[TRTLLM-5508][feat] check input tokens + improve error handling #5170
Conversation
|
/bot run |
1 similar comment
|
/bot run |
|
PR_Github #8771 [ run ] triggered by Bot |
|
PR_Github #8771 [ run ] completed with state |
ixlmar
force-pushed
the
feat/nvbug-5307858
branch
from
108db04 to
b923131
Compare
|
/bot run --stage-list "A100X-PyTorch-1" |
|
PR_Github #8808 [ run ] triggered by Bot |
|
PR_Github #8808 [ run ] completed with state |
ixlmar
force-pushed
the
feat/nvbug-5307858
branch
from
b923131 to
1e7daaa
Compare
|
/bot run --stage-list "A100X-PyTorch-1" |
|
PR_Github #8815 [ run ] triggered by Bot |
|
PR_Github #8815 [ run ] completed with state |
ixlmar
force-pushed
the
feat/nvbug-5307858
branch
from
1e7daaa to
23df114
Compare
|
/bot run --stage-list "A100X-PyTorch-1" |
|
PR_Github #8829 [ run ] triggered by Bot |
|
PR_Github #8829 [ run ] completed with state |
ixlmar
force-pushed
the
feat/nvbug-5307858
branch
from
23df114 to
e3fb4f4
Compare
📝 WalkthroughWalkthroughThis update introduces token ID range validation for requests in both the Python and C++ components of the system. It adds new methods for checking token validity, enhances error handling and reporting in the benchmarking and executor logic, and extends the test suite with cases for invalid token handling. Several methods are refactored for clarity, and type annotations are improved throughout. Changes
Sequence Diagram(s)sequenceDiagram
participant User
participant PyExecutor
participant ExecutorRequestQueue
participant LlmRequest
User->>PyExecutor: Submit requests
PyExecutor->>ExecutorRequestQueue: Enqueue requests
ExecutorRequestQueue-->>PyExecutor: Request IDs
loop On fetch/activation
PyExecutor->>ExecutorRequestQueue: Fetch new requests
ExecutorRequestQueue-->>PyExecutor: LlmRequest(s)
PyExecutor->>LlmRequest: check_token_id_range(vocab_size)
alt Token ID valid
PyExecutor->>PyExecutor: Add to active_requests
else Token ID invalid
PyExecutor->>PyExecutor: _handle_errors (fail specific requests)
end
end
Note over PyExecutor: Only validated requests proceed to generation
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20 minutes Suggested labels
Suggested reviewers
Note ⚡️ Unit Test Generation is now available in beta!Learn more here, or try it out under "Finishing Touches" below. 📜 Recent review detailsConfiguration used: .coderabbit.yaml 📒 Files selected for processing (16)
✅ Files skipped from review due to trivial changes (1)
🚧 Files skipped from review as they are similar to previous changes (13)
🧰 Additional context used📓 Path-based instructions (2)**/*.py📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
Files:
**/*.{cpp,h,cu,py,cc,cxx,hpp}📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
Files:
🧠 Learnings (3)📓 Common learnings📚 Learning: in tensorrt_llm/executor/worker.py, the lora adapter cache optimization logic that checks `is_adapte...Applied to files:
📚 Learning: in tensorrt-llm's multimodal processing pipeline, shared tensor recovery using `from_shared_tensor()...Applied to files:
🔇 Additional comments (10)
✨ Finishing Touches
🧪 Generate unit tests
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. 🪧 TipsChatThere are 3 ways to chat with CodeRabbit:
SupportNeed help? Create a ticket on our support page for assistance with any issues or questions. CodeRabbit Commands (Invoked using PR comments)
Other keywords and placeholders
Documentation and Community
|
|
/bot run --stage-list "A100X-PyTorch-1" |
![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/model_engine.py (1)
930-938: Return type annotation narrows the contract—verify all overrides/stubsAdding
-> DecoderModelForCausalLMimproves readability, but_load_modelis occasionally monkey-patched in tests (see comments around line 299) and might return stub models that are notDecoderModelForCausalLM.
