing settings remain consistent while AD-specific settings stay hidden.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
backend, cli
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Quiet
Clarity
Mostly clear
Newbie friendliness
38/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.