[None][feat] Have ability to cancel disagg request if KV cache resource are exhausted #9155
Conversation
Signed-off-by: Patrice Castonguay <55748270+pcastonguay@users.noreply.github.com>
|
/bot run --disable-fail-fast |
|
PR_Github #24499 [ run ] triggered by Bot. Commit: |
📝 WalkthroughWalkthroughThe pull request extends the Changes
Sequence Diagram(s)sequenceDiagram
actor User
participant Config as CacheTransceiverConfig
participant Transceiver as CacheTransceiver
participant Future as Future<Result>
User->>Config: Create with kvTransferSenderFutureTimeoutMs
Config->>Config: Store timeout (validate > 0)
Transceiver->>Transceiver: checkContextTransferStatus()
Transceiver->>Config: getKvTransferSenderFutureTimeoutMs()
Config-->>Transceiver: timeout value
alt timeout configured
Transceiver->>Future: wait_for(timeout)
Future-->>Transceiver: ready within timeout
Transceiver->>Transceiver: Retrieve result
Transceiver->>Transceiver: Mark kDISAGG_CONTEXT_COMPLETE
Transceiver->>Transceiver: Erase future from map
else timeout expires
Transceiver->>Future: wait_for(timeout)
Future-->>Transceiver: timeout occurred
Transceiver->>Transceiver: Log warning
Note over Transceiver: Leave future in map
else unexpected status
Transceiver->>Transceiver: Mark kDISAGG_TRANS_ERROR
Transceiver->>Transceiver: Erase future from map
end
Estimated code review effort🎯 2 (Simple) | ⏱️ ~12 minutes
Areas requiring attention:
Suggested reviewers
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 |
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 (4)
cpp/tensorrt_llm/executor/cacheTransceiverConfig.cpp (1)
34-38: operator== omits kvTransferSenderFutureTimeoutMs.Equality should include the newly added field to avoid false positives/negatives in comparisons and tests.
Apply this diff:
bool CacheTransceiverConfig::operator==(CacheTransceiverConfig const& other) const { - return mMaxTokensInBuffer == other.mMaxTokensInBuffer && mBackendType == other.mBackendType - && mKvTransferTimeoutMs == other.mKvTransferTimeoutMs; + return mMaxTokensInBuffer == other.mMaxTokensInBuffer + && mBackendType == other.mBackendType + && mKvTransferTimeoutMs == other.mKvTransferTimeoutMs + && mKvTransferSenderFutureTimeoutMs == other.mKvTransferSenderFutureTimeoutMs; }cpp/tensorrt_llm/pybind/executor/executorConfig.cpp (2)
417-427: Include the fourth parameter in pickle serialization for CacheTransceiverConfig.The field
kv_transfer_sender_future_timeout_msis lost during pickle round-trip. Extend both getstate and setstate to handle all 4 constructor parameters instead of 3.auto cacheTransceiverConfigGetstate = [](tle::CacheTransceiverConfig const& self) - { return py::make_tuple(self.getBackendType(), self.getMaxTokensInBuffer(), self.getKvTransferTimeoutMs()); }; + { + return py::make_tuple(self.getBackendType(), + self.getMaxTokensInBuffer(), + self.getKvTransferTimeoutMs(), + self.getKvTransferSenderFutureTimeoutMs()); + }; auto cacheTransceiverConfigSetstate = [](py::tuple const& state) { - if (state.size() != 3) + if (state.size() != 4) { throw std::runtime_error("Invalid CacheTransceiverConfig state!"); } - return tle::CacheTransceiverConfig(state[0].cast<tle::CacheTransceiverConfig::BackendType>(), - state[1].cast<std::optional<size_t>>(), state[2].cast<std::optional<int>>()); + return tle::CacheTransceiverConfig( + state[0].cast<tle::CacheTransceiverConfig::BackendType>(), + state[1].cast<std::optional<size_t>>(), + state[2].cast<std::optional<int>>(), + state[3].cast<std::optional<int>>()); };Apply the same fix to cpp/tensorrt_llm/nanobind/executor/executorConfig.cpp (lines ~435–445).
