[TRTLLM-7073][feat] Support torch compile for PP for Llama and DeepSeekV3

[liji-nv]

…ekV3

Summary by CodeRabbit

• New Features
  • Improved multi-GPU pipeline-parallel support with torch.compile, including backend awareness of distributed mappings for better fusions.
  • Models may now return (hidden_states, residual) to preserve necessary tensors across pipeline stages.
• Bug Fixes
  • More reliable input token detection in compiled graphs.
  • Correct handling of pipeline send/recv as in-place ops to prevent elimination during compilation.
  • Userbuffers initialization now gated to valid configurations.
• Tests
  • Expanded multi-GPU torch.compile test coverage by removing previous skips.

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.

• [x] 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.

[liji-nv]

/bot run

[tensorrt-cicd]

PR_Github #19186 [ run ] triggered by Bot

[coderabbitai[bot]]

📝 Walkthrough

Walkthrough

The change threads a Mapping object through the compilation backend and AR fusion registrations, registers pp_send/pp_recv as Torch custom ops, updates model forward returns to optionally include residuals, adapts speculative wrappers to handle tuple outputs, extends inplace op metadata, adjusts userbuffers enablement, and broadens tests to run under torch.compile with multi-GPU.

Changes

Cohort / File(s)	Summary
Backend mapping threading tensorrt_llm/_torch/compilation/backend.py	Adds mapping parameter to Backend.init and passes it into get_custom_pass; get_custom_pass signature updated to accept mapping and propagate to AR fusion registration; call detects l_position_ids_ for input token counting; minor debug hook comment.
AR fusion registration uses external Mapping tensorrt_llm/_torch/compilation/patterns/ar_residual_norm.py	All registration functions now require a Mapping argument; removes internal Mapping creation; extends allreduce.default with explicit strategy/workspace via mapping.tp_group; register_ar_fusions and register_ub_patterns accept and pass mapping.
Custom ops for pipeline send/recv + inplace info tensorrt_llm/_torch/distributed/communicator.py, tensorrt_llm/_torch/compilation/utils.py	Registers pp_recv and pp_send as Torch custom ops (trtllm::pp_recv/pp_send) mutating “tensor”; adds inplace mappings for these ops in inplace_info.
Model forward returns residuals tensorrt_llm/_torch/models/modeling_deepseekv3.py, tensorrt_llm/_torch/models/modeling_llama.py, tensorrt_llm/_torch/models/modeling_utils.py, tensorrt_llm/_torch/models/modeling_speculative.py	DeepseekV3Model and LlamaModel forward now return (hidden_states, residual); type hints updated accordingly; DecoderModel.forward return type broadened to Union[tensor, tuple]; speculative wrappers unwrap first element when model returns a tuple.
Executor wiring and userbuffers gating tensorrt_llm/_torch/pyexecutor/model_engine.py	Passes mapping into Backend constructor; enables userbuffers only when tp_size > 1 and pp_size <= 1.
Unit tests: TP mapping helper and backend wiring tests/unittest/_torch/multi_gpu/test_user_buffers.py, tests/unittest/_torch/multi_gpu/test_ar_residual_norm.py	Introduces create_tp_mapping helper; updates Backend and model inits to use mapping; adjusts a ub.initialize_userbuffers_manager call; ar_residual_norm test passes per-rank mapping to Backend.
Integration tests: remove skips under torch.compile on multi-GPU tests/integration/defs/accuracy/test_llm_api_pytorch.py	Removes conditional skips to allow tests to run with multi-GPU/pipeline-parallel configurations.

Sequence Diagram(s)

sequenceDiagram
  autonumber
  participant Engine as ModelEngine
  participant Backend as Backend
  participant Pass as PatternMatcherPass(es)
  participant Patterns as AR Residual-Norm Patterns
  participant Map as Mapping

  Engine->>Backend: __init__(..., mapping=Map)
  Backend->>Backend: get_custom_pass(enable_userbuffers, Map)
  Backend->>Pass: create pass list
  Backend->>Patterns: register_ar_fusions(Pass[], Map, enable_UB)
  Patterns-->>Pass: register residual-norm + quant patterns
  note right of Patterns: allreduce registered with mapping.tp_group

sequenceDiagram
  autonumber
  participant PP0 as Pipeline Stage 0
  participant PP1 as Pipeline Stage 1
  participant Comm as trtllm::pp_send/pp_recv (custom ops)
  participant Spec as Speculative Wrapper

  PP0->>PP0: forward() -> (hidden_states, residual)
  PP0->>Comm: pp_send(tensor=residual) [mutates tensor]
  PP0-->>PP1: pipeline boundary

  PP1->>Comm: pp_recv(tensor=residual) [mutates tensor]
  PP1->>PP1: forward(..., residual)
  Spec->>PP1: outputs = model(...)
  alt outputs is tuple
    Spec->>Spec: hidden_states = outputs[0]
  else
    Spec->>Spec: hidden_states = outputs
  end

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 11.11% 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 lacks substantive content. Only the template structure remains with blank sections for Description and Test Coverage, despite the changes affecting multiple critical files.	Fill in the Description section explaining what changes were made and why. Document the test coverage for the changes, including relevant test files like test_ar_residual_norm.py and test_user_buffers.py.

