[None][feat] DeepEP LL fp8 dispatch/combine #7927
Conversation
yilin-void
changed the title
📝 WalkthroughWalkthroughUpdates DeepEP commit pin in CMake. Enables NVFP4 AWQ with prequant scales and adds a CutlassMoeFCRunner instantiation. Adds guarded FP8 activation handling for INT4 W4 in moeOp. Renames and generalizes a low-latency combine API to support multiple precisions. Refactors fused MoE Wide EP to centralize DeepEP dispatch/adaptation and support FP8 QDQ, NVFP4, and W4A8 paths. Changes
Sequence Diagram(s)sequenceDiagram
autonumber
participant W as WideEPMoE
participant D as DeepEP Buffer
participant A2A as AllToAll
participant AD as Adapter (new)
Note over W: Dispatch with multi-precision support
W->>A2A: alltoall(x, x_sf, expert counts)
A2A-->>W: recv_x, recv_x_sf, recv_expert_count
alt Quant mode == FP8 QDQ
W->>D: dispatch(recv_x.cast(fp8), recv_x_sf)
D-->>W: token_selected, token_scales?
W->>AD: adapt outputs (fp8)
else Quant mode == NVFP4
W->>D: dispatch_nvfp4(recv_x_fp4, recv_x_sf)
D-->>W: token_selected, token_scales
W->>AD: adapt outputs (nvfp4)
else Quant mode == W4A8
W->>W: apply prequant scales (W4A8)
W->>D: dispatch(recv_x_a8, recv_x_sf)
D-->>W: token_selected, token_scales?
W->>AD: adapt outputs (w4a8)
end
AD-->>W: token_selected_slots, final_scales?, topk_idx, topk_weights
sequenceDiagram
autonumber
participant W as WideEPMoE
participant BU as VariableLengthLowLatencyBuffer
Note over W: Low-precision combine (generalized)
W->>BU: low_latency_combine_low_precision(precision, hidden_states, global_scales?, topk_idx, topk_weights, handle)
BU-->>W: combined_hidden_states
Estimated code review effort🎯 4 (Complex) | ⏱️ ~60 minutes Pre-merge checks and finishing touches❌ Failed checks (2 warnings)
✅ Passed checks (1 passed)
✨ 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. 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 (2)
cpp/tensorrt_llm/deep_ep/CMakeLists.txt (1)
112-115: Bug: STREQUAL compares against the literal string; stamp check is always trueThis compares OLD_NVSHMEM_STAMP_CONTENT to the literal “NVSHMEM_STAMP_CONTENT”. It should compare variable-to-variable (and quote values). Otherwise it will keep wiping the nvshmem project even when not needed.
Apply this diff:
-if(NOT OLD_NVSHMEM_STAMP_CONTENT STREQUAL NVSHMEM_STAMP_CONTENT) +if(NOT "${OLD_NVSHMEM_STAMP_CONTENT}" STREQUAL "${NVSHMEM_STAMP_CONTENT}") file(REMOVE_RECURSE ${CMAKE_CURRENT_BINARY_DIR}/nvshmem_project-prefix) file(WRITE ${NVSHMEM_STAMP_FILE} "${NVSHMEM_STAMP_CONTENT}") endif()tensorrt_llm/_torch/modules/fused_moe/fused_moe_wide_ep.py (1)
1-1: Add required NVIDIA Apache-2.0 header (2025).Per coding guidelines, prepend the NVIDIA Apache-2.0 copyright header with current year.
+# Copyright (c) 2025, NVIDIA CORPORATION. All rights reserved. +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. import os
🧹 Nitpick comments (7)
cpp/tensorrt_llm/deep_ep/CMakeLists.txt (1)
128-132: Avoid globally overriding compilers for the whole configure when only NVSHMEM needs GCCOverriding CMAKE_{C,CXX}_COMPILER and CMAKE_CUDA_HOST_COMPILER globally can perturb the rest of the project when users pick Clang. Prefer setting compilers via ExternalProject’s CMAKE_CACHE_ARGS only, or isolate into a dedicated toolchain for the ExternalProject.
cpp/tensorrt_llm/thop/moeOp.cpp (1)
203-217: Guarded FP8-activation path for W4A8 is consistent with kernel instantiation
- The specialization CutlassMoeFCRunner<__nv_fp8_e4m3, cutlass::uint4b_t, __nv_bfloat16, __nv_fp8_e4m3> aligns with the explicit instantiation added in moe_kernels.cu.