If any such override now returns a different subtype (e.g., a lightweighttorch.nn.Module), static checkers will emit errors. Confirm that:
- All overrides/stubs indeed inherit from
DecoderModelForCausalLM, or- You relax the hint to a wider type (e.g.,
torch.nn.Module | DecoderModelForCausalLM) to keep tests/builds warning-free.No functional issue—just a typing concern.
📜 Review details
Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (4)
tensorrt_llm/_torch/pyexecutor/model_engine.py(1 hunks)tensorrt_llm/_torch/pyexecutor/py_executor.py(11 hunks)tensorrt_llm/bench/benchmark/throughput.py(2 hunks)tensorrt_llm/bench/benchmark/utils/asynchronous.py(5 hunks)
🧰 Additional context used
🧠 Learnings (1)
tensorrt_llm/_torch/pyexecutor/py_executor.py (1)
Learnt from: amitz-nv
PR: NVIDIA/TensorRT-LLM#5616
File: tensorrt_llm/executor/worker.py:375-384
Timestamp: 2025-07-17T09:01:27.374Z
Learning: In tensorrt_llm/executor/worker.py, the LoRA adapter cache optimization logic that checks `is_adapter_in_cpu_cache()` and conditionally passes None for weights/config has a known race condition issue that cannot be solved with simple error handling or verification checks. This is a known limitation that requires a more comprehensive solution.
🧬 Code Graph Analysis (1)
tensorrt_llm/bench/benchmark/throughput.py (1)
tensorrt_llm/logger.py (1)
error(125-126)
🔇 Additional comments (10)
tensorrt_llm/bench/benchmark/throughput.py (1)
5-5: Improved error handling with explicit process termination.The addition of explicit exception handlers for
KeyboardInterruptand generic exceptions, along withsys.exit(-1)calls, ensures proper non-zero exit status on errors. This is a good practice for benchmarking tools where exit codes are important for CI/CD pipelines.Also applies to: 454-459
tensorrt_llm/bench/benchmark/utils/asynchronous.py (4)
39-40: Excellent refactoring of task lifecycle management and error handling.The worker method improvements are well-designed:
- Non-blocking queue operations with
get_nowait()+sleep(1)prevent deadlocks- Error aggregation provides detailed failure counts for better debugging
- Proper task cleanup ensures no resource leaks
- Clear separation of concerns with the
_raise_for_failedhelperAlso applies to: 97-145
198-205: Proper async shutdown implementation.Converting
stop()to async and awaiting task completion ensures graceful shutdown without race conditions. This is the correct pattern for async task management.
209-210: Fixed critical typo and improved busy state detection.
- The
busyproperty now accurately reflects ongoing work by checking both queue state and active task completion- The typo fix ensures
_iteration_log_taskis properly assigned for trackingAlso applies to: 215-215
302-302: Proper async cleanup in benchmark flow.Adding
awaitfor bothenqueue_taskcancellation andbackend.stop()ensures clean resource management and prevents task leakage.Also applies to: 304-304
tensorrt_llm/_torch/pyexecutor/py_executor.py (5)
353-383: Well-designed enqueue refactoring following DRY principles.The unified
_enqueue_implmethod elegantly handles both single and batch enqueueing with optional query support. The use ofitertools.repeat(None)for the batch case is a clean solution.
1360-1384: Excellent separation of fetch, validate, and activate phases.The
_fetch_and_activate_new_requestsmethod properly validates requests before activation, immediately failing invalid ones. This prevents invalid requests from consuming resources and provides fast feedback to users.
1981-2002: Improved granular error handling allows selective request failure.The enhanced
_handle_errorsmethod now supports failing specific requests without affecting others, which is crucial for maintaining service availability when individual requests have issues (e.g., invalid tokens).