448-454: Fix py::init template arity and pickle completeness for CacheTransceiverConfig.The C++ constructor accepts 4 parameters, but the pybind
py::inittemplate declares only 3 types while binding 4 arguments—this causes a binding error. Additionally, the pickle implementation only preserves 3 fields, droppingkv_transfer_sender_future_timeout_msduring serialization.Update lines 449–450 in
cpp/tensorrt_llm/pybind/executor/executorConfig.cppand lines 418, 421–426 to handle all 4 parameters:- py::class_<tle::CacheTransceiverConfig>(m, "CacheTransceiverConfig") - .def(py::init<std::optional<tle::CacheTransceiverConfig::BackendType>, std::optional<size_t>, - std::optional<int>>(), + py::class_<tle::CacheTransceiverConfig>(m, "CacheTransceiverConfig") + .def(py::init< + std::optional<tle::CacheTransceiverConfig::BackendType>, + std::optional<size_t>, + std::optional<int>, + std::optional<int> + >(), py::arg("backend") = std::nullopt, py::arg("max_tokens_in_buffer") = std::nullopt, py::arg("kv_transfer_timeout_ms") = std::nullopt, py::arg("kv_transfer_sender_future_timeout_ms") = std::nullopt)Also update the pickle getstate and setstate functions (lines 417–426) to include the 4th field.
cpp/tensorrt_llm/nanobind/executor/executorConfig.cpp (1)
435-445: Clarify that pybind11 has the same bug—this isn't mirroring an existing fix.The review correctly identifies that
getKvTransferSenderFutureTimeoutMs()must be added to the pickle tuple to prevent data loss. The method exists, the constructor accepts the 4th parameter, and the proposed type casting is correct. However, pybind11's getstate also only returns a 3-element tuple and setstate expects 3 elements, so this is not a mirror of an existing pybind11 fix—rather, both implementations need the same fix applied. Update the justification to reflect that both codebases require the same correction.
🧹 Nitpick comments (5)
tensorrt_llm/_torch/pyexecutor/kv_cache_transceiver.py (1)
114-114: Remove redundant attribute or document its purpose.This value isn’t used in Python; the C++ impl reads it via
_to_pybind(). Keeping it risks drift. Either remove or add a brief comment explaining why it’s stored.- self.kv_transfer_sender_future_timeout_ms = cache_transceiver_config.kv_transfer_sender_future_timeout_ms + # Note: timeout is carried via cache_transceiver_config._to_pybind() to C++. No local use here. + # self.kv_transfer_sender_future_timeout_ms = cache_transceiver_config.kv_transfer_sender_future_timeout_mscpp/tests/unit_tests/executor/serializeUtilsTest.cpp (1)
1031-1038: Add nullopt round‑trip coverage for the new field.Also test when
kvTransferSenderFutureTimeoutMsis unset to ensure serialization preservesstd::nullopt.TEST(SerializeUtilsTest, CacheTransceiverConfig) { texec::CacheTransceiverConfig cacheTransceiverConfig( tensorrt_llm::executor::CacheTransceiverConfig::BackendType::UCX, 1024, 100, 1000); auto cacheTransceiverConfig2 = serializeDeserialize(cacheTransceiverConfig); @@ EXPECT_EQ(cacheTransceiverConfig.getKvTransferSenderFutureTimeoutMs(), cacheTransceiverConfig2.getKvTransferSenderFutureTimeoutMs()); + + // Unset optional should remain unset after round-trip + texec::CacheTransceiverConfig unsetTimeout(texec::CacheTransceiverConfig::BackendType::UCX, 1024, 100, std::nullopt); + auto unsetTimeout2 = serializeDeserialize(unsetTimeout); + EXPECT_FALSE(unsetTimeout2.getKvTransferSenderFutureTimeoutMs().has_value()); }examples/disaggregated/README.md (1)
22-23: Clarify default and units for the new knob.State the default (1000 ms) and keep units consistent.
