[https://nvbugs/5517023][fix] Pass allreduce strategy and force NCCL on pre-Blackwell arch

[hyukn]

Summary by CodeRabbit

• New Features

  • Runtime-selectable all-reduce strategy: reduction method now adapts to GPU capability, with a reliable fallback on older architectures.

• Bug Fixes

  • Improved stability and performance on older GPUs by choosing an appropriate reduction strategy automatically.
  • Existing tensor mapping behavior preserved to avoid changes in distribution patterns.

Description

Pass allreduce strategy from the model config and force NCCL on pre-Blackwell arch due to perf regression observed on A100/H100.

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.

[coderabbitai]

📝 Walkthrough

Walkthrough

Add AllReduceStrategy export and pass a strategy argument when constructing AllReduce in Llama4DecoderLayer.__init__; the strategy is selected at runtime depending on get_sm_version() (fallback to AllReduceStrategy.NCCL for older SM versions). A TODO comment notes this is a temporary perf workaround.

Changes

Cohort / File(s)	Change summary
Llama decoder comms tensorrt_llm/_torch/models/modeling_llama.py	- Import AllReduceStrategy from tensorrt_llm._torch.distributed - AllReduce now constructed with strategy=... and mapping=... (strategy chosen as model_config.allreduce_strategy if get_sm_version() >= 100 else AllReduceStrategy.NCCL) - Added TODO comment noting temporary fix for pre-Blackwell perf regressions
Public API exports tensorrt_llm/_torch/distributed/...	- AllReduceStrategy added to public exports (constructor signature of AllReduce now accepts strategy)

Sequence Diagram(s)

sequenceDiagram
  autonumber
  participant App as App
  participant L4 as Llama4DecoderLayer.__init__
  participant SM as GPU SM (get_sm_version)
  participant AR as AllReduce

  App->>L4: __init__(model_config)
  L4->>SM: get_sm_version()
  SM-->>L4: sm_version
  alt sm_version >= 100
    L4->>AR: new AllReduce(strategy=model_config.allreduce_strategy, mapping=model_config.mapping)
  else sm_version < 100
    L4->>AR: new AllReduce(strategy=AllReduceStrategy.NCCL, mapping=model_config.mapping)
  end
  AR-->>L4: AllReduce instance

Loading

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

✨ Finishing touches

• 📝 Generate Docstrings

🧪 Generate unit tests

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

---

📜 Recent review details

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between 580a733 and b666707.

📒 Files selected for processing (1)

• tensorrt_llm/_torch/models/modeling_llama.py (2 hunks)

🚧 Files skipped from review as they are similar to previous changes (1)

• tensorrt_llm/_torch/models/modeling_llama.py

⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (1)

• GitHub Check: Pre-commit Check

Tip

👮 Agentic pre-merge checks are now available in preview!

Pro plan users can now enable pre-merge checks in their settings to enforce checklists before merging PRs.

• Built-in checks – Quickly apply ready-made checks to enforce title conventions, require pull request descriptions that follow templates, validate linked issues for compliance, and more.
• Custom agentic checks – Define your own rules using CodeRabbit’s advanced agentic capabilities to enforce organization-specific policies and workflows. For example, you can instruct CodeRabbit’s agent to verify that API documentation is updated whenever API schema files are modified in a PR. Note: Upto 5 custom checks are currently allowed during the preview period. Pricing for this feature will be announced in a few weeks.

Please see the documentation for more information.

Example:

reviews:
  pre_merge_checks:
    custom_checks:
      - name: "Undocumented Breaking Changes"
        mode: "warning"
        instructions: |
          Pass/fail criteria: All breaking changes to public APIs, CLI flags, environment variables, configuration keys, database schemas, or HTTP/GraphQL endpoints must be documented in the "Breaking Change" section of the PR description and in CHANGELOG.md. Exclude purely internal or private changes (e.g., code not exported from package entry points or explicitly marked as internal).

Please share your feedback with us on this Discord post.