- Erroring on non-W4A8 for FP8 activations is a good guard.
Minor:
- Consider a short comment noting SM/ENABLE_FP8 constraints to aid future refactors.
tensorrt_llm/_torch/modules/fused_moe/deep_ep_utils.py (1)
182-197: Avoid magic numbers for precision; prefer Literal or constantsThe precision parameter uses 0/1. This is brittle. Suggest tightening the type and using named constants for clarity.
Apply this diff locally to the signature:
-from typing import List, Optional, Tuple, Union +from typing import List, Optional, Tuple, Union, Literaland
- def low_latency_combine_low_precision(self, precision: int, + def low_latency_combine_low_precision(self, precision: Literal[0, 1], hidden_states: torch.Tensor, global_scales: Optional[torch.Tensor], topk_idx: torch.Tensor, topk_weights: torch.Tensor, handle: Tuple):Optional: define constants at module top for readability:
# Place near the top of the module FP8_PRECISION = 0 NVFP4_PRECISION = 1tensorrt_llm/_torch/modules/fused_moe/fused_moe_wide_ep.py (4)
188-193: Remove stray no-op expression; keep env flags.Line 188 evaluates
self.quant_config.quant_modewithout using it (Ruff B018).if self.enable_alltoall: - self.quant_config.quant_mode self.use_postquant_alltoall = (os.environ.get( "TRTLLM_MOE_POST_QUANT_ALLTOALLV", "1") == "1") self.use_low_precision_combine = (os.environ.get( "TRTLLM_MOE_USE_LOW_PRECISION_COMBINE", "0") == "1")
647-649: Shorten/standardize exception message (Ruff TRY003).Keep error messages concise; avoid long f-strings referencing large objects.
- raise ValueError( - f"unsupported quantization mode in postquant alltoall: {self.quant_config.quant_mode}" - ) + raise ValueError("Unsupported quantization mode in postquant alltoall")
708-717: Avoid magic numbers for precision; use an IntEnum.Improves readability and reduces mistakes when adding new precisions.
- assert self.has_nvfp4 or self.has_w4afp8 or self.has_fp8_qdq, "Low precision combine only supports nvfp4, w4afp8 and fp8 qdq" - precision = 0 + assert self.has_nvfp4 or self.has_w4afp8 or self.has_fp8_qdq, "Low precision combine only supports nvfp4, w4afp8 and fp8 qdq" + precision = CombinePrecision.FP8 global_scales = None if self.has_nvfp4: - precision = 1 + precision = CombinePrecision.NVFP4 global_scales = torch.ops.trtllm.calculate_nvfp4_global_scale( final_hidden_states, recv_expert_count) final_hidden_states = self.deep_ep_buffer.low_latency_combine_low_precision( precision, final_hidden_states, global_scales, deep_ep_topk_idx, deep_ep_topk_weights, deep_ep_handle)Add near the AlltoallMethodType (outside this hunk):
class CombinePrecision(IntEnum): FP8 = 0 NVFP4 = 1
311-339: Fix dtype and docstring; moe_ep_rank is a @Property
- moe_ep_rank is defined with @Property in tensorrt_llm/mapping.py — no () needed.
- Ensure torch.where doesn't upcast: make the "else" branch an int32 tensor (e.g., torch.full(..., dtype=torch.int32, device=...)); use -1 in reshapes and token_selected_slots.reshape(-1, 1).
- Add the short docstring for the DeepEP adapter describing input/output shapes and dtypes.
- Optional: prefer precomputed Mapping.slot_start/slot_end if the Mapping exposes them to avoid arithmetic on moe_ep_rank.
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (5)
cpp/tensorrt_llm/deep_ep/CMakeLists.txt(1 hunks)cpp/tensorrt_llm/kernels/cutlass_kernels/moe_gemm/moe_kernels.cu(1 hunks)cpp/tensorrt_llm/thop/moeOp.cpp(1 hunks)tensorrt_llm/_torch/modules/fused_moe/deep_ep_utils.py(2 hunks)tensorrt_llm/_torch/modules/fused_moe/fused_moe_wide_ep.py(6 hunks)
🧰 Additional context used
📓 Path-based instructions (6)
**/*.{h,hpp,hh,hxx,cpp,cxx,cc,cu,cuh}
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
**/*.{h,hpp,hh,hxx,cpp,cxx,cc,cu,cuh}: Namespace closing braces must include a trailing comment with the namespace name (e.g., '} // namespace foo').