- # Timeout in milliseconds to wait for the sender future to be ready when scheduled batch size is 0. This allows the request to be eventually cancelled by the user or because of kv_transfer_timeout_ms - kv_transfer_sender_future_timeout_ms: <int> + # Timeout (milliseconds) to wait for the sender future when scheduled batch size is 0. + # Default: 1000 ms. Enables eventual cancellation by the user or by kv_transfer_timeout_ms. + kv_transfer_sender_future_timeout_ms: <int>tensorrt_llm/llmapi/llm_args.py (1)
1566-1571: Mark as prototype and confirm cross‑lang default alignment.Consider marking this new knob as unstable for now. Also ensure C++/bindings default matches Python’s 1000 ms to avoid behavior divergence when constructed from C++.
- kv_transfer_sender_future_timeout_ms: Optional[int] = Field( - default=1000, + kv_transfer_sender_future_timeout_ms: Optional[int] = Field( + default=1000, + status="prototype", gt=0, description=... )tests/unittest/llmapi/test_llm_pytorch.py (1)
1108-1112: Stabilize stats assertion; mirror the retry used above.Avoid flaky
len(results) == 1by adding a bounded retry with sleep, liketest_llm_context_only_timed_out.- results = llm.get_stats(2) - assert len(results) == 1 + max_retries = 10 + for _ in range(max_retries): + results = llm.get_stats(2) + if len(results) == 1: + break + time.sleep(1) + else: + pytest.fail(f"Failed to get stats with len==1 after {max_retries} retries")
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (11)
cpp/include/tensorrt_llm/executor/executor.h(2 hunks)cpp/tensorrt_llm/batch_manager/cacheTransceiver.cpp(2 hunks)cpp/tensorrt_llm/executor/cacheTransceiverConfig.cpp(3 hunks)cpp/tensorrt_llm/executor/serialization.cpp(1 hunks)cpp/tensorrt_llm/nanobind/executor/executorConfig.cpp(1 hunks)cpp/tensorrt_llm/pybind/executor/executorConfig.cpp(1 hunks)cpp/tests/unit_tests/executor/serializeUtilsTest.cpp(2 hunks)examples/disaggregated/README.md(1 hunks)tensorrt_llm/_torch/pyexecutor/kv_cache_transceiver.py(1 hunks)tensorrt_llm/llmapi/llm_args.py(1 hunks)tests/unittest/llmapi/test_llm_pytorch.py(1 hunks)
🧰 Additional context used
🧠 Learnings (8)
📚 Learning: 2025-08-20T06:56:02.889Z
Learnt from: eopXD
Repo: NVIDIA/TensorRT-LLM PR: 6768
File: cpp/tensorrt_llm/batch_manager/kvCacheManager.cpp:577-579
Timestamp: 2025-08-20T06:56:02.889Z
Learning: In cpp/tensorrt_llm/batch_manager/kvCacheManager.cpp, maxSequenceLength is now enforced as a non-optional argument in the BlockManager constructor, so concerns about std::nullopt defaulting to 0 are not applicable. When windowSize > maxSequenceLength, a warning should be added instead of handling optional parameter cases.
Applied to files:
cpp/include/tensorrt_llm/executor/executor.hcpp/tensorrt_llm/executor/serialization.cppcpp/tensorrt_llm/executor/cacheTransceiverConfig.cppcpp/tensorrt_llm/pybind/executor/executorConfig.cppcpp/tensorrt_llm/batch_manager/cacheTransceiver.cppcpp/tensorrt_llm/nanobind/executor/executorConfig.cpp
📚 Learning: 2025-08-14T21:04:50.248Z
Learnt from: thorjohnsen
Repo: NVIDIA/TensorRT-LLM PR: 6910
File: cpp/tensorrt_llm/batch_manager/kvCacheManager.cpp:0-0
Timestamp: 2025-08-14T21:04:50.248Z
Learning: In KV cache onboarding logic during prefill in cpp/tensorrt_llm/batch_manager/kvCacheManager.cpp, when calculating which blocks fall within the attention window, use getTokensPerBlock() to advance token indices rather than block->getUniqueTokens().size(), because the calculation needs to consider the post-prefill state where blocks will be filled to capacity, not their current token count.