---

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.}

Pre-merge checks

❌ Failed checks (2 warnings)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% 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 is minimal and does not follow the repository template: it is missing the required templated PR title line, the Test Coverage section is empty, and the PR Checklist items are not completed. The Description is a single sentence and omits important reviewer-facing details such as measured perf data, exact SM/GPU targets, and the public API change (AllReduce constructor signature and added AllReduceStrategy). Because of these omissions, reviewers cannot fully assess impact, test coverage, or migration needs.	Add the required templated PR title at the top and expand the Description to explain what changed, why, and include concrete perf/regression evidence and exact target architectures/SM versions; explicitly document the public API change (AllReduce signature now accepts strategy and AllReduceStrategy was added) and any migration or compatibility implications. Populate the Test Coverage section with the unit/integration tests or CI stages that validate runtime strategy selection and AR-fusion disablement and include links to relevant CI/perf runs. Complete the PR Checklist and, if applicable, update docs/CHANGELOG/CODEOWNERS so reviewers can verify functional and non-functional impacts.

✅ Passed checks (1 passed)

Check name	Status	Explanation
Title Check	✅ Passed	The title succinctly and accurately captures the primary change: passing the allreduce strategy and forcing an NCCL fallback on pre-Blackwell architectures. It is a single, clear sentence that aligns with the PR objectives and the raw_summary (updating the AllReduce call site and choosing NCCL for older SMs). The included bug reference and [fix] tag act as acceptable metadata and do not obscure the intent.

[hyukn]

/bot run

[tensorrt-cicd]

PR_Github #18788 [ run ] triggered by Bot

[coderabbitai]

Actionable comments posted: 1

Caution

Some comments are outside the diff and can’t be posted inline due to platform limitations.

⚠️ Outside diff range comments (1)

tensorrt_llm/_torch/models/modeling_llama.py (1)

1-3: Missing required NVIDIA Apache‑2.0 header.

Per repo guidelines, prepend the current-year header to this file.

Apply:

+# Copyright (c) 2025, NVIDIA CORPORATION. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.

📜 Review details

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between 14aa34f and 580a733.

📒 Files selected for processing (1)

• tensorrt_llm/_torch/models/modeling_llama.py (2 hunks)

🧰 Additional context used

📓 Path-based instructions (3)

**/*.{h,hpp,hh,hxx,cpp,cxx,cc,cu,cuh,py}

📄 CodeRabbit inference engine (CODING_GUIDELINES.md)

Use only spaces, no tabs; indent with 4 spaces.

Files:

• tensorrt_llm/_torch/models/modeling_llama.py

**/*.py

📄 CodeRabbit inference engine (CODING_GUIDELINES.md)

**/*.py: Python code must target Python 3.8+.
Indent Python code with 4 spaces; do not use tabs.
Maintain module namespace when importing; prefer 'from package.subpackage import foo' then 'foo.SomeClass()' instead of importing the class directly.
Python filenames should be snake_case (e.g., some_file.py).
Python classes use PascalCase names.
Functions and methods use snake_case names.
Local variables use snake_case; prefix 'k' for variables that start with a number (e.g., k_99th_percentile).
Global variables use upper SNAKE_CASE prefixed with 'G' (e.g., G_MY_GLOBAL).
Constants use upper SNAKE_CASE (e.g., MY_CONSTANT).
Avoid shadowing variables from an outer scope.
Initialize all externally visible members of a class in the constructor.
Prefer docstrings for interfaces that may be used outside a file; comments for in-function or file-local interfaces.
Use Google-style docstrings for classes and functions (Sphinx-parsable).
Document attributes and variables inline so they render under the class/function docstring.
Avoid reflection when a simpler, explicit approach suffices (e.g., avoid dict(**locals()) patterns).
In try/except, catch the most specific exceptions possible.
For duck-typing try/except, keep the try body minimal and use else for the main logic.

Files:

• tensorrt_llm/_torch/models/modeling_llama.py

**/*.{cpp,cxx,cc,h,hpp,hh,hxx,cu,cuh,py}

📄 CodeRabbit inference engine (CODING_GUIDELINES.md)

Prepend the NVIDIA Apache-2.0 copyright header with current year to the top of all source files (e.g., .cpp, .h, .cu, .py).