✅ Passed checks (1 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly summarizes the main change: adding support for torch compile with pipeline parallelism (PP) for Llama and DeepSeekV3 models.

✨ Finishing touches

• [ ] 📝 Generate docstrings

🧪 Generate unit tests (beta)

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

[!TIP]

📝 Customizable high-level summaries are now available in beta!

You can now customize how CodeRabbit generates the high-level summary in your pull requests — including its content, structure, tone, and formatting.

• Provide your own instructions using the high_level_summary_instructions setting.
• Format the summary however you like (bullet lists, tables, multi-section layouts, contributor stats, etc.).
• Use high_level_summary_in_walkthrough to move the summary from the description to the walkthrough section.

Example instruction:

"Divide the high-level summary into five sections:

1. 📝 Description — Summarize the main change in 50–60 words, explaining what was done.
2. 📓 References — List relevant issues, discussions, documentation, or related PRs.
3. 📦 Dependencies & Requirements — Mention any new/updated dependencies, environment variable changes, or configuration updates.
4. 📊 Contributor Summary — Include a Markdown table showing contributions: | Contributor | Lines Added | Lines Removed | Files Changed |
5. ✔️ Additional Notes — Add any extra reviewer context. Keep each section concise (under 200 words) and use bullet or numbered lists for clarity."

Note: This feature is currently in beta for Pro-tier users, and pricing will be announced later.

---

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.}

[tensorrt-cicd]

PR_Github #19186 [ run ] completed with state SUCCESS /LLM/main/L0_MergeRequest_PR pipeline #14400 completed with status: 'FAILURE'

[liji-nv]

/bot run

[tensorrt-cicd]

PR_Github #19693 [ run ] triggered by Bot

[tensorrt-cicd]

PR_Github #19693 [ run ] completed with state FAILURE /LLM/main/L0_MergeRequest_PR pipeline #14821 completed with status: 'FAILURE'

[liji-nv]

/bot run

[tensorrt-cicd]

PR_Github #19800 [ run ] triggered by Bot

[tensorrt-cicd]

PR_Github #19800 [ run ] completed with state FAILURE /LLM/main/L0_MergeRequest_PR pipeline #14896 completed with status: 'FAILURE'

[liji-nv]

/bot run

[tensorrt-cicd]

PR_Github #19881 [ run ] triggered by Bot

[tensorrt-cicd]

PR_Github #19881 [ run ] completed with state SUCCESS /LLM/main/L0_MergeRequest_PR pipeline #14962 completed with status: 'FAILURE'

[liji-nv]

/bot run

[tensorrt-cicd]

PR_Github #20085 [ run ] triggered by Bot

[tensorrt-cicd]

PR_Github #20085 [ run ] completed with state SUCCESS /LLM/main/L0_MergeRequest_PR pipeline #15133 completed with status: 'FAILURE'

[liji-nv]

/bot run

[tensorrt-cicd]

PR_Github #20352 [ run ] triggered by Bot

[tensorrt-cicd]

PR_Github #20352 [ run ] completed with state SUCCESS /LLM/main/L0_MergeRequest_PR pipeline #15354 completed with status: 'FAILURE'

[liji-nv]

/bot run

[liji-nv]

/bot run

[tensorrt-cicd]

PR_Github #25694 [ run ] triggered by Bot. Commit: 79143be

[tensorrt-cicd]

PR_Github #25694 [ run ] completed with state SUCCESS. Commit: 79143be /LLM/main/L0_MergeRequest_PR pipeline #19476 completed with status: 'FAILURE'

[liji-nv]

/bot run

[tensorrt-cicd]

PR_Github #26398 [ run ] triggered by Bot. Commit: 8006410

[tensorrt-cicd]

PR_Github #26398 [ run ] completed with state SUCCESS. Commit: 8006410 /LLM/main/L0_MergeRequest_PR pipeline #20056 completed with status: 'FAILURE'

[liji-nv]

/bot run

[tensorrt-cicd]

PR_Github #26547 [ run ] triggered by Bot. Commit: 8006410

[tensorrt-cicd]

PR_Github #26547 [ run ] completed with state SUCCESS. Commit: 8006410 /LLM/main/L0_MergeRequest_PR pipeline #20189 completed with status: 'FAILURE'

[liji-nv]

/bot run