1210-1219: Clean extraction of request counting logic.Extracting
_count_active_requests_all_ranksimproves code organization and testability while maintaining the same functionality.
1350-1359: Token validation logic has an off-by-one error.The upper bound check should use
>= self.model_engine.model.lm_head.num_embeddingsinstead of> ... - 1for clarity and correctness.- if any( - min(tokens) < 0 or max(tokens) > - self.model_engine.model.lm_head.num_embeddings - 1 - for tokens in token_lists): + if any( + min(tokens) < 0 or max(tokens) >= + self.model_engine.model.lm_head.num_embeddings + for tokens in token_lists):Likely an incorrect or invalid review comment.
ixlmar
force-pushed
the
feat/nvbug-5307858
branch
from
e3fb4f4 to
10536f6
Compare
|
/bot run --stage-list "A100X-PyTorch-1" |
|
PR_Github #12209 [ run ] triggered by Bot |
|
PR_Github #12209 [ run ] completed with state |
ixlmar
force-pushed
the
feat/nvbug-5307858
branch
2 times, most recently
from
072deea to
51d8ac0
Compare
|
/bot run |
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (4)
tensorrt_llm/bench/benchmark/throughput.py (2)
454-456: Consider using exit code 130 for KeyboardInterrupt.While the implementation correctly handles
KeyboardInterruptand ensures clean termination, the standard Unix convention for keyboard interrupt (SIGINT) is exit code 130, not -1.- sys.exit(-1) + sys.exit(130)
457-459: Improve error logging and consider exit code 1.The generic exception handler has two areas for improvement:
- The error message could be more informative by including exception type
- Exit code 1 is more conventional than -1 for general errors
- except Exception as e: - logger.error(f"Error during benchmarking: {e}") - sys.exit(-1) + except Exception as e: + logger.error(f"Error during benchmarking: {type(e).__name__}: {e}") + sys.exit(1)tensorrt_llm/_torch/pyexecutor/executor_request_queue.py (2)
122-122: Consider replacing assertion with proper exceptionWhile the assertion provides clear error messaging, assertions can be disabled with Python's
-Oflag. Consider using a proper exception for more reliable error handling in production.- assert self.active, "PyExecutor has already been shutdown." + if not self.active: + raise RuntimeError("PyExecutor has already been shutdown.")
140-146: Good backward-compatible signature enhancementThe addition of the optional
queryparameter maintains backward compatibility while supporting StarAttention functionality. The method correctly delegates to the new unified implementation.Consider using a more specific type than
Optional[List]if the query structure is well-defined:query: Optional[List[int]] = None # if it's token IDs
📜 Review details
Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (6)
tensorrt_llm/_torch/pyexecutor/executor_request_queue.py(7 hunks)tensorrt_llm/_torch/pyexecutor/model_engine.py(1 hunks)tensorrt_llm/_torch/pyexecutor/py_executor.py(8 hunks)tensorrt_llm/bench/benchmark/throughput.py(2 hunks)tensorrt_llm/bench/benchmark/utils/asynchronous.py(5 hunks)tests/unittest/conftest.py(2 hunks)
🚧 Files skipped from review as they are similar to previous changes (4)
- tests/unittest/conftest.py
- tensorrt_llm/_torch/pyexecutor/model_engine.py
- tensorrt_llm/_torch/pyexecutor/py_executor.py
- tensorrt_llm/bench/benchmark/utils/asynchronous.py
⏰ 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 (6)
tensorrt_llm/bench/benchmark/throughput.py (1)
5-5: LGTM: Import added for process termination.The
sysimport is correctly added to support the newsys.exit(-1)calls in the exception handlers.tensorrt_llm/_torch/pyexecutor/executor_request_queue.py (5)
8-9: LGTM: Import improvements enhance type safetyThe addition of
repeatfrom itertools and comprehensive type imports support the refactored enqueue methods and improved type annotations throughout the file.Also applies to: 17-18
116-147: Excellent refactoring: Unified enqueue implementationThe introduction of
_enqueue_implas a private helper eliminates code duplication while maintaining the same functionality. The use ofzip(requests, repeat(None))for batch operations is elegant and efficient.
