[https://nvbugs/5599515][fix] Fix PP bubbles. #8687
Conversation
Signed-off-by: Yuxian Qiu <[email protected]>
|
/bot run --disable-fail-fast |
📝 WalkthroughWalkthroughThe changes add NVTX profiling instrumentation to request queue and pipeline parallel executor operations, reorganize microbatch synchronization timing in the executor loop to perform sampler event synchronization before inter-PP communication, and expand test coverage by adding new multi-GPU throughput test configurations. Changes
Sequence DiagramsequenceDiagram
participant Executor as Executor Loop (PP)
participant Stage2 as Stage 2:<br/>Inter-PP Comm
participant Sampler as Sampler
participant NextStage as Next PP Stage
Note over Executor,Sampler: Microbatch N Processing
Executor->>Sampler: Schedule microbatch N
Sampler-->>Executor: sampler_event_N (async)
Note over Executor,Sampler: Before Stage 2: Sync Previous Batch
rect rgb(220, 240, 255)
Note over Executor: NEW: synchronize<br/>sampler_event_(N-1)<br/>before Stage 2
Executor->>Sampler: Wait for sampler_event_(N-1)
Sampler-->>Executor: Complete
end
Note over Executor,Stage2: Stage 2: Inter-PP Communication
rect rgb(240, 220, 255)
Note over Stage2: Wrap send in nvtx_range<br/>("send_sample_state")
Stage2->>NextStage: isend_object(sample_state_N)
NextStage-->>Stage2: Sent
end
Note over Executor,Sampler: Continue with next microbatch
Estimated code review effort🎯 2 (Simple) | ⏱️ ~12 minutes
Pre-merge checks and finishing touches❌ Failed checks (2 warnings)
✅ Passed checks (1 passed)
✨ Finishing touches
🧪 Generate unit tests (beta)
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. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 0
🧹 Nitpick comments (1)
tensorrt_llm/_torch/pyexecutor/py_executor.py (1)
927-933: Consider renaming the variable to clarify intent.The variable
prev_microbatch_idis calculated here and then recalculated with a different offset at line 938-940. While the logic appears correct (synchronizing the sampler event for microbatch N-1 before processing Stage 2 for a different microbatch), the variable name reuse could be confusing for maintainability.Consider using distinct variable names to clarify the different semantic purposes:
- prev_microbatch_id = (microbatch_id - - 1) % self.num_micro_batches - previous_batch = self.micro_batches[prev_microbatch_id] + sync_microbatch_id = (microbatch_id - + 1) % self.num_micro_batches + previous_batch = self.micro_batches[sync_microbatch_id] if previous_batch is not None: with nvtx_range("sync_previous_sampler_event"): previous_batch.sample_state.sampler_event.synchronize()
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (6)
tensorrt_llm/_torch/pyexecutor/executor_request_queue.py(2 hunks)tensorrt_llm/_torch/pyexecutor/py_executor.py(2 hunks)tests/integration/defs/accuracy/test_llm_api_pytorch.py(1 hunks)tests/integration/test_lists/qa/llm_function_core.txt(1 hunks)tests/integration/test_lists/qa/llm_function_core_sanity.txt(1 hunks)tests/integration/test_lists/test-db/l0_dgx_b300.yml(1 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/pyexecutor/executor_request_queue.pytensorrt_llm/_torch/pyexecutor/py_executor.pytests/integration/defs/accuracy/test_llm_api_pytorch.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/pyexecutor/executor_request_queue.pytensorrt_llm/_torch/pyexecutor/py_executor.pytests/integration/defs/accuracy/test_llm_api_pytorch.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/pyexecutor/executor_request_queue.pytensorrt_llm/_torch/pyexecutor/py_executor.pytests/integration/defs/accuracy/test_llm_api_pytorch.py
🧠 Learnings (1)
📚 Learning: 2025-09-09T09:40:45.658Z
Learnt from: fredricz-20070104
PR: NVIDIA/TensorRT-LLM#7645
File: tests/integration/test_lists/qa/llm_function_core.txt:648-648
Timestamp: 2025-09-09T09:40:45.658Z
Learning: In TensorRT-LLM test lists, it's common and intentional for the same test to appear in multiple test list files when they serve different purposes (e.g., llm_function_core.txt for comprehensive core functionality testing and llm_function_core_sanity.txt for quick sanity checks). This duplication allows tests to be run in different testing contexts.