Prefer const or constexpr variables over #define for constants.
Declare variables that are not modified after initialization as const.
Avoid magic literals in code; except for 0, nullptr, true, false. Use named constants for comparisons and logic.
Use Allman brace style for formatting.
Place the semicolon of an empty for/while loop on a new line.
Bodies of switch/while/do-while/for must be compound statements (brace-delimited), and if/else must always be followed by brace-delimited statements.
Type names (e.g., classes) must be CamelCase starting with an uppercase letter (e.g., FooBar).
Local variables, methods, and namespaces use lowerCamelCase (e.g., localFooBar).
Non-magic-number global variables that are non-static and not in an anonymous namespace must be lowerCamelCase prefixed with 'g' (e.g., gDontUseGlobalFoos).
Non-magic-number globals that are static or in an anonymous namespace use lowerCamelCase prefixed with 's' (e.g., sMutableStaticGlobal).
Locally visible static variables use lowerCamelCase with 's' prefix (e.g., static std::once_flag sFlag).
Private/protected member variables use 'm' prefix with CamelCase (e.g., mNbFooValues). Public members may omit, but 'm' is encouraged for clarity.
Constants (enums, global constants, static constants, and function-scope magic/literal constants) use uppercase SNAKE_CASE with 'k' prefix (e.g., kDIGIT_NUM).
Function-scope constants that are not magic numbers or literals are named like non-constant variables (e.g., bool const pass = a && b).
If macros are necessary, name them in UPPER_SNAKE_CASE (e.g., FOO_VERSION) and prefer constants over #define.
Use LLVM clang-format; wrap lines at a maximum of 120 columns; use '// clang-format off/on' sparingly with justification.
Use smart pointers for heap allocations; prefer unique_ptr for sole ownership, shared_ptr for shared...
Files:
cpp/tensorrt_llm/thop/moeOp.cppcpp/tensorrt_llm/kernels/cutlass_kernels/moe_gemm/moe_kernels.cu
**/*.{cpp,cxx,cc,cu,h,hpp,hh,hxx,cuh}
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
C++ filenames should be lowerCamelCase (first letter lowercase) and must be case-insensitive unique within a compilation target.
Files:
cpp/tensorrt_llm/thop/moeOp.cppcpp/tensorrt_llm/kernels/cutlass_kernels/moe_gemm/moe_kernels.cu
**/*.{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:
cpp/tensorrt_llm/thop/moeOp.cpptensorrt_llm/_torch/modules/fused_moe/deep_ep_utils.pytensorrt_llm/_torch/modules/fused_moe/fused_moe_wide_ep.pycpp/tensorrt_llm/kernels/cutlass_kernels/moe_gemm/moe_kernels.cu
**/*.{h,hpp,hh,hxx,cpp,cxx,cc}
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
**/*.{h,hpp,hh,hxx,cpp,cxx,cc}: Prefer anonymous namespaces over 'static' for internal linkage of functions.
All templates (class/function/member/static) must be instantiated at least once; non-POD classes should have private data members.
Files:
cpp/tensorrt_llm/thop/moeOp.cpp
**/*.{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:
cpp/tensorrt_llm/thop/moeOp.cpptensorrt_llm/_torch/modules/fused_moe/deep_ep_utils.pytensorrt_llm/_torch/modules/fused_moe/fused_moe_wide_ep.pycpp/tensorrt_llm/kernels/cutlass_kernels/moe_gemm/moe_kernels.cu
**/*.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/modules/fused_moe/deep_ep_utils.pytensorrt_llm/_torch/modules/fused_moe/fused_moe_wide_ep.py
🧠 Learnings (4)
📚 Learning: 2025-08-08T22:03:40.707Z
Learnt from: sklevtsov-nvidia
PR: NVIDIA/TensorRT-LLM#3294
File: cpp/tensorrt_llm/kernels/cutlass_kernels/moe_gemm/moe_kernels.cu:1198-1209
Timestamp: 2025-08-08T22:03:40.707Z
Learning: In the CUTLASS MoE kernels (cpp/tensorrt_llm/cutlass_extensions), when `layout_info.fusion` is set to `TmaWarpSpecializedGroupedGemmInput::EpilogueFusion::FINALIZE`, the `router_scales` parameter must be non-null by design. The fused finalize kernel epilogue does not perform nullptr checks and requires valid router scales to function correctly. This is an implicit contract that callers must satisfy when enabling the FINALIZE fusion mode.
