[https://nvbugs/5481434][feat] cherry-pick fix to reuse pytorch memory segments occupied by cudagraph #7747
Conversation
… by cudagraph pool (NVIDIA#7457) Signed-off-by: Hui Gao <huig@nvidia.com>
📝 WalkthroughWalkthroughIntroduces buffer pooling in DeepGemmFusedMoE: adds class-level caches for graph-capture and runtime, refactors get_workspace to reuse/prefer larger preallocated CUDA tensors via helpers (select_buffer_with_more_elements, get_empty), and replaces direct torch.empty allocations for workspace_0/1/sf with cached buffers. Adds math import. Changes
Sequence Diagram(s)sequenceDiagram
participant Caller
participant DeepGemmFusedMoE
participant get_workspace
participant get_empty
participant GraphPool as Graph Pool (cache)
participant RuntimePool as Runtime Pool (cache)
Caller->>DeepGemmFusedMoE: forward(...)
DeepGemmFusedMoE->>get_workspace: request workspaces
rect rgba(220,235,245,0.4)
note right of get_workspace: For each of workspace_0, workspace_1, workspace_sf
get_workspace->>get_empty: (shape, dtype, cache_name)
alt Capturing graph
get_empty->>GraphPool: find suitable buffer
opt none suitable
get_empty->>RuntimePool: select larger available buffer
opt none suitable
get_empty-->>get_empty: allocate new CUDA tensor
end
get_empty->>GraphPool: promote/store buffer
end
else Runtime
get_empty->>RuntimePool: find/select suitable buffer
opt none suitable
get_empty-->>get_empty: allocate new CUDA tensor
get_empty->>RuntimePool: store buffer
end
end
get_empty-->>get_workspace: return buffer
end
get_workspace-->>DeepGemmFusedMoE: workspace_0/1/sf
DeepGemmFusedMoE-->>Caller: continue compute with reused buffers
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20 minutes ✨ Finishing touches
🧪 Generate unit tests
Tip 👮 Agentic pre-merge checks are now available in preview!Pro plan users can now enable pre-merge checks in their settings to enforce checklists before merging PRs.
Please see the documentation for more information. Example: reviews:
pre_merge_checks:
custom_checks:
- name: "Undocumented Breaking Changes"
mode: "warning"
instructions: |
Pass/fail criteria: All breaking changes to public APIs, CLI flags, environment variables, configuration keys, database schemas, or HTTP/GraphQL endpoints must be documented in the "Breaking Change" section of the PR description and in CHANGELOG.md. Exclude purely internal or private changes (e.g., code not exported from package entry points or explicitly marked as internal).Please share your feedback with us on this Discord post. 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 Pre-merge checks❌ Failed checks (2 warnings)
✅ Passed checks (1 passed)
|
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 3
🧹 Nitpick comments (3)
tensorrt_llm/_torch/modules/fused_moe/fused_moe_deepgemm.py (3)
419-482: Optional: prefer “best fit” in graph pool to reduce waste; add pool reset helper.
- Iterate graph candidates and pick the smallest adequate buffer (min by numel) to reduce fragmentation.
- Add a classmethod to clear pools (useful between runs/tests).
Example (outside the changed hunk, for clarity):
# Pick smallest adequate candidate graph_buffer = min( (b for b in candidate_buffers if b.dtype == dtype and b.numel() >= numel_like), key=lambda t: t.numel(), default=None, ) @classmethod def reset_buffer_pools(cls) -> None: cls.allocated_buffer_in_runtime.clear() cls.allocated_buffer_in_graph_pool.clear()
489-496: Consider torch.empty over torch.zeros if downstream kernels fully overwrite.Zero-filling costs time/bandwidth in hot paths. If all elements are later written by kernels, prefer empty for speed.
Apply this diff if safe:
- workspace_0 = get_empty((num_experts * m_max * fp8_dim, ), - dtype=torch.float8_e4m3fn, - cache_name='workspace_0') + workspace_0 = get_empty((num_experts * m_max * fp8_dim, ), + dtype=torch.float8_e4m3fn, + cache_name='workspace_0') @@ - workspace_1 = get_empty( + workspace_1 = get_empty( (num_experts * m_max * max(intermediate_size * 2, hidden_size), ), dtype=torch.bfloat16, cache_name='workspace_1')(Note: change is conceptual; get_empty currently zeros. Switch new_buffer to torch.empty if correctness permits.)
502-506: Same note for scaling-factor workspace.If fully written, consider empty allocation to avoid memset during capture/warmup.
- (num_experts * (scale_k_padded // 4) * m_padded, ), + (num_experts * (scale_k_padded // 4) * m_padded, ), dtype=torch.int32, cache_name='workspace_sf')(Adopt empty in get_empty if safe.)
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (1)
tensorrt_llm/_torch/modules/fused_moe/fused_moe_deepgemm.py(3 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:
tensorrt_llm/_torch/modules/fused_moe/fused_moe_deepgemm.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:
tensorrt_llm/_torch/modules/fused_moe/fused_moe_deepgemm.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:
tensorrt_llm/_torch/modules/fused_moe/fused_moe_deepgemm.py
🧠 Learnings (1)
📚 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:
tensorrt_llm/_torch/modules/fused_moe/fused_moe_deepgemm.py
🧬 Code graph analysis (1)
tensorrt_llm/_torch/modules/fused_moe/fused_moe_deepgemm.py (2)
tensorrt_llm/_torch/auto_deploy/compile/backends/torch_cudagraph.py (1)
capture_graph(76-119)tensorrt_llm/quantization/utils/fp8_utils.py (2)
align(24-25)ceil_div(10-21)
🪛 Ruff (0.12.2)
tensorrt_llm/_torch/modules/fused_moe/fused_moe_deepgemm.py
370-370: Mutable class attributes should be annotated with typing.ClassVar
(RUF012)
371-371: Mutable class attributes should be annotated with typing.ClassVar
(RUF012)
⏰ 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 (1)
tensorrt_llm/_torch/modules/fused_moe/fused_moe_deepgemm.py (1)
1-1: LGTM: import math is justified.Used for math.prod in get_empty.
|
/bot run --disable-fail-fast |
|
PR_Github #18692 [ run ] triggered by Bot |
|
PR_Github #18692 [ run ] completed with state |
|
/bot run --disable-fail-fast |
|
PR_Github #18734 [ run ] triggered by Bot |
|
PR_Github #18734 [ run ] completed with state |
Signed-off-by: Hui Gao <huig@nvidia.com>
|
/bot run --disable-fail-fast |
|
PR_Github #19148 [ run ] triggered by Bot |
|
PR_Github #19148 [ run ] completed with state |
|
/bot run --stage-list="GB200-4_GPUs-PyTorch-1" --only-multi-gpu-test |
|
/bot run --stage-list="GB200-4_GPUs-PyTorch-1" |
|
PR_Github #19220 [ run ] triggered by Bot |
|
PR_Github #19221 [ run ] triggered by Bot |
|
PR_Github #19220 [ run ] completed with state |
|
PR_Github #19221 [ run ] completed with state |
|
/bot skip --comment "the timed out stage passed in last round" |
|
PR_Github #19241 [ skip ] triggered by Bot |
|
PR_Github #19241 [ skip ] completed with state |
…y segments occupied by cudagraph (NVIDIA#7747) Signed-off-by: Hui Gao <huig@nvidia.com>
…y segments occupied by cudagraph (NVIDIA#7747) Signed-off-by: Hui Gao <huig@nvidia.com>
…y segments occupied by cudagraph (NVIDIA#7747) Signed-off-by: Hui Gao <huig@nvidia.com>
4 participants
#7457
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.