[https://nvbugs/5501557][fix] Fix nemotron build error #7644
Conversation
* Root cause is Nemotron-nas model includes some no-op attention layers. Signed-off-by: Wanli Jiang <[email protected]>
📝 WalkthroughWalkthroughUpdated inflight batching to compute window size grouping using the count of local KV-head layers, deriving numLayers from numKvHeadsPerLayer.size(), and to call KVCacheManager::groupLayersByWindowSize(maxAttentionWindowVec, numLayers) instead of BaseKVCacheManager::groupLayersByWindowSize(..., modelConfig.getNbLayers()). Changes
Sequence Diagram(s)sequenceDiagram
autonumber
actor Caller as InflightBatching
participant Cfg as modelConfig
participant KV as KVCacheManager
Caller->>Caller: numLayers = numKvHeadsPerLayer.size()
Caller->>KV: groupLayersByWindowSize(maxAttentionWindowVec, numLayers)
KV-->>Caller: windowSizeLayers
opt Prior flow (reference)
Caller->>Cfg: getNbLayers()
Cfg-->>Caller: totalModelLayers
Caller->>KV: groupLayersByWindowSize(maxAttentionWindowVec, totalModelLayers) note right of KV: Previously via BaseKVCacheManager
end
Estimated code review effort🎯 2 (Simple) | ⏱️ ~10 minutes 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.
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. Pre-merge checks (1 passed, 2 warnings)❌ 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
🧹 Nitpick comments (1)
cpp/tensorrt_llm/batch_manager/trtGptModelInflightBatching.cpp (1)
272-276: Avoid the temporary vector; compute numLayers without allocation.Minor allocation you can skip; distance on the iterator range suffices.
- auto [numKvHeadsPerLayerBegin, numKvHeadsPerLayerEnd] = modelConfig.getNumKvHeadsPerLayerLocalRange( - worldConfig.getPipelineParallelism(), worldConfig.getPipelineParallelRank(), isCrossAttention); - auto numKvHeadsPerLayer = std::vector<SizeType32>(numKvHeadsPerLayerBegin, numKvHeadsPerLayerEnd); - auto const numLayers = static_cast<SizeType32>(numKvHeadsPerLayer.size()); + auto [numKvHeadsPerLayerBegin, numKvHeadsPerLayerEnd] = modelConfig.getNumKvHeadsPerLayerLocalRange( + worldConfig.getPipelineParallelism(), worldConfig.getPipelineParallelRank(), isCrossAttention); + auto const numLayers = static_cast<SizeType32>( + std::distance(numKvHeadsPerLayerBegin, numKvHeadsPerLayerEnd)); auto const windowSizeLayers = KVCacheManager::groupLayersByWindowSize(maxAttentionWindowVec, numLayers);
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (1)
cpp/tensorrt_llm/batch_manager/trtGptModelInflightBatching.cpp(1 hunks)
🧰 Additional context used
📓 Path-based instructions (5)
**/*.{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/batch_manager/trtGptModelInflightBatching.cpp
**/*.{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/batch_manager/trtGptModelInflightBatching.cpp
**/*.{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/batch_manager/trtGptModelInflightBatching.cpp
**/*.{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/batch_manager/trtGptModelInflightBatching.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/batch_manager/trtGptModelInflightBatching.cpp
🧠 Learnings (5)
📓 Common learnings
Learnt from: thorjohnsen
PR: NVIDIA/TensorRT-LLM#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.
📚 Learning: 2025-08-14T21:04:50.248Z
Learnt from: thorjohnsen
PR: NVIDIA/TensorRT-LLM#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/tensorrt_llm/batch_manager/trtGptModelInflightBatching.cpp
📚 Learning: 2025-08-20T06:56:02.889Z
Learnt from: eopXD
PR: NVIDIA/TensorRT-LLM#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/tensorrt_llm/batch_manager/trtGptModelInflightBatching.cpp
📚 Learning: 2025-08-15T06:46:54.897Z
Learnt from: eopXD
PR: NVIDIA/TensorRT-LLM#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/batch_manager/trtGptModelInflightBatching.cpp
📚 Learning: 2025-08-21T09:41:49.347Z
Learnt from: eopXD
PR: NVIDIA/TensorRT-LLM#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/batch_manager/trtGptModelInflightBatching.cpp
🧬 Code graph analysis (1)
cpp/tensorrt_llm/batch_manager/trtGptModelInflightBatching.cpp (1)
cpp/tensorrt_llm/batch_manager/kvCacheManager.cpp (2)
groupLayersByWindowSize(2227-2242)groupLayersByWindowSize(2227-2228)
⏰ 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 (3)
cpp/tensorrt_llm/batch_manager/trtGptModelInflightBatching.cpp (3)
275-276: Correct window-size grouping using local KV-head layers.Deriving numLayers from the local KV-head range and grouping via KVCacheManager fixes the mismatch when models include no-op attention layers. Looks good.