Applied to files:
cpp/tensorrt_llm/kernels/cutlass_kernels/moe_gemm/moe_kernels.cu
📚 Learning: 2025-08-09T20:57:04.084Z
Learnt from: sklevtsov-nvidia
PR: NVIDIA/TensorRT-LLM#3294
File: cpp/tensorrt_llm/kernels/cutlass_kernels/moe_gemm/moe_gemm_tma_warp_specialized_input.cu:118-127
Timestamp: 2025-08-09T20:57:04.084Z
Learning: In the CUTLASS MoE finalize fusion implementation (cpp/tensorrt_llm/kernels/cutlass_kernels/moe_gemm/moe_gemm_tma_warp_specialized_input.cu), when setting `fused_finalize_epilogue.stride_final_output` with shape `(hidden_size, num_output_tokens, 1)`, the `num_rows_in_final_output` should be set to `num_output_tokens` (not `hidden_size`) because of a swap+transpose operation that maps rows of the output tensor to `hidden_size` and columns to `num_output_tokens`.
Applied to files:
cpp/tensorrt_llm/kernels/cutlass_kernels/moe_gemm/moe_kernels.cu
📚 Learning: 2025-09-19T21:28:13.751Z
Learnt from: jhaotingc
PR: NVIDIA/TensorRT-LLM#7856
File: cpp/tensorrt_llm/thop/fp8BlockScaleMoe.cpp:159-166
Timestamp: 2025-09-19T21:28:13.751Z
Learning: In TensorRT-LLM blockScaleMoe routing (cpp/tensorrt_llm/kernels/trtllmGenKernels/blockScaleMoe/runner.cu), the DeepSeek routing method performs reinterpret_cast<float*>(routingLogits) at line 89, which could cause issues if routing_logits are BF16. However, Qwen3-FP8 models use RenormalizeNaive routing method and are not affected by this dtype casting issue.
Applied to files:
cpp/tensorrt_llm/kernels/cutlass_kernels/moe_gemm/moe_kernels.cu
📚 Learning: 2025-08-19T03:35:20.866Z
Learnt from: djns99
PR: NVIDIA/TensorRT-LLM#6915
File: cpp/tensorrt_llm/kernels/cutlass_kernels/moe_gemm/moe_kernels.cu:4616-4626
Timestamp: 2025-08-19T03:35:20.866Z
Learning: In the MOE profiler TMA workspace preparation (cpp/tensorrt_llm/kernels/cutlass_kernels/moe_gemm/moe_kernels.cu), the overlapping of TMA WS regions for NONE and FINALIZE variants is deliberate design to save memory space, as confirmed by djns99. The comment "reuse the same pointers to save space" reflects this intentional behavior.
Applied to files:
cpp/tensorrt_llm/kernels/cutlass_kernels/moe_gemm/moe_kernels.cu
🧬 Code graph analysis (3)
cpp/tensorrt_llm/thop/moeOp.cpp (1)
cpp/tests/unit_tests/kernels/mixtureOfExpertsTest.cu (1)
ENABLE_FP8(216-225)
tensorrt_llm/_torch/modules/fused_moe/deep_ep_utils.py (1)
tensorrt_llm/_common.py (1)
precision(111-116)
tensorrt_llm/_torch/modules/fused_moe/fused_moe_wide_ep.py (6)
tensorrt_llm/_torch/modules/fused_moe/quantization.py (2)
FP8QDQFusedMoEMethod(493-601)FusedMoEQuantScalesW4A8(57-65)tensorrt_llm/_torch/modules/fused_moe/interface.py (4)
MoEWeightLoadingMode(16-22)has_any_quant(277-280)has_fp8_qdq(284-287)has_nvfp4(296-299)tensorrt_llm/mapping.py (1)
moe_ep_rank(360-361)tensorrt_llm/_torch/custom_ops/torch_custom_ops.py (9)
_(249-302)_(378-386)_(464-474)_(643-670)_(703-713)_(787-797)_(887-903)_(983-991)_(1024-1035)tensorrt_llm/_torch/modules/fused_moe/deep_ep_utils.py (3)
low_latency_dispatch(139-155)low_latency_dispatch_fp4(169-180)low_latency_combine_low_precision(182-197)tensorrt_llm/_torch/modules/fused_moe/fused_moe_cutlass.py (1)
has_w4afp8(181-184)
🪛 Ruff (0.13.1)
tensorrt_llm/_torch/modules/fused_moe/fused_moe_wide_ep.py
188-188: Found useless expression. Either assign it to a variable or remove it.