Applied to files:
cpp/include/tensorrt_llm/executor/executor.hcpp/tensorrt_llm/executor/serialization.cppcpp/tensorrt_llm/executor/cacheTransceiverConfig.cppcpp/tensorrt_llm/batch_manager/cacheTransceiver.cpp
📚 Learning: 2025-08-15T06:46:54.897Z
Learnt from: eopXD
Repo: NVIDIA/TensorRT-LLM PR: 6767
File: cpp/tensorrt_llm/batch_manager/kvCacheManager.cpp:0-0
Timestamp: 2025-08-15T06:46:54.897Z
Learning: In cpp/tensorrt_llm/batch_manager/kvCacheManager.cpp addToken function, newly allocated blocks are unshared by design. The beam search path in addToken (when sequence.getNumTokens() > windowSize) is currently broken/non-functional with SWA, so the block allocation doesn't follow a shared-then-unshared pattern.
Applied to files:
cpp/tensorrt_llm/executor/serialization.cppcpp/tensorrt_llm/executor/cacheTransceiverConfig.cppcpp/tests/unit_tests/executor/serializeUtilsTest.cpp
📚 Learning: 2025-08-21T09:41:49.347Z
Learnt from: eopXD
Repo: NVIDIA/TensorRT-LLM PR: 6768
File: cpp/tensorrt_llm/batch_manager/kvCacheManager.cpp:2010-2045
Timestamp: 2025-08-21T09:41:49.347Z
Learning: In cpp/tensorrt_llm/batch_manager/kvCacheManager.cpp, updateSequenceCacheBlockOffsets is specifically for updating bookkeeping when blocks are added during the context phase, not for refreshing offsets after detach operations. During detach operations, GenerationRequest::removeFrontBlock handles the necessary cache block bookkeeping internally.
Applied to files:
cpp/tensorrt_llm/executor/serialization.cppcpp/tensorrt_llm/batch_manager/cacheTransceiver.cpp
📚 Learning: 2025-08-06T08:18:28.669Z
Learnt from: zhengd-nv
Repo: NVIDIA/TensorRT-LLM PR: 6633
File: cpp/tensorrt_llm/batch_manager/dataTransceiverImpl.cpp:145-155
Timestamp: 2025-08-06T08:18:28.669Z
Learning: In cpp/tensorrt_llm/batch_manager/dataTransceiverImpl.cpp, the existing `mMtxForMap` mutex in DataSenderImpl is sufficient to synchronize measurement file operations in the `release` method, as all file operations occur within the same critical section that protects the `mRequestToSession` map access.
Applied to files:
cpp/tensorrt_llm/batch_manager/cacheTransceiver.cpp
📚 Learning: 2025-08-20T06:48:45.368Z
Learnt from: eopXD
Repo: NVIDIA/TensorRT-LLM PR: 6768
File: cpp/include/tensorrt_llm/batch_manager/kvCacheManager.h:0-0
Timestamp: 2025-08-20T06:48:45.368Z
Learning: In cpp/tensorrt_llm/batch_manager/kvCacheManager.cpp, updateSequenceCacheBlockOffsets is only called when adding a sequence, not during detach operations. During detach, the cache block bookkeeping is handled by GenerationRequest::removeFrontBlock.
Applied to files:
cpp/tensorrt_llm/batch_manager/cacheTransceiver.cpp
📚 Learning: 2025-07-28T17:06:08.621Z
Learnt from: moraxu
Repo: NVIDIA/TensorRT-LLM PR: 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.