Applied to files:
tests/integration/test_lists/qa/llm_function_core.txttests/integration/test_lists/qa/llm_function_core_sanity.txt
🧬 Code graph analysis (2)
tensorrt_llm/_torch/pyexecutor/executor_request_queue.py (3)
tensorrt_llm/_utils.py (1)
nvtx_range(904-923)tensorrt_llm/_torch/distributed/communicator.py (9)
recv_object(376-377)recv_object(592-605)prev_pp_rank(95-96)is_last_pp_rank(79-80)isend_object(373-374)isend_object(613-625)next_pp_rank(91-92)send_object(370-371)send_object(608-610)tensorrt_llm/mapping.py (3)
prev_pp_rank(268-272)is_last_pp_rank(256-257)next_pp_rank(274-278)
tensorrt_llm/_torch/pyexecutor/py_executor.py (3)
tensorrt_llm/_utils.py (1)
nvtx_range(904-923)tensorrt_llm/_torch/distributed/communicator.py (3)
isend_object(373-374)isend_object(613-625)next_pp_rank(91-92)tensorrt_llm/mapping.py (1)
next_pp_rank(274-278)
⏰ 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 (7)
tests/integration/test_lists/qa/llm_function_core_sanity.txt (1)
42-42: Sanity list includes throughput_pp4_mtp — OK.Duplication with core list is intentional for different testing contexts. Based on learnings.
tests/integration/test_lists/qa/llm_function_core.txt (1)
491-491: throughput_pp4_mtp entry verified — ready to merge.The entry is correctly defined in the test parameters and consistently referenced across all configuration files (core list, sanity list, and test-db).
tests/integration/defs/accuracy/test_llm_api_pytorch.py (2)
2080-2082: Appended ID verified across all test definitions and lists—no inconsistencies detected.The verification confirms
throughput_pp4_mtpis correctly:
- Defined in the test parameter list (line 2081)
- Referenced consistently in all test configuration files with uniform naming and ordering
- Properly integrated with other throughput variants
2064-2075: Configuration verified as correct and properly integrated.All technical checks pass:
- Param tuple correctly placed with tp=1, pp=4, ep=1, mtp_nextn=1 (line 2064–2075)
- enable_lm_head_tp_in_adp=False is safe with tp=1 (no-op)
- skip_less_mpi_world_size(4) matches pp_size=4
- Corresponding ID "throughput_pp4_mtp" exists (line 2081)
The attention_dp=True with mtp_nextn=1 combination is a design choice; confirm with domain experts if this is the intended NVFP4 throughput profile for PP4.
tests/integration/test_lists/test-db/l0_dgx_b300.yml (1)
76-76: Test entry is valid and correctly configured — approve.The test ID
throughput_pp4_mtpis confirmed in the parametrization at line 2080 of test_llm_api_pytorch.py. The timeout value (180s) matches adjacent test variants. The GB110 hardware filtering mentioned in the original review is already explicitly defined in the file-level condition block (gpu wildcard:*gb110*, system_gpu_count: 4), confirming this is intentional by design for thel0_dgx_b300.ymltarget configuration.tensorrt_llm/_torch/pyexecutor/py_executor.py (1)
961-966: LGTM! NVTX instrumentation improves observability.The added profiling instrumentation for the inter-PP sample state communication will help identify communication bottlenecks in pipeline parallel execution.
tensorrt_llm/_torch/pyexecutor/executor_request_queue.py (1)
559-559: LGTM! NVTX instrumentation enhances profiling capabilities.The added NVTX profiling around request broadcasting and inter-PP communication provides valuable observability for diagnosing pipeline parallel bottlenecks. The instrumentation is non-invasive and follows the same pattern used in
py_executor.py.Also applies to: 578-591
|
PR_Github #22643 [ run ] triggered by Bot. Commit: |
|
PR_Github #22643 [ run ] completed with state |
Signed-off-by: Yuxian Qiu <[email protected]>
|
/bot run --disable-fail-fast |
|
PR_Github #22695 [ run ] triggered by Bot. Commit: |
Signed-off-by: Yuxian Qiu <[email protected]>
|
/bot run --disable-fail-fast |
|
PR_Github #22699 [ run ] triggered by Bot. Commit: |
|
PR_Github #22695 [ run ] completed with state |
|
PR_Github #22699 [ run ] completed with state |
|
/bot run |
|
PR_Github #22792 [ run ] triggered by Bot. Commit: |
|
/bot run --disable-fail-fast |
|
PR_Github #22826 [ run ] triggered by Bot. Commit: |
|
PR_Github #22792 [ run ] completed with state |
|
PR_Github #22826 [ run ] completed with state |
|
/bot run --disable-fail-fast |
|
PR_Github #22863 [ run ] triggered by Bot. Commit: |
|
PR_Github #22863 [ run ] completed with state |
|
/bot run --disable-fail-fast |
|
PR_Github #22963 [ run ] triggered by Bot. Commit: |
|
PR_Github #22963 [ run ] completed with state |
|
/bot run --disable-fail-fast |
|
PR_Github #23021 [ run ] triggered by Bot. Commit: |
|
PR_Github #23021 [ run ] completed with state |
|
/bot skip --comment "The failed test accuracy/test_llm_api_pytorch.py::TestDeepSeekV32::test_nvfp4_multi_gpus[baseline] is unrelated to this PR, skip CI" |
|
PR_Github #23093 [ skip ] triggered by Bot. Commit: |
|
PR_Github #23093 [ skip ] completed with state |
Signed-off-by: Yuxian Qiu <[email protected]> Signed-off-by: FredricZ-2007 <[email protected]>
5 participants
Summary by CodeRabbit
Tests
Chores
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 thestage-listparameter 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.mdand the
scripts/test_to_stage_mapping.pyhelper.kill
killKill all running builds associated with pull request.
skip
skip --comment COMMENTSkip 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-pipelineReuse 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.