[#8389][fix] Update group attention matching to first map to custom torch attention #8638
Conversation
…en patterns with repeat_kv Signed-off-by: Fridah-nv <[email protected]>
Fridah-nv
changed the title
Fridah-nv
changed the title
📝 WalkthroughWalkthroughRefactored attention pattern matching transforms to rename grouped attention patterns, explicitly separating SDPA-to-torch_attention and repeat_kv-with-torch_attention cases. Updated configuration keys, function names, class names, and registrations to reflect new semantic distinctions in pattern matching strategies. Changes
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20 minutes Rationale: Changes involve semantic renaming across multiple related components (configuration, core implementation, tests) with new function and class definitions. While the edits follow a consistent pattern of renaming existing functionality, the implementation of two new generator functions ( Pre-merge checks and finishing touches❌ Failed checks (2 warnings, 1 inconclusive)
✅ Passed checks (2 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/auto_deploy/transform/library/attention.py (1)
620-653: Fix docstring reference to use correct transform key.Line 629 mentions
match_repeat_kv_with_torch_attentionbut the actual registered key ismatch_grouped_attention(line 656).Apply this diff:
- This transform should run BEFORE match_repeat_kv_with_torch_attention to ensure + This transform should run BEFORE match_grouped_attention to ensure SDPA calls are converted first.
🧹 Nitpick comments (4)
tensorrt_llm/_torch/auto_deploy/transform/library/attention.py (4)
376-428: Consider addingstrict=Truetozip()calls for safer argument matching.The
zip()calls at lines 400 and 415 should includestrict=Trueto ensureargsandargnameshave matching lengths, preventing silent bugs if they become misaligned.Apply this diff:
- m = dict(zip(argnames, args)) + m = dict(zip(argnames, args, strict=True))(Apply to both pattern_fn and replacement_fn at lines 400 and 415)
Based on static analysis hints.
431-490: Consider improving exception chain handling.At line 473, the exception raised in the
except KeyErrorclause should useraise ... from errto preserve the exception chain, aiding debugging.Apply this diff:
- except KeyError: - raise RuntimeError(f"Unexpected arg name: {name}") + except KeyError as err: + raise RuntimeError(f"Unexpected arg name: {name}") from errBased on static analysis hints.
493-555: Addstrict=Truetozip()calls for argument safety.The
zip()calls at lines 517 and 541 should includestrict=Trueto catch argument count mismatches early.Apply this diff:
- m = dict(zip(argnames, args)) + m = dict(zip(argnames, args, strict=True))(Apply to both pattern_fn and replacement_fn at lines 517 and 541)
Based on static analysis hints.
1-1: Copyright header requirement.As per coding guidelines, Python source files should include the NVIDIA Apache-2.0 copyright header with the current year. This appears to be missing from this file.
If this file is newly created or significantly modified, consider adding the copyright header at the top of the file.
As per coding guidelines.
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (4)
tensorrt_llm/_torch/auto_deploy/config/default.yaml(1 hunks)tensorrt_llm/_torch/auto_deploy/transform/library/attention.py(7 hunks)tests/unittest/_torch/auto_deploy/unit/singlegpu/transformations/library/test_attention_matcher.py(1 hunks)tests/unittest/_torch/auto_deploy/unit/singlegpu/transformations/library/test_attention_matcher_hf.py(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:
tests/unittest/_torch/auto_deploy/unit/singlegpu/transformations/library/test_attention_matcher_hf.pytests/unittest/_torch/auto_deploy/unit/singlegpu/transformations/library/test_attention_matcher.pytensorrt_llm/_torch/auto_deploy/transform/library/attention.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:
tests/unittest/_torch/auto_deploy/unit/singlegpu/transformations/library/test_attention_matcher_hf.pytests/unittest/_torch/auto_deploy/unit/singlegpu/transformations/library/test_attention_matcher.pytensorrt_llm/_torch/auto_deploy/transform/library/attention.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:
tests/unittest/_torch/auto_deploy/unit/singlegpu/transformations/library/test_attention_matcher_hf.pytests/unittest/_torch/auto_deploy/unit/singlegpu/transformations/library/test_attention_matcher.pytensorrt_llm/_torch/auto_deploy/transform/library/attention.py
🧬 Code graph analysis (1)
tensorrt_llm/_torch/auto_deploy/transform/library/attention.py (2)
tensorrt_llm/_torch/auto_deploy/transform/interface.py (4)
get(519-521)TransformRegistry(503-531)register(509-516)BaseTransform(213-500)tensorrt_llm/_torch/auto_deploy/utils/pattern_matcher.py (1)
register_ad_pattern(99-182)
🪛 Ruff (0.14.1)
tensorrt_llm/_torch/auto_deploy/transform/library/attention.py
399-399: Avoid specifying long messages outside the exception class
(TRY003)
400-400: zip() without an explicit strict= parameter
Add explicit value for parameter strict=
(B905)
414-414: Avoid specifying long messages outside the exception class
(TRY003)
415-415: zip() without an explicit strict= parameter
Add explicit value for parameter strict=
(B905)
473-473: Within an except clause, raise exceptions with raise ... from err or raise ... from None to distinguish them from errors in exception handling
(B904)
473-473: Avoid specifying long messages outside the exception class
(TRY003)
540-540: Avoid specifying long messages outside the exception class
(TRY003)
541-541: 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). (2)
- GitHub Check: Check PR Checklist Resolution
- GitHub Check: Pre-commit Check
🔇 Additional comments (5)
tensorrt_llm/_torch/auto_deploy/config/default.yaml (1)
37-39: LGTM! Configuration keys renamed to reflect new attention matching semantics.The renaming clearly distinguishes SDPA-to-torch_attention mapping from repeat_kv handling, aligning with the refactored transform implementation.
tests/unittest/_torch/auto_deploy/unit/singlegpu/transformations/library/test_attention_matcher_hf.py (1)
44-47: LGTM! Test configuration updated to use renamed matcher keys.The configuration now correctly references
match_sdpa_to_torch_attentionandmatch_grouped_attention, consistent with the refactored transform registration.tensorrt_llm/_torch/auto_deploy/transform/library/attention.py (2)
558-617: LGTM! Repeat KV pattern generation correctly handles n_rep parameter.The function properly enumerates pattern variations and includes
n_repin scalar workarounds, which should help correctly identify GQA scenarios where repeat_kv is applied.
656-691: LGTM! Transform correctly handles repeat_kv with torch_attention patterns.The class properly registers the
match_grouped_attentiontransform and its docstring accurately describes the ordering dependency withmatch_sdpa_to_torch_attention.tests/unittest/_torch/auto_deploy/unit/singlegpu/transformations/library/test_attention_matcher.py (1)
449-452: LGTM! Test configuration correctly updated with new matcher keys.The test now uses
match_sdpa_to_torch_attentionandmatch_grouped_attention, aligning with the refactored transform registration and config updates.
|
If there is no change in the final graph and no change in the custom ops, I think the perf side should be good. |
3 participants
Details in/ closes #8389
TODO: manual perf tests after I get access to the clusters
Summary by CodeRabbit
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.