[TRTLLM-6342][feat] Support custom sharding config source

[greg-kwasniewski1]

This PR enables user to specify a custom sharding config file, for example:

$ cat tp_sharding.yaml

head_dim : 8
tp_plan :
  gate: colwise
  up: colwise
  down: rowwise
  "*": gather

The config is expected to be a parsable .yaml or .json dictionary, with at least one required key: tp_plan, which should be a dictionary [(partial)_module_name] : [sharding_action]

The relevant tests in test_tp_sharding.py have been updated.

Summary by CodeRabbit

• New Features

  • Added flexible sharding configuration with selectable sources (heuristic, custom, factory).
  • Enabled loading custom sharding configs from JSON/YAML.
  • Expanded sharding dimensions support (including TP, EP, BMM).
  • Improved node filtering by allowing multiple predicate targets.
  • Enhanced transform aggregation for clearer sharding summaries.

• Refactor

  • Simplified configuration by replacing legacy flags with enums and consolidated fields.
  • Default behavior now favors heuristic-based sharding with partial config support.

• Tests

  • Updated unit tests to use the new sharding configuration.
  • Added scenarios validating custom YAML-based TP sharding.

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)

• Any new dependencies have been scanned for license and vulnerabilities

• CODEOWNERS updated if ownership changes

• Documentation updated as needed

• 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

/bot [-h] ['run', 'kill', 'skip', 'reuse-pipeline'] ...

Provide a user friendly way for developers to interact with a Jenkins server.

Run /bot [-h|--help] to print this help message.

See details below for each supported subcommand.

run [--reuse-test (optional)pipeline-id --disable-fail-fast --skip-test --stage-list "A10-PyTorch-1, xxx" --gpu-type "A30, H100_PCIe" --test-backend "pytorch, cpp" --add-multi-gpu-test --only-multi-gpu-test --disable-multi-gpu-test --post-merge --extra-stage "H100_PCIe-TensorRT-Post-Merge-1, xxx" --detailed-log --debug(experimental)]

Launch build/test pipelines. All previously running jobs will be killed.

--reuse-test (optional)pipeline-id (OPTIONAL) : Allow the new pipeline to reuse build artifacts and skip successful test stages from a specified pipeline or the last pipeline if no pipeline-id is indicated. If the Git commit ID has changed, this option will be always ignored. The DEFAULT behavior of the bot is to reuse build artifacts and successful test results from the last pipeline.

--disable-reuse-test (OPTIONAL) : Explicitly prevent the pipeline from reusing build artifacts and skipping successful test stages from a previous pipeline. Ensure that all builds and tests are run regardless of previous successes.

--disable-fail-fast (OPTIONAL) : Disable fail fast on build/tests/infra failures.

--skip-test (OPTIONAL) : Skip all test stages, but still run build stages, package stages and sanity check stages. Note: Does NOT update GitHub check status.

--stage-list "A10-PyTorch-1, xxx" (OPTIONAL) : Only run the specified test stages. Examples: "A10-PyTorch-1, xxx". Note: Does NOT update GitHub check status.

--gpu-type "A30, H100_PCIe" (OPTIONAL) : Only run the test stages on the specified GPU types. Examples: "A30, H100_PCIe". Note: Does NOT update GitHub check status.

--test-backend "pytorch, cpp" (OPTIONAL) : Skip test stages which don't match the specified backends. Only support [pytorch, cpp, tensorrt, triton]. Examples: "pytorch, cpp" (does not run test stages with tensorrt or triton backend). Note: Does NOT update GitHub pipeline status.

--only-multi-gpu-test (OPTIONAL) : Only run the multi-GPU tests. Note: Does NOT update GitHub check status.

--disable-multi-gpu-test (OPTIONAL) : Disable the multi-GPU tests. Note: Does NOT update GitHub check status.

--add-multi-gpu-test (OPTIONAL) : Force run the multi-GPU tests in addition to running L0 pre-merge pipeline.