661-662: Consistent local-layer grouping for KV-cache sizing.Using the local numLayers here aligns block computation with the actual managed layers on this PP rank. Good change.
661-663: Remove sanity-check: no instances of groupLayersByWindowSize called with modelConfig.getNbLayers()
Wanli-Jiang
force-pushed
the
user/williamj/fix-nemotron-trt
branch
from
9324575 to
b6d8016
Compare
|
/bot run |
|
PR_Github #18203 [ run ] triggered by Bot |
|
PR_Github #18203 [ run ] completed with state |
| auto windowSizeLayers | ||
| = BaseKVCacheManager::groupLayersByWindowSize(maxAttentionWindowVec, modelConfig.getNbLayers()); | ||
| auto const numLayers = static_cast<SizeType32>(numKvHeadsPerLayer.size()); | ||
| auto const windowSizeLayers = KVCacheManager::groupLayersByWindowSize(maxAttentionWindowVec, numLayers); |
There was a problem hiding this comment.
Here's my issue with groupLayersByWindowSize: It doesn't take PP into account and leads to incorrect memory computation.
https://github.com/NVIDIA/TensorRT-LLM/blob/main/cpp/tensorrt_llm/batch_manager/kvCacheManager.cpp#L2177-L2192
Let's take an example of a model with 4 layers and maxAttentionWindowVec = {2, 3, 4, 5}
Case1: PP=1
ppRank 0; layers = {l0, l1, l2, l3}
Input to groupLayersByWindowSize: maxAttentionWindowVec={2, 3, 4, 5}, numLayers=4.
Output from groupLayersByWindowSize: {2:{0}, 3:{1}, 4:{2}, 5{3}}.
Case1: PP=2
ppRank 0; layers = {l0, l1}
Input to groupLayersByWindowSize: maxAttentionWindowVec={2, 3, 4, 5}, numLayers=2.
Output from groupLayersByWindowSize: {2:{0}, 3:{1}}.
ppRank 1; layers = {l2, l3}
Input to groupLayersByWindowSize: maxAttentionWindowVec={2, 3, 4, 5}, numLayers=2.
Output from groupLayersByWindowSize: {2:{0}, 3:{1}}. // THIS SHOULD'VE BEEN {4{0}, 5{1}} WHERE 0,1 ARE LOCAL LAYER IDS.
Problem is it's just taking numLayers as input. Instead it should take a layer range (where layerIds are global) to get the math right. [OR] maxAttentionWindowVec must be different for each PP rank.
Please let me know if I'm missing something.
| auto const windowSizeLayers = KVCacheManager::groupLayersByWindowSize(maxAttentionWindowVec, numLayers); | ||
| std::map<SizeType32, SizeType32> cacheSizeBytesPerTokenPerWindow; | ||
| for (auto const& [windowSize, managedLayers] : windowSizeLayers) | ||
| { |
There was a problem hiding this comment.
There's a similar issue with calculateCacheSizePerTokenForSingleWindowSize which further calls getNumKvForGivenLayers.
To my understanding, getNumKvForGivenLayers expects global layer ids. Currently, we pass in local layer ids when using PP.
|
/bot skip --comment "Verified by dev testing" |
|
PR_Github #18252 [ skip ] triggered by Bot |
|
PR_Github #18252 [ skip ] completed with state |
|
/bot run |
|
PR_Github #18281 [ run ] triggered by Bot |
|
Closing since this was manually merged to the branch |
|
PR_Github #18281 [ run ] completed with state |
4 participants
Summary by CodeRabbit
Bug Fixes
Performance
Stability
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.