Skip to content

docs: Link HSTU AOTI example from PyTorch backend README - #201

Merged
yinggeh merged 4 commits into
mainfrom
yinggeh/tri-1616-create-hstu-support-readme-in-triton
Jul 22, 2026
Merged

docs: Link HSTU AOTI example from PyTorch backend README#201
yinggeh merged 4 commits into
mainfrom
yinggeh/tri-1616-create-hstu-support-readme-in-triton

Conversation

@yinggeh

@yinggeh yinggeh commented Jul 22, 2026

Copy link
Copy Markdown
Contributor

What does the PR do?

Adds an HSTU (Generative Recommenders) subsection at the end of the AOT Inductor Support section, pointing readers at the tutorials HSTU guide and the recsys-examples HSTU inference docs for end-to-end platform: "torch_aoti" serving.

Checklist

  • PR title reflects the change and is of format <commit_type>: <Title>
  • Changes are described in the pull request.
  • Related issues are referenced.
  • Populated github labels field
  • Added test plan and verified test passes.
  • Verified that the PR passes existing CI.
  • Verified copyright is correct on all changed files.
  • Added succinct git squash message before merging ref.
  • All template sections are filled out.
  • Optional: Additional screenshots for behavior/output changes with before/after.

Commit Type:

Check the conventional commit type
box here and add the label to the github PR.

  • build
  • ci
  • docs
  • feat
  • fix
  • perf
  • refactor
  • revert
  • style
  • test

Related PRs:

Where should the reviewer start?

  • README.md### HSTU (Generative Recommenders) under AOT Inductor Support

Test plan:

  • Confirm the new subsection sits after MODEL_INIT_LIBRARY and before PyTorch 2.0 / TorchScript

  • After the tutorials HSTU guide lands, verify both linked URLs resolve

  • CI Pipeline ID:
    N/A (docs-only)

Caveats:

  • The tutorials HSTU page should merge before or with this PR so the tutorial URL is live.

Background

HSTU generative recommenders are served on Triton via Torch AOTI. This cross-link surfaces that path from the backend README without duplicating recsys-examples runbooks.

Related Issues: (use one of the action keywords Closes / Fixes / Resolves / Relates to)

  • Relates to accompanying tutorials and server documentation PRs for HSTU / Torch AOTI

Comment thread README.md
Signed-off-by: Yingge He <yinggeh@nvidia.com>
@greptile-apps

greptile-apps Bot commented Jul 22, 2026

Copy link
Copy Markdown

Greptile Summary

This PR adds a new ### HSTU (Generative Recommenders) subsection to the ## AOT Inductor Support (Beta) section of the README, explaining that HSTU models can be served via platform: "torch_aoti" and pointing readers to an external tutorial and the recsys-examples inference docs.

  • Adds a brief description of HSTU (Hierarchical Sequential Transduction Unit) and its role in Generative Recommenders, followed by links to the tutorials guide and NVIDIA's recsys-examples docs.
  • The new subsection is placed after ### Model Initialization Hook (MODEL_INIT_LIBRARY) and before ### PyTorch 2.0 Models, keeping AOTI-specific content grouped together.

Confidence Score: 5/5

This is a documentation-only change with no code modifications; safe to merge once the linked tutorials PR is live.

The change is a single 12-line documentation addition. The prose is technically accurate, the markdown formatting is consistent with surrounding sections, and the HSTU subsection is correctly placed within the AOT Inductor section it describes.

No files require special attention.

Important Files Changed

Filename Overview
README.md Adds a 12-line HSTU subsection under AOT Inductor Support; prose is accurate, placement is appropriate, and formatting is consistent with surrounding sections.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[HSTU Generative Recommender Model] --> B[Export via torch.export]
    B --> C[Compile via torch._inductor.compile_and_package]
    C --> D[model.pt2 archive]
    D --> E[Triton Model Repository]
    E --> F["config.pbtxt: platform: 'torch_aoti'"]
    F --> G[Triton PyTorch Backend — AOT Inductor]
    G --> H[Inference Response]
Loading

Reviews (3): Last reviewed commit: "docs: Refine HSTU sequence encoder wordi..." | Re-trigger Greptile

Comment thread README.md
Comment on lines +390 to +392
[HSTU tutorial](https://github.com/triton-inference-server/tutorials/tree/main/Popular_Models_Guide/HSTU)
and the
[recsys-examples HSTU inference docs](https://github.com/NVIDIA/recsys-examples/blob/main/examples/hstu/inference/README.md).

Copy link
Copy Markdown

Choose a reason for hiding this comment

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

P2 Tutorial link may not resolve at merge time

The HSTU tutorial URL points to a path in the tutorials repo that depends on a companion PR which the author acknowledges has not yet merged. If this PR lands before that tutorials PR, readers will hit a 404 on the primary guide link. Consider merging the tutorials PR first, or gating this PR's merge on that URL becoming live.

mudit-eng
mudit-eng previously approved these changes Jul 22, 2026
Signed-off-by: Yingge He <yinggeh@nvidia.com>
Signed-off-by: Yingge He <yinggeh@nvidia.com>
@yinggeh
yinggeh requested a review from mudit-eng July 22, 2026 19:28
@yinggeh
yinggeh merged commit e76eb86 into main Jul 22, 2026
2 checks passed
@yinggeh
yinggeh deleted the yinggeh/tri-1616-create-hstu-support-readme-in-triton branch July 22, 2026 21:58
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Development

Successfully merging this pull request may close these issues.

2 participants