Applied to files:
tests/unittest/llmapi/test_llm_pytorch.py
📚 Learning: 2025-08-14T15:38:01.771Z
Learnt from: MatthiasKohl
Repo: NVIDIA/TensorRT-LLM PR: 6904
File: cpp/tensorrt_llm/pybind/thop/bindings.cpp:55-57
Timestamp: 2025-08-14T15:38:01.771Z
Learning: In TensorRT-LLM Python bindings, tensor parameter collections like mla_tensor_params and spec_decoding_tensor_params are kept as required parameters without defaults to maintain API consistency, even when it might affect backward compatibility.
Applied to files:
tensorrt_llm/llmapi/llm_args.py
🧬 Code graph analysis (8)
cpp/include/tensorrt_llm/executor/executor.h (2)
cpp/tensorrt_llm/executor/cacheTransceiverConfig.cpp (15)
CacheTransceiverConfig(24-32)other(34-38)other(34-34)setKvTransferTimeoutMs(50-57)setKvTransferTimeoutMs(50-50)setKvTransferSenderFutureTimeoutMs(59-66)setKvTransferSenderFutureTimeoutMs(59-59)getMaxTokensInBuffer(73-76)getMaxTokensInBuffer(73-73)getBackendType(68-71)getBackendType(68-68)getKvTransferTimeoutMs(78-81)getKvTransferTimeoutMs(78-78)getKvTransferSenderFutureTimeoutMs(83-86)getKvTransferSenderFutureTimeoutMs(83-83)tensorrt_llm/llmapi/llm_args.py (1)
CacheTransceiverConfig(1545-1579)
cpp/tensorrt_llm/executor/cacheTransceiverConfig.cpp (1)
tensorrt_llm/llmapi/llm_args.py (1)
CacheTransceiverConfig(1545-1579)
cpp/tensorrt_llm/pybind/executor/executorConfig.cpp (1)
cpp/tensorrt_llm/executor/cacheTransceiverConfig.cpp (8)
getKvTransferTimeoutMs(78-81)getKvTransferTimeoutMs(78-78)setKvTransferTimeoutMs(50-57)setKvTransferTimeoutMs(50-50)getKvTransferSenderFutureTimeoutMs(83-86)getKvTransferSenderFutureTimeoutMs(83-83)setKvTransferSenderFutureTimeoutMs(59-66)setKvTransferSenderFutureTimeoutMs(59-59)
cpp/tensorrt_llm/batch_manager/cacheTransceiver.cpp (1)
cpp/tests/unit_tests/multi_gpu/cacheTransceiverTest.cpp (7)
future(167-176)request(893-943)request(893-893)request(945-952)request(945-945)request(954-1005)request(954-954)
cpp/tests/unit_tests/executor/serializeUtilsTest.cpp (3)
cpp/tensorrt_llm/executor/cacheTransceiverConfig.cpp (1)
CacheTransceiverConfig(24-32)tensorrt_llm/llmapi/llm_args.py (1)
CacheTransceiverConfig(1545-1579)cpp/tests/unit_tests/multi_gpu/cacheTransceiverTest.cpp (2)
serializeDeserialize(72-81)serializeDeserialize(72-72)
tests/unittest/llmapi/test_llm_pytorch.py (4)
tensorrt_llm/llmapi/llm_args.py (2)
KvCacheConfig(1366-1510)CacheTransceiverConfig(1545-1579)tensorrt_llm/llmapi/llm.py (3)
LLM(1101-1117)generate(259-341)get_stats(517-528)tensorrt_llm/sampling_params.py (1)
SamplingParams(113-540)tensorrt_llm/disaggregated_params.py (1)
DisaggregatedParams(15-92)
tensorrt_llm/llmapi/llm_args.py (1)
tensorrt_llm/builder.py (1)
default(45-50)
cpp/tensorrt_llm/nanobind/executor/executorConfig.cpp (1)
cpp/tensorrt_llm/executor/cacheTransceiverConfig.cpp (16)
getBackendType(68-71)getBackendType(68-68)setBackendType(40-43)setBackendType(40-40)getMaxTokensInBuffer(73-76)getMaxTokensInBuffer(73-73)setMaxTokensInBuffer(45-48)setMaxTokensInBuffer(45-45)getKvTransferTimeoutMs(78-81)getKvTransferTimeoutMs(78-78)setKvTransferTimeoutMs(50-57)setKvTransferTimeoutMs(50-50)getKvTransferSenderFutureTimeoutMs(83-86)getKvTransferSenderFutureTimeoutMs(83-83)setKvTransferSenderFutureTimeoutMs(59-66)setKvTransferSenderFutureTimeoutMs(59-59)
⏰ 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 (8)
cpp/tests/unit_tests/executor/serializeUtilsTest.cpp (1)
792-793: Constructor arity update looks correct.UCX + (1024, 100, 1000) matches the extended signature. LGTM.
