[https://nvbugs/5494718][fix] Fix Single GPU Multi-node issue and OOM on DGX Spark

[SimengLiu-nv]

… with DGX Spark.

Summary by CodeRabbit

• Bug Fixes

  • Optimized memory transfers to prevent unnecessary device transfers on integrated GPU systems, reducing potential memory overhead.

• Performance

  • Added automatic integrated GPU detection to optimize device handling and improve overall system performance on compatible hardware.

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.

Details

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.

[SimengLiu-nv]

/bot run --disable-fail-fast

[coderabbitai]

📝 Walkthrough

Walkthrough

The changes introduce integrated GPU detection to optimize memory usage during weight transfers. A new is_device_integrated() function is added to detect and cache whether the GPU is integrated, and the linear module uses this to conditionally skip device transfers. Additionally, is_multi_device_enable() is removed and replaced with a direct check against global_mpi_size().

Changes

Cohort / File(s)	Summary
Integrated GPU Detection tensorrt_llm/_utils.py	Added _device_integrated cache variable and is_device_integrated() public function to detect and cache integrated GPU status. Function returns False if CUDA unavailable; on first call checks torch.cuda.is_available() and reads device properties.
Linear Module Transfer Guard tensorrt_llm/_torch/modules/linear.py	Added guard in load_weight_shard to skip device transfers when using integrated GPUs. Imports is_device_integrated and logger. When device types differ and is_device_integrated() is true, logs warning and uses weight's current device instead of transferring.
Multi-Device Check Replacement tensorrt_llm/_utils.py	Removed is_multi_device_enable() function. Replaced its usage in mpi_broadcast with direct check: broadcast occurs when global_mpi_size() > 1.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~12 minutes

Pre-merge checks and finishing touches

❌ Failed checks (2 warnings)

Check name	Status	Explanation	Resolution
Description Check	⚠️ Warning	The pull request description is largely incomplete. The author provided only a fragment ("… with DGX Spark.") instead of a substantive description, and both required sections—Description and Test Coverage—are entirely empty with only template comments. While one checkbox in the PR Checklist is marked as completed, the core sections that should explain what is being changed and why, along with relevant test coverage information, are absent. This significantly lacks the information needed to understand the PR's rationale and testing approach.
Docstring Coverage	⚠️ Warning	Docstring coverage is 12.50% which is insufficient. The required threshold is 80.00%.	You can run @coderabbitai generate docstrings to improve docstring coverage.

✅ Passed checks (1 passed)

Check name	Status	Explanation
Title Check	✅ Passed	The pull request title follows the correct format from the template, using the NVBugs ID [https://nvbugs/5494718], valid type [fix], and providing a specific summary referencing the DGX Spark environment and issues being addressed. The title is related to the changeset—the modifications add integrated GPU detection and optimize device transfers to address OOM issues on integrated GPU systems like DGX Spark. While the title emphasizes the problems being solved rather than the technical solution, it provides sufficient context to understand the fix is relevant to integrated GPU environments, which aligns with the actual code changes.

✨ 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.}

[coderabbitai]

Actionable comments posted: 2

Caution

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

⚠️ Outside diff range comments (1)

tensorrt_llm/_utils.py (1)

1-1: Update copyright year to 2025.

Per coding guidelines, modified source files should have the copyright header updated with the current year.

As per coding guidelines

Apply this diff to update the copyright year:

-# SPDX-FileCopyrightText: Copyright (c) 2022-2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-FileCopyrightText: Copyright (c) 2022-2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.

📜 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 984d4fe and 6d9422c.

📒 Files selected for processing (2)

• tensorrt_llm/_torch/modules/linear.py (2 hunks)
• tensorrt_llm/_utils.py (3 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/modules/linear.py
• tensorrt_llm/_utils.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/modules/linear.py
• tensorrt_llm/_utils.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/modules/linear.py
• tensorrt_llm/_utils.py

🧬 Code graph analysis (2)

tensorrt_llm/_torch/modules/linear.py (2)

tensorrt_llm/_utils.py (1)

• is_device_integrated (1205-1222)

tensorrt_llm/logger.py (1)

• warning (132-133)

tensorrt_llm/_utils.py (1)

cpp/tensorrt_llm/runtime/utils/mpiUtils.cpp (4)

• bcast (287-295)
• bcast (287-287)
• bcast (297-301)
• bcast (297-297)

⏰ 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 (3)

tensorrt_llm/_torch/modules/linear.py (1)

17-17: LGTM!

The new imports support the integrated GPU detection logic introduced in this PR.

Also applies to: 20-20

tensorrt_llm/_utils.py (2)

53-55: LGTM!

The module-level cache variable follows Python naming conventions and has a clear explanatory comment.

---

585-585: LGTM!

Replacing is_multi_device_enable() with global_mpi_size() > 1 provides a more precise check that avoids unnecessary broadcast operations when running with a single process.