(B018)
647-649: Avoid specifying long messages outside the exception class
(TRY003)
🔇 Additional comments (8)
cpp/tensorrt_llm/deep_ep/CMakeLists.txt (2)
1-1: Verified DeepEP pin — commit exists and nvshmem.patch presentCommit 6e134bbd0a1e2d3c9ab58908de4eb40aa446ba17 is reachable on GitHub and the tarball contains third-party/nvshmem.patch — NVSHMEM patch flow OK.
25-30: Confirmed — "100f" is valid for CMake 3.31+
CMake 3.31+ accepts "100f" (it forwards the token to the CUDA toolchain). "f" = family‑specific (matches the major SM and allows equal/higher minor); "a" = architecture‑specific (exact SM only).cpp/tensorrt_llm/kernels/cutlass_kernels/moe_gemm/moe_kernels.cu (1)
4737-4738: Explicit instantiation for FP8-act × W4 weights (BF16 out) looks correctMatches the new runner used in moeOp and is properly gated by ENABLE_FP8 and ENABLE_BF16. No issues from a linkage/ODR perspective.
tensorrt_llm/_torch/modules/fused_moe/deep_ep_utils.py (2)
5-5: Import addition is correctOptional is needed; no concerns.
192-194: Confirm deep_ep.Buffer API parity and callsite updatesVariableLengthLowLatencyBuffer.low_latency_combine_low_precision delegates to self.buffer.low_latency_combine_low_precision; callsites found at tensorrt_llm/_torch/modules/fused_moe/fused_moe_wide_ep.py (~L715, L719). No occurrences of low_latency_combine_fp4 remain in the repo. The deep_ep extension/bindings are not in this repo — confirm the native deep_ep.Buffer implements low_latency_combine_low_precision (or provide updated bindings).
tensorrt_llm/_torch/modules/fused_moe/fused_moe_wide_ep.py (3)
20-22: Imports look correct.New imports (FP8 QDQ, W4A8 scales) match usages below.
521-524: Adapter invocation LGTM.Correctly adapts DeepEP low-latency outputs to fused_moe inputs for the non-postquant path.
610-651: Confirm Tensor.view(dtype) usage is supported in your PyTorch/buildRepo-wide search shows many .view(torch.) calls (e.g. tensorrt_llm/_torch/modules/fused_moe/fused_moe_wide_ep.py around lines 616–619 and 643–645). If your PyTorch build does not accept a dtype argument for .view, replace with .to(dtype) or an explicit zero-copy reinterpretation helper to avoid runtime errors.
Signed-off-by: Yilin Zhang <18275976+yilin-void@users.noreply.github.com>
yilin-void
force-pushed
the
deep_ep/hopper_fp8
branch
from
ee6ec2c to
38b6215
Compare
|
/bot run |
|
PR_Github #19689 [ run ] triggered by Bot |
yilin-void
changed the title
|
PR_Github #19689 [ run ] completed with state |
|
/bot run |
|
PR_Github #19767 [ run ] triggered by Bot |
|
PR_Github #19767 [ run ] completed with state |
4 participants
DeepEP diff: https://github.com/deepseek-ai/DeepEP/compare/515a311f290eb6d9592fcccfcc80c40f5123ca72...be2582ffe69b5e7d61c3bc9bf7a5316bc48261f9?expand=1
Summary by CodeRabbit
New Features
Improvements
Chores
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.