Files:

• tensorrt_llm/_torch/models/modeling_llama.py

🧠 Learnings (1)

📚 Learning: 2025-08-14T06:36:40.701Z

Learnt from: timlee0212
PR: NVIDIA/TensorRT-LLM#6886
File: tensorrt_llm/_torch/models/modeling_deepseekv3.py:0-0
Timestamp: 2025-08-14T06:36:40.701Z
Learning: In DeepSeek V3 model (tensorrt_llm/_torch/models/modeling_deepseekv3.py), the disagreement between AllReduce.__init__ guard and _compute_mlp_tp_size logic for MNNVL usage is expected by design. The AllReduce component and MLP TP-size computation intentionally use different criteria for MNNVL availability decisions.

Applied to files:

• tensorrt_llm/_torch/models/modeling_llama.py

🧬 Code graph analysis (1)

tensorrt_llm/_torch/models/modeling_llama.py (2)

tensorrt_llm/_torch/distributed/ops.py (1)

• AllReduce (362-491)

tensorrt_llm/_utils.py (1)

• get_sm_version (681-683)

⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (1)

• GitHub Check: Pre-commit Check

🔇 Additional comments (1)

tensorrt_llm/_torch/models/modeling_llama.py (1)

649-652: Require explicit AllReduce.strategy or a safe ModelConfig default

Script output shows AllReduce() invocations without a strategy kwarg at:

• ./tensorrt_llm/_torch/distributed/ops.py:287
• ./tensorrt_llm/_torch/distributed/ops.py:362
• ./tensorrt_llm/_torch/distributed/ops.py:424
• ./tensorrt_llm/_torch/distributed/ops.py:494
• ./tensorrt_llm/_torch/models/modeling_deepseekv3.py:688
• ./tensorrt_llm/_torch/models/modeling_llama.py:454
• ./tensorrt_llm/_torch/models/modeling_qwen3_moe.py:206
• ./tests/unittest/_torch/auto_deploy/unit/multigpu/transformations/library/test_collective_fusion.py:20
• ./tests/unittest/_torch/multi_gpu/test_allreduce.py:129
• ./tests/unittest/_torch/multi_gpu/test_allreduce.py:340
• ./tests/unittest/_torch/multi_gpu/test_allreduce.py:502
• ./tests/unittest/_torch/multi_gpu/test_user_buffers.py:695
• ./tests/unittest/_torch/multi_gpu/test_user_buffers.py:696
• ./tests/unittest/_torch/multi_gpu/test_user_buffers.py:697

Action: update these call sites to pass strategy (e.g., strategy=model_config.allreduce_strategy or an explicit AllReduceStrategy like AUTO), or ensure ModelConfig supplies that safe default.

⛔ Skipped due to learnings

Learnt from: timlee0212
PR: NVIDIA/TensorRT-LLM#6886
File: tensorrt_llm/_torch/models/modeling_deepseekv3.py:0-0
Timestamp: 2025-08-14T06:36:40.701Z
Learning: In DeepSeek V3 model (tensorrt_llm/_torch/models/modeling_deepseekv3.py), the disagreement between AllReduce.__init__ guard and _compute_mlp_tp_size logic for MNNVL usage is expected by design. The AllReduce component and MLP TP-size computation intentionally use different criteria for MNNVL availability decisions.

[hyukn]

/bot run

[tensorrt-cicd]

PR_Github #18794 [ run ] triggered by Bot

[tensorrt-cicd]

PR_Github #18788 [ run ] completed with state ABORTED
LLM/release-1.0/L0_MergeRequest_PR #396 (Blue Ocean) completed with status: ABORTED

[tensorrt-cicd]

PR_Github #18794 [ run ] completed with state SUCCESS
/LLM/release-1.0/L0_MergeRequest_PR pipeline #399 completed with status: 'FAILURE'

[hyukn]

/bot run

[tensorrt-cicd]

PR_Github #18848 [ run ] triggered by Bot

[tensorrt-cicd]

PR_Github #18848 [ run ] completed with state SUCCESS
/LLM/release-1.0/L0_MergeRequest_PR pipeline #403 completed with status: 'SUCCESS'