--post-merge (OPTIONAL) : Run the L0 post-merge pipeline instead of the ordinary L0 pre-merge pipeline.

--extra-stage "H100_PCIe-TensorRT-Post-Merge-1, xxx" (OPTIONAL) : Run the ordinary L0 pre-merge pipeline and specified test stages. Examples: --extra-stage "H100_PCIe-TensorRT-Post-Merge-1, xxx".

--detailed-log (OPTIONAL) : Enable flushing out all logs to the Jenkins console. This will significantly increase the log volume and may slow down the job.

--debug (OPTIONAL) : Experimental feature. Enable access to the CI container for debugging purpose. Note: Specify exactly one stage in the stage-list parameter to access the appropriate container environment. Note: Does NOT update GitHub check status.

For guidance on mapping tests to stage names, see docs/source/reference/ci-overview.md
and the scripts/test_to_stage_mapping.py helper.

kill

kill

Kill all running builds associated with pull request.

skip

skip --comment COMMENT

Skip testing for latest commit on pull request. --comment "Reason for skipping build/test" is required. IMPORTANT NOTE: This is dangerous since lack of user care and validation can cause top of tree to break.

reuse-pipeline

reuse-pipeline

Reuse a previous pipeline to validate current commit. This action will also kill all currently running builds associated with the pull request. IMPORTANT NOTE: This is dangerous since lack of user care and validation can cause top of tree to break.

[greg-kwasniewski1]

\bot run

[coderabbitai]

📝 Walkthrough

Walkthrough

Refactors sharding configuration to enum-based sources and dimensions, introduces multi-source sharding apply flow (factory/custom/heuristic), adds custom config loading and TP append utilities, updates TransformInfo with += merge, removes legacy fields from AutoDeployConfig, adjusts default/config YAML, extends node filtering, and updates unit tests accordingly.

Changes

Cohort / File(s)	Summary
Config defaults tensorrt_llm/_torch/auto_deploy/config/default.yaml	Switch detect_sharding to sharding_source: ['heuristic'], enable support_partial_config: true, retain sharding_dims and requires_shape_prop, remove use_sharding_from_factory.
Public config fields removal tensorrt_llm/_torch/auto_deploy/llm_args.py	Remove simple_shard_only, use_sharding_from_factory, and sharding_dims from AutoDeployConfig.
Transform info in-place merge tensorrt_llm/_torch/auto_deploy/transform/interface.py	Add TransformInfo.__iadd__ returning a new merged instance combining flags via AND and summing num_matches.
Sharding library refactor tensorrt_llm/_torch/auto_deploy/transform/library/sharding.py	Introduce ShardingSource/ShardingDim, rework ShardingTransformConfig fields, overhaul _apply to process FACTORY/CUSTOM/HEURISTIC sources, add custom-config path, refactor factory detection and counters, update logging and shard accounting.
Sharding utilities upgrade tensorrt_llm/_torch/auto_deploy/utils/sharding_utils.py	Add enums ShardingSource, ShardingDim; extend ShardingConfig with custom_sharding_config, sharding_source, ep_transforms; add read_custom_sharding_config and append_TP.
Node utils enhancement tensorrt_llm/_torch/auto_deploy/utils/node_utils.py	Allow iterable of callables as target to match nodes if any predicate returns true.
Test helpers tests/unittest/_torch/auto_deploy/_utils_test/_graph_test_helpers.py	Improve assertion message to include actual vs. expected sets.
Tests: BMM sharding tests/unittest/_torch/auto_deploy/unit/multigpu/transformations/library/test_bmm_sharding.py	Replace use_sharding_from_factory with sharding_source: ["heuristic"]; add support_partial_config: False.
Tests: EP sharding tests/unittest/_torch/auto_deploy/unit/multigpu/transformations/library/test_ep_sharding.py	Switch to sharding_source: ["heuristic"], specify sharding_dims: ["ep"], set support_partial_config: False.
Tests: TP sharding + custom config tests/unittest/_torch/auto_deploy/unit/multigpu/transformations/library/test_tp_sharding.py	Introduce YAML-based predefined TP config; use sharding_source (["custom"] or ["heuristic"]), custom_sharding_config, sharding_dims: ["tp"]; propagate predefined config into optimizer; remove main block; add yaml import.