121-121: Excellent modernization: Context manager for lock safetyConverting to context managers (
with self.enqueue_lock:) ensures locks are always properly released, even during exceptions. This eliminates potential deadlock scenarios and follows modern Python concurrency best practices.Also applies to: 149-149, 154-154, 159-160
202-202: LGTM: Comprehensive type annotationsThe addition of explicit return type annotations (
-> List[LlmRequest]) and modern parameter type syntax improves type safety and IDE support without changing functionality.Also applies to: 210-210, 224-224, 429-430, 448-449
116-147: Well-coordinated refactoring supports broader improvementsThis refactoring provides a solid foundation for the request validation and error handling improvements mentioned in the PR objectives, while maintaining backward compatibility and improving code maintainability.
|
PR_Github #12553 [ run ] triggered by Bot |
|
PR_Github #12553 [ run ] completed with state |
|
PR_Github #14013 [ run ] completed with state |
ixlmar
force-pushed
the
feat/nvbug-5307858
branch
from
6af3918 to
f20658b
Compare
|
/bot run --stage-list="A100X-PyTorch-1" |
|
PR_Github #14023 [ run ] triggered by Bot |
|
PR_Github #14019 [ run ] completed with state |
There was a problem hiding this comment.
Actionable comments posted: 2
📜 Review details
Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (16)
cpp/include/tensorrt_llm/batch_manager/llmRequest.h(1 hunks)cpp/tensorrt_llm/batch_manager/llmRequest.cpp(1 hunks)cpp/tensorrt_llm/nanobind/batch_manager/bindings.cpp(1 hunks)cpp/tensorrt_llm/pybind/batch_manager/bindings.cpp(1 hunks)tensorrt_llm/_torch/pyexecutor/executor_request_queue.py(7 hunks)tensorrt_llm/_torch/pyexecutor/model_engine.py(1 hunks)tensorrt_llm/_torch/pyexecutor/py_executor.py(9 hunks)tensorrt_llm/bench/benchmark/throughput.py(2 hunks)tensorrt_llm/bench/benchmark/utils/asynchronous.py(6 hunks)tensorrt_llm/executor/proxy.py(1 hunks)tensorrt_llm/executor/worker.py(1 hunks)tests/integration/defs/test_e2e.py(2 hunks)tests/integration/defs/trt_test_alternative.py(1 hunks)tests/integration/test_lists/test-db/l0_a10.yml(1 hunks)tests/unittest/conftest.py(2 hunks)tests/unittest/llmapi/test_llm_pytorch.py(2 hunks)
✅ Files skipped from review due to trivial changes (1)
- cpp/include/tensorrt_llm/batch_manager/llmRequest.h
🚧 Files skipped from review as they are similar to previous changes (14)
- tests/integration/defs/trt_test_alternative.py
- cpp/tensorrt_llm/batch_manager/llmRequest.cpp
- tensorrt_llm/executor/proxy.py
- tensorrt_llm/bench/benchmark/throughput.py
- tests/unittest/conftest.py
- cpp/tensorrt_llm/nanobind/batch_manager/bindings.cpp
- tensorrt_llm/executor/worker.py
- tensorrt_llm/_torch/pyexecutor/model_engine.py
- tests/integration/test_lists/test-db/l0_a10.yml
- tests/integration/defs/test_e2e.py
- cpp/tensorrt_llm/pybind/batch_manager/bindings.cpp
- tensorrt_llm/_torch/pyexecutor/py_executor.py
- tensorrt_llm/_torch/pyexecutor/executor_request_queue.py
- tests/unittest/llmapi/test_llm_pytorch.py
🧰 Additional context used
📓 Path-based instructions (2)
**/*.py
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
**/*.py: The 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 filenames should use snake_case (e.g., some_file.py).
Python classes should use PascalCase (e.g., class SomeClass).
