[None][fix] avoid implicit cudaStreamSynchronize in sample_async.#10120
[None][fix] avoid implicit cudaStreamSynchronize in sample_async.#10120yuxianq merged 3 commits intoNVIDIA:mainfrom
Conversation
Signed-off-by: Yuxian Qiu <142763828+yuxianq@users.noreply.github.com>
|
/bot run --disable-fail-fast |
📝 WalkthroughWalkthroughModified stop-word detection and index retrieval logic in a sampler utility. Changed Changes
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: 0
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/pyexecutor/sampler.py (1)
2626-2648: Critical: Logic error and performance regression from implicit synchronizations.This refactored implementation has two major issues:
Logic bug (line 2647): Direct assignment
per_step[step_idx, request_idx, beam_idx] = matchwill overwrite previous True values when checking multiple stop words. Use OR assignment instead:if match: per_step[step_idx, request_idx, beam_idx] = TruePerformance regression: The nested loops with tensor indexing and
.all()calls (lines 2641-2647) cause multiple implicitcudaStreamSynchronizeoperations per iteration—exactly what this PR aims to eliminate. Based on learnings, tensor access inside loops must be avoided.The current implementation contradicts the PR objective and will severely degrade performance with overlap scheduling.
🔎 Suggested vectorized approach to avoid synchronization
Refactor to use batch tensor operations instead of element-wise access:
# Pseudo-code outline: # 1. Build all stop word sequences as a batched tensor # 2. Build all token sequences (across steps) as a batched tensor # 3. Use batched comparison operations # 4. Reduce results without accessing individual elements # This keeps all operations on GPU without sync pointsThe exact implementation requires careful handling of variable-length stop words and sequences, but the key principle is: avoid
.item(), element-wise assignment, and.all()calls inside loops over requests/beams/steps.Based on learnings, this pattern appears in similar code paths that were previously optimized to avoid tensor access in loops.
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (1)
tensorrt_llm/_torch/pyexecutor/sampler.py(3 hunks)
🧰 Additional context used
📓 Path-based instructions (2)
**/*.py
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
**/*.py: Code developed for TensorRT-LLM should conform to Python 3.8+
Indent Python code with 4 spaces. Do not use tabs
Always maintain the namespace when importing in Python, even if only one class or function from a module is used
Python files should use snake_case naming:some_file.py
Python classes should use PascalCase naming:class SomeClass
Python functions and methods should use snake_case naming:def my_awesome_function():
Python local variables should use snake_case naming:my_variable = ...
Python variable names that start with a number should be prefixed with 'k':k_99th_percentile = ...
Python global variables should use upper snake_case with prefix 'G':G_MY_GLOBAL = ...
Python constants should use upper snake_case naming:MY_CONSTANT = ...
Avoid shadowing variables declared in an outer scope in Python
Initialize all externally visible members of a Python class in the constructor
For Python interfaces that may be used outside a file, prefer docstrings over comments
Python comments should be reserved for code within a function, or interfaces that are local to a file
Use Google style docstrings in Python for classes and functions, which can be parsed by Sphinx
Python attributes and variables can be documented inline with type and description
Avoid using reflection in Python when functionality can be easily achieved without reflection
When using try-except blocks in Python, limit the except to the smallest set of errors possible
When using try-except blocks in Python to handle multiple possible variable types (duck-typing), keep the body of the try as small as possible, using the else block for logic
Files:
tensorrt_llm/_torch/pyexecutor/sampler.py
**/*.{cpp,h,cu,cuh,py}
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
All TensorRT-LLM Open Source Software code should contain an NVIDIA copyright header that includes the year of its latest meaningful modification
Files:
tensorrt_llm/_torch/pyexecutor/sampler.py
🧠 Learnings (7)
📓 Common learnings
Learnt from: tongyuantongyu
Repo: NVIDIA/TensorRT-LLM PR: 9655
File: tensorrt_llm/_torch/pyexecutor/sampler.py:3031-3031
Timestamp: 2025-12-12T03:27:18.859Z
Learning: In tensorrt_llm/_torch/pyexecutor/sampler.py, when reviewing code that iterates through requests, ensure it does not convert excessive data into Python lists. Instead, the code should use torch.gather or indexing to gather only the data that will be used in the for loop before converting to Python lists. This minimizes data movement and improves performance.
Learnt from: tongyuantongyu
Repo: NVIDIA/TensorRT-LLM PR: 9655
File: tensorrt_llm/_torch/pyexecutor/sampler.py:3031-3031
Timestamp: 2025-12-12T03:27:18.859Z
Learning: In tensorrt_llm/_torch/pyexecutor/sampler.py, when reviewing code that iterates through requests, ensure it does not access torch.Tensor objects (CPU or GPU) inside the loop. Instead, the code should use .tolist() to convert batched data tensors to Python lists beforehand, and then access the list in the for loop. This is a critical performance consideration.
Learnt from: ixlmar
Repo: NVIDIA/TensorRT-LLM PR: 7294
File: tensorrt_llm/_torch/pyexecutor/sampler.py:1191-1197
Timestamp: 2025-08-28T10:22:02.288Z
Learning: In tensorrt_llm/_torch/pyexecutor/sampler.py, the object identity comparison `softmax_req_indices is not group_req_indices_cuda` on line ~1191 is intentional and used as an optimization to determine whether to reuse an existing indexer or create a new one, based on which code path was taken during tensor assignment.
Learnt from: ixlmar
Repo: NVIDIA/TensorRT-LLM PR: 7294
File: tensorrt_llm/_torch/pyexecutor/sampler.py:368-392
Timestamp: 2025-08-27T15:03:57.149Z
Learning: In TensorRT-LLM's sampler.py, int32 usage for softmax_indices and related tensor indexing is intentional and should not be changed to int64. The torch.IntTensor type hint is correct for the sample() function's softmax_indices parameter.