tensorrt_llm/llmapi/llm_args.py (1)
1577-1580: Pybind forwarding LGTM.Field is correctly forwarded to
_CacheTransceiverConfig.cpp/tensorrt_llm/executor/serialization.cpp (1)
1293-1296: Wire-format extension adds two fields; confirm compatibility and add round‑trip tests.Appending kvTransferTimeoutMs and kvTransferSenderFutureTimeoutMs is fine, but older payloads will be shorter and older readers may choke on the longer layout. Please confirm producer/consumer version lockstep or add versioning/tolerant reads, and add a serialize/deserialize round‑trip test covering nullopt and set values.
Also applies to: 1302-1304, 1311-1313
cpp/tensorrt_llm/executor/cacheTransceiverConfig.cpp (1)
24-31: Ctor/getter/setter for sender-future timeout: LGTM.Validation and plumbing look correct.
Also applies to: 59-66, 83-86
cpp/tensorrt_llm/pybind/executor/executorConfig.cpp (1)
460-462: New property binding: LGTM.Getter/setter wiring for kv_transfer_sender_future_timeout_ms looks correct.
cpp/include/tensorrt_llm/executor/executor.h (1)
1468-1470: API surface for sender-future timeout: LGTM.Signatures and docs are consistent with implementation. Ensure equality includes this field (addressed in cpp).
Also applies to: 1475-1475, 1480-1481, 1489-1492
cpp/tensorrt_llm/nanobind/executor/executorConfig.cpp (2)
466-472: 4‑arg constructor binding: LGTM.
479-481: New property binding: LGTM.
Signed-off-by: Patrice Castonguay <55748270+pcastonguay@users.noreply.github.com>
Signed-off-by: Patrice Castonguay <55748270+pcastonguay@users.noreply.github.com>
|
/bot run --disable-fail-fast |
|
PR_Github #24502 [ run ] triggered by Bot. Commit: |
|
PR_Github #24499 [ run ] completed with state |
|
PR_Github #24502 [ run ] completed with state |
Signed-off-by: Patrice Castonguay <55748270+pcastonguay@users.noreply.github.com>
|
/bot run --disable-fail-fast |
|
PR_Github #24529 [ run ] triggered by Bot. Commit: |
|
PR_Github #24529 [ run ] completed with state |
|
/bot run --disable-fail-fast |
|
PR_Github #24609 [ run ] triggered by Bot. Commit: |
|
PR_Github #24609 [ run ] completed with state |
Superjomn
left a comment
There was a problem hiding this comment.
Overall LGTM on the llmapi changes.
…ce are exhausted (NVIDIA#9155) Signed-off-by: Patrice Castonguay <55748270+pcastonguay@users.noreply.github.com> Signed-off-by: lkomali <lkomali@nvidia.com>
|
Hi @pcastonguay @chuangz0 @Tabrizian. I noticed that when kv cache resource is exhausted, decoding node will also block on future.get(), could you please take a look? |
Summary by CodeRabbit
New Features
Documentation
Tests
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.