Python functions and methods should use snake_case (e.g., def my_awesome_function():).
Python local variables should use snake_case. Prefix k for variable names that start with a number (e.g., k_99th_percentile = ...).
Python global variables should use upper snake_case and prefix G (e.g., G_MY_GLOBAL = ...).
Python constants should use upper snake_case (e.g., 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 interfaces that may be used outside a file, prefer docstrings over comments in Python.
Comments in Python should be reserved for code within a function, or interfaces that are local to a file.
Use Google style docstrings for classes and functions in Python, which can be parsed by Sphinx.
Attributes and variables in Python can be documented inline; attribute docstrings will be rendered under the docstring for the class.
Avoid using reflection in Python when functionality can be easily achieved without it.
When using try-except blocks in Python, limit the except to the smallest set of errors possible.
When using try-except blocks to handle multiple possible variable types in Python, keep the body of the try as small as possible, using the else block to implement the logic.
Files:
tensorrt_llm/bench/benchmark/utils/asynchronous.py
**/*.{cpp,h,hpp,cc,cxx,cu,py}
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
All TensorRT-LLM Open Source Software code should contain an NVIDIA copyright header that includes the current year. This includes .cpp, .h, .cu, .py, and any other source files which are compiled or interpreted.
Files:
tensorrt_llm/bench/benchmark/utils/asynchronous.py
🧠 Learnings (1)
📓 Common learnings
Learnt from: moraxu
PR: NVIDIA/TensorRT-LLM#6303
File: tests/integration/test_lists/qa/examples_test_list.txt:494-494
Timestamp: 2025-07-28T17:06:08.621Z
Learning: In TensorRT-LLM testing, it's common to have both CLI flow tests (test_cli_flow.py) and PyTorch API tests (test_llm_api_pytorch.py) for the same model. These serve different purposes: CLI flow tests validate the traditional command-line workflow, while PyTorch API tests validate the newer LLM API backend. Both are legitimate and should coexist.
🔇 Additional comments (9)
tensorrt_llm/bench/benchmark/utils/asynchronous.py (9)
7-7: LGTM: Import addition supports new type hints.The addition of
Setto the typing imports is necessary for the newself._tasks: Set[asyncio.Task]type annotation.
39-41: Good: Enhanced error tracking and task management.The new instance variables improve error handling by:
_task_errors: Collecting all task exceptions instead of failing fast_backend_taskand_iteration_log_task: Proper type hints for better async task management
80-80: Minor improvement: More efficient token extraction.Converting the list comprehension to a generator expression with
chainis more memory-efficient for large token sequences.
98-116: Excellent: Comprehensive error aggregation and reporting.The new error handling approach provides several benefits:
- Collects all task errors instead of stopping on first failure
- Groups identical errors with counts for cleaner reporting
- Provides detailed error summary in the exception message
This aligns well with the
--disable-fail-fastbehavior mentioned in the PR comments.
138-144: Good: Proper cleanup with task cancellation and error propagation.The cleanup logic in the finally block correctly:
- Cancels all remaining tasks
- Waits for their completion
- Raises any accumulated errors
This ensures no tasks are left hanging and all errors are properly reported.
197-204: LGTM: Proper async conversion with task awaiting.Converting
stop()to async and awaiting the tasks instead of cancelling them ensures proper cleanup. The assertion on_backend_taskis appropriate since it should always be set afterrun()is called.
208-209: Good: More accurate busy state detection.The updated
busyproperty now correctly checks both queue state and running tasks, providing a more accurate representation of the manager's activity state.
214-214: LGTM: Type hint consistency.Adding the explicit type hint for
_iteration_log_taskmaintains consistency with the new type annotations.
