[None][doc] Add docs for AutoDeploy transforms

[bmarimuthu-nv]

Summary by CodeRabbit

• Documentation
  • Added comprehensive AutoDeploy transforms documentation including descriptions of all pipeline stages (factory, export, post-export, pattern matching, sharding, weight loading, post-load fusion, cache initialization, visualization, and compilation) with auto-generated API references.

Description

Test Coverage

PR Checklist

Please review the following before submitting your PR:

• PR description clearly explains what and why. If using CodeRabbit's summary, please make sure it makes sense.

• PR Follows TRT-LLM CODING GUIDELINES to the best of your knowledge.

• Test cases are provided for new code paths (see test instructions)

• If PR introduces API changes, an appropriate PR label is added - either api-compatible or api-breaking. For api-breaking, include BREAKING in the PR title.

• Any new dependencies have been scanned for license and vulnerabilities

• CODEOWNERS updated if ownership changes

• Documentation updated as needed

• Update tava architecture diagram if there is a significant design change in PR.

• The reviewers assigned automatically/manually are appropriate for the PR.

• Please check this after reviewing the above items as appropriate for this PR.

GitHub Bot Help

To see a list of available CI bot commands, please comment /bot help.

[bmarimuthu-nv]

@CodeRabbit summary

[coderabbitai]

✅ Action performed

Summary regeneration triggered.

[bmarimuthu-nv]

Example docs: http://cw-dfw-cs-001-vscode-01:14465/transforms.html

[coderabbitai]

📝 Walkthrough

Walkthrough

Adds a Sphinx extension that discovers AutoDeploy transforms from source and YAML configs and generates transform-stage documentation; integrates the extension into Sphinx and adds a landing page plus per-stage RST pages that invoke the new directives.

Changes

AutoDeploy Transform Documentation with Sphinx Extension

Layer / File(s)	Summary
Sphinx Extension for AutoDeploy Transforms docs/source/_ext/trtllm_auto_deploy.py	Implements a 470-line Sphinx extension that discovers transform modules by filesystem scanning (without imports), parses Python source with AST to extract TransformRegistry.register() decorators and get_config_class uses, resolves config-class inheritance with cycle detection, loads and validates YAML stage/mode mappings from checked-in AutoDeploy configs, and registers two directives: trtllm_auto_deploy_transform_stage and trtllm_auto_deploy_additional_transforms to render stage-specific and unassigned transforms.
Documentation Infrastructure Setup docs/source/conf.py, .gitignore	Registers the trtllm_auto_deploy extension in Sphinx configuration and updates .gitignore to un-ignore transforms.rst while ignoring the transforms/ directory and its *.rst files.
Transform Stage Documentation Pages docs/source/features/auto_deploy/transforms.rst, .../transforms/core.rst, .../transforms/export.rst, .../transforms/factory.rst, .../transforms/pattern_matcher.rst, .../transforms/post_export.rst, .../transforms/sharding.rst, .../transforms/compile.rst, .../transforms/cache_init.rst, .../transforms/weight_load.rst, .../transforms/post_load_fusion.rst, .../transforms/visualize.rst, .../transforms/additional.rst	Adds a landing page and per-stage RST pages that invoke the new directives to render discovered transform classes and their resolved config models, and documents core transform APIs via autopydantic_model/automodule directives.
Documentation Navigation Integration docs/source/features/auto_deploy/auto-deploy.md, docs/source/index.rst	Adds an "API Reference" link to AutoDeploy Transforms in the AutoDeploy beta docs and inserts features/auto_deploy/transforms.rst into the main documentation toctree.

🎯 3 (Moderate) | ⏱️ ~20 minutes

🚥 Pre-merge checks | ✅ 3 | ❌ 2

❌ Failed checks (2 warnings)

Check name	Status	Explanation	Resolution
Description check	⚠️ Warning	The PR description is incomplete. It only contains the repository template with no substantive content in required sections like Description, Test Coverage, or implementation details.	Fill in the Description section explaining what documentation was added and why. Add Test Coverage section documenting build verification steps. Complete the PR checklist comments to explain design decisions and dependency checks.
Docstring Coverage	⚠️ Warning	Docstring coverage is 30.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (3 passed)

Check name	Status	Explanation
Title check	✅ Passed	The PR title '[None][doc] Add docs for AutoDeploy transforms' is directly related to the main change—comprehensive documentation addition for AutoDeploy transforms.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[govind-ramnarayan]

LGTM, docs look clean. I guess one question is how we will keep this updated - if we add a new transform, will we get a KeyError or something in this trtllm_auto_deploy script? Does it depend on whether it is in an existing stage or in a new stage?

