[TRTLLM-6747][feat] Merge add sparse exp and shared exp into local reduction #7369
Conversation
|
/bot run --disable-fail-fast |
📝 WalkthroughWalkthroughAdds a do_reduce flag to MNNVL MoE alltoallv combine to optionally skip reduction and return a 3D tensor. Updates the fused MoE EP path to request no reduction. Adjusts DeepseekV3 MoE forward to handle 3D routed outputs via a compiled helper that reduces along top-k before adding to shared output. Changes
Sequence Diagram(s)sequenceDiagram
autonumber
participant EP as FusedMoEWideEP
participant M as MnnvlMoe
participant D as DeepseekV3MoE.forward
participant H as _reduce_add_shared_output
rect rgb(245,248,255)
note over EP,M: Alltoallv combine
EP->>M: mnnvl_moe_alltoallv_combine(..., do_reduce=False)
M-->>EP: routed_output (shape: [tokens, top_k, hidden])
end
rect rgb(245,255,245)
note over EP,D: Finalization
EP->>D: routed_output (3D), shared_output (2D)
alt routed_output is 3D
D->>D: assert shared.numel * top_k == routed_output.numel
D->>H: reduce over dim=1 and add to shared_output
H-->>D: final_hidden_states (2D)
else routed_output is 2D
D->>D: assert sizes match
D-->>D: final_hidden_states = shared_output + routed_output
end
end
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Tip 🔌 Remote MCP (Model Context Protocol) integration is now available!Pro plan users can now connect to remote MCP servers from the Integrations page. Connect with popular remote MCPs such as Notion and Linear to add more context to your reviews and chats. ✨ 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/Issue comments)Type Other keywords and placeholders
Status, Documentation and Community
|
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🧹 Nitpick comments (1)
tensorrt_llm/_torch/models/modeling_deepseekv3.py (1)
601-611: Strengthen shape assertions to catch mismatched dims (not just numel).Assert each dimension explicitly for clearer errors and earlier detection.
- if routed_output.dim() == 3: - assert shared_output.numel( - ) * self.top_k == routed_output.numel( - ), f'unmatched tensor shape' - final_hidden_states = _reduce_add_shared_output( - routed_output, shared_output) - else: - assert shared_output.size() == routed_output.size( - ), f'unmatched tensor shape' - final_hidden_states = shared_output + routed_output + if routed_output.dim() == 3: + assert ( + routed_output.shape[0] == shared_output.shape[0] + and routed_output.shape[1] == self.top_k + and routed_output.shape[2] == shared_output.shape[1] + ), f"routed_output {tuple(routed_output.shape)} incompatible with shared_output {tuple(shared_output.shape)} and top_k={self.top_k}" + final_hidden_states = _reduce_add_shared_output(routed_output, shared_output) + else: + assert routed_output.shape == shared_output.shape, ( + f"routed_output {tuple(routed_output.shape)} != shared_output {tuple(shared_output.shape)}" + ) + final_hidden_states = shared_output + routed_output
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
💡 Knowledge Base configuration:
- MCP integration is disabled by default for public repositories
- Jira integration is disabled by default for public repositories
- Linear integration is disabled by default for public repositories
You can enable these sources in your CodeRabbit configuration.
📒 Files selected for processing (3)
tensorrt_llm/_mnnvl_utils.py(2 hunks)tensorrt_llm/_torch/models/modeling_deepseekv3.py(2 hunks)tensorrt_llm/_torch/modules/fused_moe/fused_moe_wide_ep.py(1 hunks)
🧰 Additional context used
📓 Path-based instructions (2)
**/*.{cpp,cc,cxx,cu,h,hpp,hh,hxx,cuh,py}
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
**/*.{cpp,cc,cxx,cu,h,hpp,hh,hxx,cuh,py}: Use spaces only; no tabs; indent with 4 spaces
Prepend NVIDIA copyright header (current year) to all source files (.cpp, .h, .cu, .py, etc.)
Files:
tensorrt_llm/_mnnvl_utils.pytensorrt_llm/_torch/modules/fused_moe/fused_moe_wide_ep.pytensorrt_llm/_torch/models/modeling_deepseekv3.py
**/*.py
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
**/*.py: Python code must target Python 3.8+
Indent Python with 4 spaces; no tabs
Preserve module namespaces when importing: from package.subpackage import foo; then call foo.SomeClass() instead of importing the class directly
Python naming: files snake_case; classes PascalCase; functions/methods snake_case; locals snake_case (prefix k_ when starting with a number); globals UPPER_SNAKE_CASE with G_ prefix; constants UPPER_SNAKE_CASE
Avoid shadowing outer-scope variables; initialize all externally visible members in init
Prefer docstrings for interfaces used outside a file; limit comments to function-internal or file-local interfaces
Use Google-style docstrings for classes and functions; document attributes/variables inline so Sphinx can render them
Avoid reflection when simpler alternatives exist; prefer explicit parameters and return dicts over locals()/dynamic tricks
In try/except, catch the narrowest exceptions possible; keep try bodies minimal and use else for the main logic when doing duck-typing checks
Files:
tensorrt_llm/_mnnvl_utils.pytensorrt_llm/_torch/modules/fused_moe/fused_moe_wide_ep.pytensorrt_llm/_torch/models/modeling_deepseekv3.py
🧠 Learnings (2)
📓 Common learnings
Learnt from: timlee0212
PR: NVIDIA/TensorRT-LLM#6886
File: tensorrt_llm/_torch/models/modeling_deepseekv3.py:0-0
Timestamp: 2025-08-14T06:36:40.701Z
Learning: In DeepSeek V3 model (tensorrt_llm/_torch/models/modeling_deepseekv3.py), the disagreement between AllReduce.__init__ guard and _compute_mlp_tp_size logic for MNNVL usage is expected by design. The AllReduce component and MLP TP-size computation intentionally use different criteria for MNNVL availability decisions.