📚 Learning: 2025-12-12T03:27:18.859Z
Learnt from: tongyuantongyu
Repo: NVIDIA/TensorRT-LLM PR: 9655
File: tensorrt_llm/_torch/pyexecutor/sampler.py:3031-3031
Timestamp: 2025-12-12T03:27:18.859Z
Learning: In tensorrt_llm/_torch/pyexecutor/sampler.py, when reviewing code that iterates through requests, ensure it does not convert excessive data into Python lists. Instead, the code should use torch.gather or indexing to gather only the data that will be used in the for loop before converting to Python lists. This minimizes data movement and improves performance.
Applied to files:
tensorrt_llm/_torch/pyexecutor/sampler.py
📚 Learning: 2025-12-12T03:27:08.565Z
Learnt from: tongyuantongyu
Repo: NVIDIA/TensorRT-LLM PR: 9655
File: tensorrt_llm/_torch/pyexecutor/sampler.py:3031-3031
Timestamp: 2025-12-12T03:27:08.565Z
Learning: In files under tensorrt_llm/_torch/pyexecutor, avoid accessing torch.Tensor objects inside for-loops when iterating over requests. Convert batched tensors to Python lists beforehand using tensor.tolist(), and then iterate over those lists. This improves performance by reducing tensor-bound operations inside hot loops. Apply this pattern to similar code paths that process batches to access simple Python data structures (lists) inside loops.
Applied to files:
tensorrt_llm/_torch/pyexecutor/sampler.py
📚 Learning: 2025-08-28T10:22:02.288Z
Learnt from: ixlmar
Repo: NVIDIA/TensorRT-LLM PR: 7294
File: tensorrt_llm/_torch/pyexecutor/sampler.py:1191-1197
Timestamp: 2025-08-28T10:22:02.288Z
Learning: In tensorrt_llm/_torch/pyexecutor/sampler.py, the object identity comparison `softmax_req_indices is not group_req_indices_cuda` on line ~1191 is intentional and used as an optimization to determine whether to reuse an existing indexer or create a new one, based on which code path was taken during tensor assignment.
Applied to files:
tensorrt_llm/_torch/pyexecutor/sampler.py
📚 Learning: 2025-08-27T15:03:57.149Z
Learnt from: ixlmar
Repo: NVIDIA/TensorRT-LLM PR: 7294
File: tensorrt_llm/_torch/pyexecutor/sampler.py:368-392
Timestamp: 2025-08-27T15:03:57.149Z
Learning: In TensorRT-LLM's sampler.py, int32 usage for softmax_indices and related tensor indexing is intentional and should not be changed to int64. The torch.IntTensor type hint is correct for the sample() function's softmax_indices parameter.
Applied to files:
tensorrt_llm/_torch/pyexecutor/sampler.py
📚 Learning: 2025-08-28T10:25:22.370Z
Learnt from: ixlmar
Repo: NVIDIA/TensorRT-LLM PR: 7294
File: tensorrt_llm/_torch/pyexecutor/sampler.py:887-891
Timestamp: 2025-08-28T10:25:22.370Z
Learning: In tensorrt_llm/_torch/pyexecutor/sampler.py, the draft_probs and target_probs tensors have shapes [1, steps] not [steps, vocab_size] as might be expected, making the .squeeze(0) operations appropriate for removing the batch dimension of size 1.
Applied to files:
tensorrt_llm/_torch/pyexecutor/sampler.py
📚 Learning: 2025-08-13T16:20:37.987Z
Learnt from: dcampora
Repo: NVIDIA/TensorRT-LLM PR: 6867
File: tensorrt_llm/_torch/pyexecutor/sampler.py:67-72
Timestamp: 2025-08-13T16:20:37.987Z
Learning: In TensorRT-LLM sampler code, performance is prioritized over additional validation checks. The beam_width helper method intentionally returns the first request's beam_width without validating consistency across all requests to avoid performance overhead from iterating through the entire batch.
Applied to files:
tensorrt_llm/_torch/pyexecutor/sampler.py
🪛 Ruff (0.14.8)
tensorrt_llm/_torch/pyexecutor/sampler.py
2644-2644: zip() without an explicit strict= parameter
Add explicit value for parameter strict=
(B905)
⏰ 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/pyexecutor/sampler.py (1)
2413-2422: LGTM - Correctly avoids implicit synchronization.The change from returning a Python list to a CUDA tensor prevents implicit synchronization. The
non_blocking=Truetransfer ensures the operation remains asynchronous, aligning with the PR's objective to maintain overlap scheduler performance.
|
PR_Github #28978 [ run ] triggered by Bot. Commit: |
|
PR_Github #28978 [ run ] completed with state
|
Signed-off-by: Yuxian Qiu <142763828+yuxianq@users.noreply.github.com>
|
/bot run --disable-fail-fast |
|
PR_Github #29116 [ run ] triggered by Bot. Commit: |
|
PR_Github #29116 [ run ] completed with state |
|
@dcampora No e2e perf data, but I provide nsys result in the description, please check it. |
…IDIA#10120) Signed-off-by: Yuxian Qiu <142763828+yuxianq@users.noreply.github.com>
…IDIA#10120) Signed-off-by: Yuxian Qiu <142763828+yuxianq@users.noreply.github.com> Signed-off-by: Daniil Kulko <kulkodaniil@gmail.com>
6 participants
Description
Any cudaStreamSynchronize in sample_async will break the overlap feature of overlap scheduler and pipeline parallelism.
This PR fix it by eliminating all implicit cudaStreamSynchronize.
Here is nsys for LLaMA 405B TP2PP2:
Before fix:
After fix:
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.
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 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.