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@docs/source/_ext/trtllm_auto_deploy.py`:
- Around line 1-2: Replace the current two-line SPDX-only header at the top of
docs/source/_ext/trtllm_auto_deploy.py with the repository's full NVIDIA
copyright/license header block required for Python files (i.e., the standard
multi-line NVIDIA copyright header that includes the company copyright line,
years, and the SPDX-License-Identifier), ensuring the header appears before any
imports or code in the trtllm_auto_deploy module.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 852038e4-5964-44eb-a18b-90b650271ab2

📥 Commits

Reviewing files that changed from the base of the PR and between 9827c21 and f56aaa0.

📒 Files selected for processing (18)

• .gitignore
• docs/source/_ext/trtllm_auto_deploy.py
• docs/source/conf.py
• docs/source/features/auto_deploy/auto-deploy.md
• docs/source/features/auto_deploy/transforms.rst
• docs/source/features/auto_deploy/transforms/additional.rst
• docs/source/features/auto_deploy/transforms/cache_init.rst
• docs/source/features/auto_deploy/transforms/compile.rst
• docs/source/features/auto_deploy/transforms/core.rst
• docs/source/features/auto_deploy/transforms/export.rst
• docs/source/features/auto_deploy/transforms/factory.rst
• docs/source/features/auto_deploy/transforms/pattern_matcher.rst
• docs/source/features/auto_deploy/transforms/post_export.rst
• docs/source/features/auto_deploy/transforms/post_load_fusion.rst
• docs/source/features/auto_deploy/transforms/sharding.rst
• docs/source/features/auto_deploy/transforms/visualize.rst
• docs/source/features/auto_deploy/transforms/weight_load.rst
• docs/source/index.rst

[bmarimuthu-nv]

LGTM, docs look clean. I guess one question is how we will keep this updated - if we add a new transform, will we get a KeyError or something in this trtllm_auto_deploy script? Does it depend on whether it is in an existing stage or in a new stage?

For a new transform in an existing configured stage the doc generation is automatic. But a new stage needs a docs page/toctree entry for it to show up as a stage page.

[tburt-nv]

/bot run --stages "A10-Build_Docs"

[tensorrt-cicd]

PR_Github #52885 Bot args parsing error: usage: /bot [-h]
{run,kill,skip,submit,reviewers,reuse-pipeline,reuse-review} ...
/bot: error: unrecognized arguments: --stages A10-Build_Docs

Link to invocation

[govind-ramnarayan]

LGTM, docs look clean. I guess one question is how we will keep this updated - if we add a new transform, will we get a KeyError or something in this trtllm_auto_deploy script? Does it depend on whether it is in an existing stage or in a new stage?

For a new transform in an existing configured stage the doc generation is automatic. But a new stage needs a docs page/toctree entry for it to show up as a stage page.

Hmm, is there any way to write some sort of validation or test that would tell us if a stage was missing in the docs? Not blocking the merging of this PR because I think it will likely be obvious and I don't know if adding a stage will be common, but it could be nice.

[bmarimuthu-nv]

is there any way to write some sort of validation or test that would tell us if a stage was missing in the docs?

I don't know if there's something that checks for docs against code, but that could be nice. But yeah, doc generation for new transforms to existing stages is automated which is the common case.

[bmarimuthu-nv]

/bot run --extra-stage "A10-Build_Docs"

[tensorrt-cicd]

PR_Github #52904 [ run ] triggered by Bot. Commit: f56aaa0 Link to invocation

[tensorrt-cicd]

PR_Github #52904 [ run ] completed with state SUCCESS. Commit: f56aaa0
/LLM/main/L0_MergeRequest_PR pipeline #42151 completed with status: 'SUCCESS'
Pipeline passed with automatic retried tests. Check the rerun report for details.

CI Report

Link to invocation