📚 Learning: 2025-08-14T06:36:40.701Z
Learnt from: timlee0212
PR: NVIDIA/TensorRT-LLM#6886
File: tensorrt_llm/_torch/models/modeling_deepseekv3.py:0-0
Timestamp: 2025-08-14T06:36:40.701Z
Learning: In DeepSeek V3 model (tensorrt_llm/_torch/models/modeling_deepseekv3.py), the disagreement between AllReduce.__init__ guard and _compute_mlp_tp_size logic for MNNVL usage is expected by design. The AllReduce component and MLP TP-size computation intentionally use different criteria for MNNVL availability decisions.
Applied to files:
tensorrt_llm/_mnnvl_utils.pytensorrt_llm/_torch/modules/fused_moe/fused_moe_wide_ep.pytensorrt_llm/_torch/models/modeling_deepseekv3.py
🧬 Code graph analysis (1)
tensorrt_llm/_torch/models/modeling_deepseekv3.py (3)
tensorrt_llm/functional.py (1)
sum(3253-3275)tensorrt_llm/_utils.py (1)
numel(959-960)cpp/tensorrt_llm/kernels/trtllmGenKernels/blockScaleMoe/runner.h (1)
top_k(221-221)
⏰ 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 (2)
tensorrt_llm/_mnnvl_utils.py (1)
618-622: I’ve added scripts to locate the CUDA/C++ implementation and inspect anyuse_low_precisionlogic in the operator. Once we see whether the kernel itself handles reduction under low-precision combine, we can confirm if the host-side sum is necessary or would double-accumulate.tensorrt_llm/_torch/modules/fused_moe/fused_moe_wide_ep.py (1)
909-911: No other alltoall_combine callers or 2D shape expectations found. rg confirms only the call in fused_moe_wide_ep.py and no downstream view/reshape checks expect 2D.
|
PR_Github #16972 [ run ] triggered by Bot |
zongfeijing
changed the title
|
PR_Github #16972 [ run ] completed with state |
|
/bot run --disable-fail-fast |
|
PR_Github #16989 [ run ] triggered by Bot |
zongfeijing
force-pushed
the
user/zongfeij/fuse_reduce_add
branch
from
987b9e3 to
a6b5917
Compare
|
/bot run |
|
PR_Github #17002 [ run ] triggered by Bot |
|
PR_Github #16989 [ run ] completed with state |
|
PR_Github #17002 [ run ] completed with state |
zongfeijing
force-pushed
the
user/zongfeij/fuse_reduce_add
branch
2 times, most recently
from
30eb5f8 to
47f96a9
Compare
|
/bot run |
|
PR_Github #17052 [ run ] triggered by Bot |
|
PR_Github #17052 [ run ] completed with state |
|
/bot run --disable-fail-fast |
|
PR_Github #17060 [ run ] triggered by Bot |
|
PR_Github #17060 [ run ] completed with state |
|
/bot run |
|
PR_Github #17073 [ run ] triggered by Bot |
|
PR_Github #17073 [ run ] completed with state |
…reduction Signed-off-by: Zongfei Jing <[email protected]>
Signed-off-by: Zongfei Jing <[email protected]>
Signed-off-by: Zongfei Jing <[email protected]>
zongfeijing
force-pushed
the
user/zongfeij/fuse_reduce_add
branch
from
47f96a9 to
32994db
Compare
|
/bot run --disable-fail-fast |
|
PR_Github #17078 [ run ] triggered by Bot |
|
PR_Github #17078 [ run ] completed with state |
|
/bot run |
|
PR_Github #17092 [ run ] triggered by Bot |
|
PR_Github #17092 [ run ] completed with state |
|
/bot run |
|
PR_Github #17113 [ run ] triggered by Bot |
|
PR_Github #17113 [ run ] completed with state |
|
/bot skip "The test cases have passed, although the CI pipeline failed due to an known issue." |
GitHub Bot Help
Provide a user friendly way for developers to interact with a Jenkins server. Run See details below for each supported subcommand.
Launch build/test pipelines. All previously running jobs will be killed.
kill
Kill all running builds associated with pull request. skip
Skip testing for latest commit on pull request. reuse-pipeline
Reuse 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. |
|
/bot skip --comment "The test cases have passed, although the CI pipeline failed due to an known issue." |
|
PR_Github #17124 [ skip ] triggered by Bot |
|
PR_Github #17124 [ skip ] completed with state |
…duction (NVIDIA#7369) Signed-off-by: Zongfei Jing <[email protected]>
6 participants
…reduction
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.