Sequence Diagram(s)

sequenceDiagram
  autonumber
  participant User
  participant InferenceOptimizer as InferenceOptimizer
  participant ShardingExec as ShardingTransformExecutor
  participant ShardingCfg as ShardingConfig
  participant Heuristics as HeuristicDetector
  participant Factory as FactoryConfig
  participant FS as Filesystem

  User->>InferenceOptimizer: configure detect_sharding (sharding_source, sharding_dims, custom_config)
  InferenceOptimizer->>ShardingExec: _apply(shared_config)
  ShardingExec->>ShardingCfg: read sharding_source, sharding_dims
  alt CUSTOM in sharding_source
    ShardingExec->>FS: read_custom_sharding_config(path)
    FS-->>ShardingExec: config loaded / error
    opt config loaded
      ShardingExec->>Factory: map custom to predefined_config
      Factory-->>ShardingExec: factory sharding plan
      ShardingExec->>ShardingCfg: append_TP()/record EP/BMM
    end
  end
  alt FACTORY in sharding_source
    ShardingExec->>Factory: load factory config if available
    Factory-->>ShardingExec: plan or none
    opt plan
      ShardingExec->>ShardingCfg: append_TP()/record EP/BMM
    end
  end
  alt HEURISTIC in sharding_source
    ShardingExec->>Heuristics: detect TP/EP/BMM per sharding_dims
    Heuristics-->>ShardingExec: detected shards and counts
    ShardingExec->>ShardingCfg: append_TP()/record EP/BMM
  end
  ShardingExec-->>InferenceOptimizer: TransformInfo (merged via +=)
  InferenceOptimizer-->>User: result

Loading

Estimated code review effort

🎯 4 (Complex) | ⏱️ ~60 minutes

Pre-merge checks and finishing touches

❌ Failed checks (2 warnings)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 36.84% which is insufficient. The required threshold is 80.00%.	You can run @coderabbitai generate docstrings to improve docstring coverage.
Description Check	⚠️ Warning	The PR description includes a brief free-form summary and the raw template text but lacks the required structured title and fails to populate the Description and Test Coverage sections, leaving them as empty placeholders.	Please add a title in the prescribed format at the top of the description, fill in the Description section with a concise explanation of the change and its motivation, and complete the Test Coverage section by listing the relevant tests that verify the new custom sharding configuration feature.

✅ Passed checks (1 passed)

Check name	Status	Explanation
Title Check	✅ Passed	The title succinctly and accurately describes the addition of support for a custom sharding configuration source by ticket and feature, matching the primary change of the pull request.

✨ Finishing touches

• 📝 Generate docstrings

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

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

[greg-kwasniewski1]

/bot run

[tensorrt-cicd]

PR_Github #20667 [ run ] triggered by Bot

[tensorrt-cicd]

PR_Github #20667 [ run ] completed with state FAILURE
/LLM/main/L0_MergeRequest_PR pipeline #15610 completed with status: 'FAILURE'

[greg-kwasniewski1]

/bot run

[greg-kwasniewski1]

\bot run

[tensorrt-cicd]

PR_Github #20779 [ run ] triggered by Bot

[tensorrt-cicd]

PR_Github #20779 [ run ] completed with state SUCCESS
/LLM/main/L0_MergeRequest_PR pipeline #15707 completed with status: 'FAILURE'

[greg-kwasniewski1]

/bot run

[tensorrt-cicd]

PR_Github #20797 [ run ] triggered by Bot

[tensorrt-cicd]

PR_Github #20797 [ run ] completed with state SUCCESS
/LLM/main/L0_MergeRequest_PR pipeline #15722 completed with status: 'FAILURE'