117-144: Critical: Ensure non-blocking worker loop doesn’t spin the CPUWhile switching from blocking
Queue.get()toget_nowait()plusawait asyncio.sleep(0)yields better task interleaving, an always‐empty inbox could still cause a tight yield loop under high concurrency. We do have benchmarks in place—
- tests/unittest/scaffolding/test_bench.py:
async_scaffolding_benchmark- tensorrt_llm/bench/benchmark/throughput.py & low_latency.py invoking
async_benchmark—but none explicitly measure CPU utilization when the queue is empty. Please add CPU‐profiling or utilization metrics to your existing benchmarks to confirm there’s no busy‐waiting regression.
• tensorrt_llm/bench/benchmark/utils/asynchronous.py (worker: lines 117–144)
• tests/unittest/scaffolding/test_bench.py (async_scaffolding_benchmark)
• tensorrt_llm/bench/benchmark/throughput.py & low_latency.py (async_benchmark callers)
|
PR_Github #14023 [ run ] completed with state |
ixlmar
force-pushed
the
feat/nvbug-5307858
branch
from
f20658b to
99093d3
Compare
|
/bot run --stage-list="A100X-PyTorch-1" |
|
PR_Github #14103 [ run ] triggered by Bot |
|
@FrankD412 Did a short sanity-check using: $ cat /tmp/input.yml
cuda_graph_config:
enable_padding: true
batch_sizes:
- 1
- 2
- 4
- 8
- 16
- 32
- 64
- 128
- 256
- 384
$ trtllm-bench \
--model TinyLlama-1.1B-Chat-v1.0 \
--model_path /models/llama-models-v2/TinyLlama-1.1B-Chat-v1.0 \
throughput \
--dataset tiny_input1_output1_requests10000_good.json \
--backend pytorch \
--extra_llm_api_options /tmp/input.yml \
--concurrency 1000Results with new [old] code: Re-running with the old code suggests that the observed differences are within the spread. |
|
PR_Github #14103 [ run ] completed with state |
|
/bot run |
|
PR_Github #14117 [ run ] triggered by Bot |
|
PR_Github #14117 [ run ] completed with state |
…IA#5170) Signed-off-by: ixlmar <[email protected]> Signed-off-by: Lanyu Liao <[email protected]>
…IA#5170) Signed-off-by: ixlmar <[email protected]>
6 participants
[TRTLLM-5508] feat: check input tokens + improve error handling
Description
Fixes https://nvbugspro.nvidia.com/bug/5307858 and implements necessary error handling improvements in PyTorch backend and
trtllm-bench.Test Coverage
Tests covering invalid requests still need to be designed and implemented. Tests should best cover the LLM API as well astrtllm-bench.Added test cases:
tests/unittest/llmapi/test_llm_pytorch.py::test_llm_invalid_input_token_async,tests/unittest/llmapi/test_llm_pytorch.py::test_llm_invalid_input_tokentests/integration/defs/test_e2e.py::test_trtllm_bench_invalid_token_pytorchGitHub 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 [--disable-fail-fast --skip-test --stage-list "A10-1, xxx" --gpu-type "A30, H100_PCIe" --add-multi-gpu-test --only-multi-gpu-test --disable-multi-gpu-test --post-merge --extra-stage "H100_PCIe-[Post-Merge]-1, xxx"]Launch build/test pipelines. All previously running jobs will be killed.
--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 "A10-1, xxx"(OPTIONAL) : Only run the specified test stages. Examples: "A10-1, 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.--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. Will also run 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-[Post-Merge]-1, xxx"(OPTIONAL) : Run the ordinary L0 pre-merge pipeline and specified test stages. Examples: --extra-stage "H100_PCIe-[Post-Merge]-1, xxx".For guidance on mapping tests to stage names, see
docs/source/reference/ci-overview.md.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.
Summary by CodeRabbit
New Features
Bug Fixes
Tests
Refactor
Style