[TRTLLM-7963][feat] Use CUDAGraph to improve the tuning accuracy for AutoTuner. #9089
Conversation
📝 WalkthroughWalkthroughAdded CUDA graph support to autotuning profiling with a new Changes
Sequence Diagram(s)sequenceDiagram
participant Profiler as Profiler
participant Kernel as _profile_single_kernel
participant CUDA as CUDA Runtime
Note over Profiler,CUDA: With use_cudagraph=True (new path)
Profiler->>Kernel: call with use_cudagraph=True
Kernel->>CUDA: create stream context
Kernel->>CUDA: warmup runs in stream
Kernel->>CUDA: create CUDAGraph
loop repeated execution
Kernel->>CUDA: replay graph
end
Kernel->>CUDA: synchronize & record time
Kernel-->>Profiler: return timing
Note over Profiler,CUDA: With use_cudagraph=False (legacy path)
Profiler->>Kernel: call with use_cudagraph=False
loop repeated runs
Kernel->>CUDA: launch kernel
Kernel->>CUDA: host delay (stream_delay_micro_secs)
end
Kernel->>CUDA: synchronize & record time
Kernel-->>Profiler: return timing
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20 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: 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/_torch/autotuner.py (1)
859-881: Update docstring to document the new parameter.The
use_cudagraphparameter was added to the method signature but is not documented in the docstring. The Args section should include this new parameter.Apply this diff to update the docstring:
def _profile_single_kernel( self, runner: TunableRunner, inputs: List[torch.Tensor], tactic: Any, use_cudagraph: bool = True, **kwargs, ) -> float: """Profile a single kernel implementation for performance measurement. Args: runner (TunableRunner): The runner implementation to profile inputs (List[torch.Tensor]): Input tensors for the kernel tactic (Any): Tactic to use for this profiling run + use_cudagraph (bool): Whether to use CUDA graphs for profiling (default: True) + When True, captures and replays kernels using CUDA graphs for more accurate timing. + When False, uses sequential execution with stream delay to reduce host overhead. Returns: Average execution time in milliseconds
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (1)
tensorrt_llm/_torch/autotuner.py(5 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/autotuner.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/autotuner.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/autotuner.py
🧠 Learnings (2)
📓 Common learnings
Learnt from: MrGeva
Repo: NVIDIA/TensorRT-LLM PR: 7219
File: tensorrt_llm/_torch/auto_deploy/compile/backends/torch_cudagraph.py:162-168
Timestamp: 2025-09-04T07:33:10.618Z
Learning: When users explicitly provide cuda_graph_batch_sizes in TorchCudagraphCompiler, respect their choices and only sanitize the values (clamp, dedupe, sort) without forcing additional batch sizes like 1 or max_batch_size. Only add commonly-used batch sizes when falling back to the heuristic.
📚 Learning: 2025-08-26T09:37:10.463Z
Learnt from: jiaganc
Repo: NVIDIA/TensorRT-LLM PR: 7031
File: tensorrt_llm/bench/dataclasses/configuration.py:90-104
Timestamp: 2025-08-26T09:37:10.463Z
Learning: In TensorRT-LLM, the `get_pytorch_perf_config()` method returns `self.pytorch_config` which can contain default `cuda_graph_config` values, so `llm_args` may already have this config before the extra options processing.
Applied to files:
tensorrt_llm/_torch/autotuner.py
🧬 Code graph analysis (1)
tensorrt_llm/_torch/autotuner.py (1)
tensorrt_llm/_torch/pyexecutor/model_engine.py (2)
warmup(86-92)warmup(522-550)
⏰ 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/autotuner.py (3)
102-102: LGTM! Field addition looks good.The new
use_cudagraphfield appropriately controls CUDA graph usage during profiling, with a sensible default ofTruefor improved accuracy.
813-818: LGTM! Parameter forwarding is correct.The
use_cudagraphflag is appropriately forwarded fromtuning_configto the profiling method.
888-916: LGTM! Profiling logic is correctly implemented.The profiling flow appropriately handles both CUDA graph and non-graph paths:
- Warmup runs occur first in the stream context
- CUDA graphs capture repeated executions when enabled
- Stream delay is correctly applied only for non-graph profiling to reduce host overhead
- Timing and synchronization are properly implemented for both paths
|
/bot run --disable-fail-fast |
|
PR_Github #24279 [ run ] triggered by Bot. Commit: |
|
PR_Github #24279 [ run ] completed with state |
hyukn
force-pushed
the
feat/autotune_profile_cudagraph
branch
from
5572f86 to
90e428e
Compare
|
/bot run --disable-fail-fast |
|
PR_Github #24384 [ run ] triggered by Bot. Commit: |
|
PR_Github #24384 [ run ] completed with state |
|
/bot run --disable-fail-fast |
|
PR_Github #24433 [ run ] triggered by Bot. Commit: |
|
PR_Github #24433 [ run ] completed with state |
|
/bot run --disable-fail-fast |
|
PR_Github #24482 [ run ] triggered by Bot. Commit: |
|
PR_Github #24482 [ run ] completed with state |
hyukn
force-pushed
the
feat/autotune_profile_cudagraph
branch
from
90e428e to
f4bc776
Compare
|
/bot run --disable-fail-fast |
|
PR_Github #24517 [ run ] triggered by Bot. Commit: |
|
PR_Github #24517 [ run ] completed with state |
hyukn
force-pushed
the
feat/autotune_profile_cudagraph
branch
3 times, most recently
from
e15ff36 to
7a3429e
Compare
|
/bot run --disable-fail-fast |
|
PR_Github #24728 [ run ] completed with state |
|
/bot run --disable-fail-fast |
|
PR_Github #24768 [ run ] triggered by Bot. Commit: |
hyukn
force-pushed
the
feat/autotune_profile_cudagraph
branch
from
7b62cec to
447f0f0
Compare
|
/bot run --disable-fail-fast |
|
PR_Github #24822 [ run ] triggered by Bot. Commit: |
|
PR_Github #24768 [ run ] completed with state |
hyukn
force-pushed
the
feat/autotune_profile_cudagraph
branch
from
447f0f0 to
d623c7d
Compare
|
/bot run --disable-fail-fast |
|
PR_Github #24825 [ run ] triggered by Bot. Commit: |
|
PR_Github #24822 [ run ] completed with state |
…AutoTuner. Signed-off-by: Yukun He <[email protected]>
hyukn
force-pushed
the
feat/autotune_profile_cudagraph
branch
from
d623c7d to
b964920
Compare
|
/bot run --disable-fail-fast |
|
PR_Github #24839 [ run ] triggered by Bot. Commit: |
|
PR_Github #24825 [ run ] completed with state |
|
PR_Github #24839 [ run ] completed with state |
|
/bot run --disable-fail-fast |
|
PR_Github #24915 [ run ] triggered by Bot. Commit: |
|
PR_Github #24915 [ run ] completed with state |
|
/bot run --disable-fail-fast |
|
PR_Github #24967 [ run ] triggered by Bot. Commit: |
|
PR_Github #24967 [ run ] completed with state |
Signed-off-by: Yukun He <[email protected]>
|
/bot run --disable-fail-fast |
|
PR_Github #25060 [ run ] triggered by Bot. Commit: |
|
PR_Github #25060 [ run ] completed with state |
6 participants
Summary by CodeRabbit
New Features
use_cudagraphoption to control CUDA graph-based profiling during autotuning.Improvements
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
Update tava architecture diagram if there is a significant design change in PR.
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.