Skip to content

docs: describe local_vllm_model and extend docs for vllm_model#1430

Open
marta-sd wants to merge 5 commits into
mainfrom
martas/1106
Open

docs: describe local_vllm_model and extend docs for vllm_model#1430
marta-sd wants to merge 5 commits into
mainfrom
martas/1106

Conversation

@marta-sd
Copy link
Copy Markdown
Contributor

This PR solves issue #1106

@copy-pr-bot
Copy link
Copy Markdown

copy-pr-bot Bot commented May 27, 2026

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

@marta-sd
Copy link
Copy Markdown
Contributor Author

/ok to test 86b5e63

@cwing-nvidia
Copy link
Copy Markdown
Contributor

could we also include local_vllm_model_proxy in this PR or plan how to structure in the docs if it will be a follow-on PR? could you describe rationale for separate pages for each vllm variant vs a combined docs page for all vllm variants?

@marta-sd
Copy link
Copy Markdown
Contributor Author

Thanks for feedback @cwing-nvidia !

could we also include local_vllm_model_proxy in this PR or plan how to structure in the docs if it will be a follow-on PR?

I have added it to this PR, please take a look 🙂

could you describe rationale for separate pages for each vllm variant vs a combined docs page for all vllm variants?

My only motivation was that the existing vllm page was already quite long and I felt that it would be easier to navigate if they were separated. I don't see any blockers to have them all on one page, whichever you think works best for the users.

I think another options to consider is having 3 pages:

  1. overview (how Gym connects to vLLM, how model servers differ and when to choose which)
  2. self-managed vllm server (the existing page)
  3. deploy vllm server with Gym (shared page for local and local proxy)

Let me know what's your preferred structure 🙂

marta-sd added 4 commits May 29, 2026 10:14
Signed-off-by: Marta Stepniewska-Dziubinska <martas@nvidia.com>
… local_vllm_model)

Signed-off-by: Marta Stepniewska-Dziubinska <martas@nvidia.com>
…h no browser)

Signed-off-by: Marta Stepniewska-Dziubinska <martas@nvidia.com>
Signed-off-by: Marta Stepniewska-Dziubinska <martas@nvidia.com>
@marta-sd marta-sd marked this pull request as ready for review May 29, 2026 08:26
@marta-sd
Copy link
Copy Markdown
Contributor Author

/ok to test 1bd2b63

Style-guide-only cleanup stacked on top of #1430. No content or meaning
changes — this is purely an NVIDIA style-guide pass over the docs that
#1430 introduces.

## Fixes

| File | Fix |
|------|-----|
| `local-vllm.mdx` | `e.g.` → "for example" (×2); `on/off` slash → "on
or off"; `you've` → "you have" |
| `local-vllm-proxy.mdx` | typo `extising` → "existing"; `disabled in
through` → "disabled through"; `etc.` → "and so on" |
| `vllm.mdx` | `you'd` → "you would" (in the new `<Note>`) |
| `responses_api_models/vllm_model/README.md` | removed exclamation mark
|

## Scope notes

- Scoped to content **#1430 adds/changes**. Pre-existing violations in
`vllm.mdx` (the `Please…!` line, `HuggingFace` spelling, bare `[here]`
links) are out of scope for a style check of this PR.
- Left heading-case and `see → refer to` untouched — both are
inconsistent repo-wide, so changing only these pages would worsen
consistency.

Targets `martas/1106` so it merges into #1430 before that PR lands on
`main`.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Signed-off-by: Lawrence Lane <llane@nvidia.com>
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Copy link
Copy Markdown
Contributor

@cmunley1 cmunley1 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

can we give an example or link to an example of how to use this with training frameworks like nemo rl and verl?

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.

4 participants