[tensorrt-cicd]

PR_Github #21958 [ run ] triggered by Bot. Commit: 6d9422c

[tensorrt-cicd]

PR_Github #21958 [ run ] completed with state SUCCESS. Commit: 6d9422c
/LLM/main/L0_MergeRequest_PR pipeline #16555 completed with status: 'FAILURE'

[SimengLiu-nv]

PR_Github #21958 [ run ] completed with state SUCCESS. Commit: 6d9422c /LLM/main/L0_MergeRequest_PR pipeline #16555 completed with status: 'FAILURE'

Infra failure: E: Failed to fetch https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/x86_64/Packages.gz File has unexpected size (959242 != 963109).

[SimengLiu-nv]

/bot run --disable-fail-fast

[tensorrt-cicd]

PR_Github #22071 [ run ] triggered by Bot. Commit: f6902c4

[tensorrt-cicd]

PR_Github #22071 [ run ] completed with state SUCCESS. Commit: f6902c4
/LLM/main/L0_MergeRequest_PR pipeline #16643 completed with status: 'FAILURE'

[SimengLiu-nv]

/bot run

[tensorrt-cicd]

PR_Github #22096 [ run ] triggered by Bot. Commit: 36d66ac

[tensorrt-cicd]

PR_Github #22096 [ run ] completed with state SUCCESS. Commit: 36d66ac
/LLM/main/L0_MergeRequest_PR pipeline #16660 completed with status: 'FAILURE'

[SimengLiu-nv]

PR_Github #22096 [ run ] completed with state SUCCESS. Commit: 36d66ac /LLM/main/L0_MergeRequest_PR pipeline #16660 completed with status: 'FAILURE'

Hit new test failure unrelated to the change, filed ttps://nvbugs/5601909 and waive it from the new commit. Relaunch a new CI.

[SimengLiu-nv]

/bot run

[tensorrt-cicd]

PR_Github #22124 [ run ] triggered by Bot. Commit: d06bc34

[tensorrt-cicd]

PR_Github #22124 [ run ] completed with state SUCCESS. Commit: d06bc34
/LLM/main/L0_MergeRequest_PR pipeline #16682 completed with status: 'FAILURE'

[SimengLiu-nv]

PR_Github #22124 [ run ] completed with state SUCCESS. Commit: d06bc34 /LLM/main/L0_MergeRequest_PR pipeline #16682 completed with status: 'FAILURE'

Apt installation failed.
[2025-10-22T06:27:23.802Z] Err:103 https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/sbsa libcusparse-dev-12-9 12.5.10.65-1

[2025-10-22T06:27:23.802Z] Connection timed out [IP: 23.221.22.207 443]

[SimengLiu-nv]

/bot run

[tensorrt-cicd]

PR_Github #22203 [ run ] triggered by Bot. Commit: d06bc34

[tensorrt-cicd]

PR_Github #22332 [ run ] triggered by Bot. Commit: e841c17

[tensorrt-cicd]

PR_Github #22332 [ run ] completed with state SUCCESS. Commit: e841c17
/LLM/main/L0_MergeRequest_PR pipeline #16836 completed with status: 'FAILURE'

[SimengLiu-nv]

PR_Github #22332 [ run ] completed with state SUCCESS. Commit: e841c17 /LLM/main/L0_MergeRequest_PR pipeline #16836 completed with status: 'FAILURE'

Hit failure documented with nvbug but somehow the test is unwaived.

[SimengLiu-nv]

/bot run

[tensorrt-cicd]

PR_Github #22456 [ run ] triggered by Bot. Commit: da3d7f8

[tensorrt-cicd]

PR_Github #22456 [ run ] completed with state SUCCESS. Commit: da3d7f8
/LLM/main/L0_MergeRequest_PR pipeline #16924 completed with status: 'FAILURE'

[SimengLiu-nv]

/bot run

[SimengLiu-nv]

PR_Github #22456 [ run ] completed with state SUCCESS. Commit: da3d7f8 /LLM/main/L0_MergeRequest_PR pipeline #16924 completed with status: 'FAILURE'

Hit a documented nvbug, but somehow it is not waived.

[tensorrt-cicd]

PR_Github #22482 [ run ] triggered by Bot. Commit: 2a0a0ef

[zeroepoch]

/bot reuse-pipeline

[zeroepoch]

/bot skip --comment "Passed previous run before rebasing"

[tensorrt-cicd]

PR_Github #22485 [ reuse-pipeline ] triggered by Bot. Commit: 2a0a0ef

[tensorrt-cicd]

PR_Github #22482 [ run ] completed with state ABORTED. Commit: 2a0a0ef
LLM/main/L0_MergeRequest_PR #16942 (Blue Ocean) completed with status: ABORTED

[tensorrt-cicd]

PR_Github #22485 [ reuse-pipeline ] completed with state SUCCESS. Commit: 2a0a0ef
Can't reuse PR_Github #22482 with status: ABORTED

[tensorrt-cicd]

PR_Github #22486 [ skip ] triggered by Bot. Commit: 2a0a0ef

[tensorrt-cicd]

PR_Github #22486 [ skip ] completed with state SUCCESS. Commit: 2a0a0ef
Skipping testing for commit 2a0a0ef