Skip to content

[Docs] Clarify model platform support and Ascend setup - #201

Merged
bjf-frz merged 6 commits into
mainfrom
docs/platform-support
Nov 20, 2026
Merged

bjf-frz merged 6 commits into
mainfrom
docs/platform-support

Conversation

@bjf-frz

@bjf-frz bjf-frz commented Nov 20, 2026 •

Copy link
Copy Markdown
Collaborator

Required human contributor answers

1. What outcome are you trying to achieve?

Make supported platforms easy to find in the README and documentation, including the Ascend NPU setup.

2. What is the highest-leverage variable for that outcome?

The model-to-platform mapping in the README's Model Support table.

3. How will this work move that variable rather than create busywork?

Add a Platforms column to the existing model table and link the existing device guides from Getting Started, the documentation index, and the first-run and serving guides.

4. Who is the customer or user, and what do they actually need?

Readers selecting a model and device need to see the supported platform and the corresponding setup. Ascend users need the supported Ouro-1.4B configuration and a checkout of main.

PR necessity and bugfix evidence

A documentation PR was requested. The release entry point accepts CPU and CUDA. The Ascend integration landed after v0.2.0. I verified the distinction by comparing the tag and the current main branch.

Purpose

  • List the platforms for each checkpoint in the README and retain the existing qualification notes.
  • Add setup entry points for NVIDIA CUDA and Huawei Ascend NPU.
  • Document which checkout to use for the Ascend recipe and link it from the documentation index, first-run guide, and serving guide.

Test Plan

vllm-rlt Version / Commit: 364ee4c on efb5cd0.

  • Run pre-commit run --all-files and git diff --check.
  • Check relative file links and heading anchors, the seven model table rows, and the Ascend setup block's shell syntax.
  • Compare the device arguments in the v0.2.0 and main serving entry points.

Test Result

  • Ruff lint, Ruff formatting, and the CPU regression suite passed through pre-commit on the initial documentation commit e69ff58.
  • Checked the changed relative links and anchors, the Ascend setup's shell syntax, and GitHub's rendering of all seven model rows. The latest check confirmed the direct README link to the Torch-NPU installation steps.
  • Confirmed that --device npu is available on main and was added after v0.2.0.

@bjf-frz
bjf-frz merged commit ce200f0 into main Nov 20, 2026
1 check passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant