[https://nvbugs/5451373][fix] : Fix the accuracy issue when using FP8 context MLA #6881
Conversation
📝 WalkthroughWalkthroughRecomputes and assigns FP8 BMM1 scaling and additional FP8 quant/dequant scales in the MLA-enabled attention path on the host, and extends the MLA kernel to accept/produce FP8 bmm scales and additional quant/dequant scale parameters; host invocations updated to pass the new parameters. No public API signature changes. Changes
Sequence Diagram(s)sequenceDiagram
participant Caller
participant AttentionOp
participant CUDA as Kernel
Note over AttentionOp: Host prepares MLA params
Caller->>AttentionOp: enqueueContext(...)
AttentionOp->>AttentionOp: compute host_bmm1_scale\nset quant/dequant scale fields
AttentionOp->>CUDA: launch applyMLARopeAndAssignQKVKernelOptContext(..., bmm1_scale, bmm2_scale, quant_scale_o, quant_scale_kv, dequant_scale_q, dequant_scale_kv, host_bmm1_scale)
CUDA->>CUDA: (first thread) if FP8: compute bmm1_scale, bmm2_scale
CUDA-->>AttentionOp: kernel returns (bmm1_scale/bmm2_scale available on device)
AttentionOp-->>Caller: return status
Estimated code review effort🎯 4 (Complex) | ⏱️ ~35 minutes Possibly related PRs
Suggested labels
Suggested reviewers
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: 1
🔭 Outside diff range comments (1)
cpp/tensorrt_llm/common/attentionOp.cpp (1)
768-770: Workspace under-allocation for FP8 context MLA BMM scalesIn getWorkspaceSizeForContext(), the BMM scale buffers are only reserved when mFP8ContextFMHA is true. The new MLA code path writes two floats to fmha_bmm1_scale_ptr (and uses fmha_bmm2_scale_ptr). If mFP8ContextMLA is enabled without mFP8ContextFMHA, the workspace will be under-sized, leading to out-of-bounds writes.
Update the allocations to include mFP8ContextMLA:
// Before: size_t const fmha_bmm1_scale_size = mFP8ContextFMHA ? sizeof(float) * 2 : 0; size_t const fmha_bmm2_scale_size = mFP8ContextFMHA ? sizeof(float) : 0; // After: size_t const fmha_bmm1_scale_size = (mFP8ContextFMHA || mFP8ContextMLA) ? sizeof(float) * 2 : 0; size_t const fmha_bmm2_scale_size = (mFP8ContextFMHA || mFP8ContextMLA) ? sizeof(float) : 0;Follow-up: Please verify other workspace size helpers do not miss similar FP8ContextMLA cases. I can help run a repo-wide scan if needed.
🧹 Nitpick comments (1)
cpp/tensorrt_llm/common/attentionOp.cpp (1)
1777-1782: Prefer sanitized memcpy and remove TODO-like comment
- Use tensorrt_llm::common::cudaMemcpyAsyncSanitized or at least TLLM_CUDA_CHECK for consistency and immediate error surfacing.
- The “Need to figure out the correct bmm1 scales” comment should be removed once the computation is used (see previous suggestion).
📜 Review details
Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (1)
cpp/tensorrt_llm/common/attentionOp.cpp(1 hunks)
🧰 Additional context used
📓 Path-based instructions (4)
**/*.{cpp,cxx,cc,cu,h,hpp,hxx,hh,cuh}
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
**/*.{cpp,cxx,cc,cu,h,hpp,hxx,hh,cuh}: In C++, close namespaces with a comment naming the namespace (e.g., } // namespace foo)
Prefer const/constexpr variables over #define for constants
Declare variables const if not modified after initialization
Use Allman brace style in C++
C++ filenames use lowerCamelCase and must be case-insensitively unique within a build target
C++ type names use UpperCamelCase
Local variables, methods, and namespaces use lowerCamelCase
Global non-static variables not in anonymous namespace use gPrefix lowerCamelCase (e.g., gExample)
Static globals or globals in anonymous namespaces use sPrefix lowerCamelCase
Locally visible static variables start with 's' (e.g., static std::once_flag sFlag;)
Member variables use mPrefix lowerCamelCase; public members may omit but are encouraged to use 'm'
Constants (enums, global/static/function-scope magic numbers) use kPREFIXED_UPPER_SNAKE (e.g., kDIGIT_NUM)
If macros are unavoidable, use UPPER_SNAKE_CASE (prefer constants over #define)
Constructor parameter that conflicts with a public member name gets trailing underscore (foo_)
Literal suffixes should be uppercase (e.g., 1234L not 1234l)
C++: use spaces only; indent 4 spaces
Run clang-format (LLVM style) before submitting; wrap lines at 120 characters
If formatting must be bypassed, use // clang-format off/on around the section
Prefer smart pointers; use unique_ptr for sole ownership, shared_ptr for shared; weak_ptr only in exceptional cases
Do not use deprecated pre-C++11 smart pointers
Use C++ style comments; avoid C comments except special inline cases; prefer // single-line
Capitalize and punctuate full-sentence comments
Follow Doxygen rules: use //! for comments and //!< for members in C++
Disable code with #if/#endif and mnemonic conditions; avoid commented-out code; avoid dead code
Do not throw exceptions across library boundaries
Use least-forceful casts; avoid removing const/volatile; avoid C-style and functional casts (except constructors); p...
Files:
cpp/tensorrt_llm/common/attentionOp.cpp
**/*.{cpp,cxx,cc,cu}
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
**/*.{cpp,cxx,cc,cu}: Avoid literal values except for 0, nullptr, true, false; use named constexpr for other literals
Place semicolon of empty for/while loop on a new line
Always use brace-delimited bodies for switch/while/do-for/if/else
Use inline C comments in argument lists when parameter meaning is unclear (e.g., /* checkForErrors = */ false)
Do not use assignment in subexpressions (e.g., if (x = y) ... is forbidden)
Switch on enums should enumerate all values and omit default to catch new values at compile time
Structure switch statements; prohibit fallthrough except between empty cases; each case ends with break or throw; return at end of case not allowed; put break inside braces for compound case
Prefer anonymous namespaces over static for internal linkage of functions
Every defined function must be called at least once (no unused methods)
Files:
cpp/tensorrt_llm/common/attentionOp.cpp
**/*.{h,hpp,hxx,hh,cuh,cpp,cxx,cc,cu}
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
Parameter names must be consistent between declarations and definitions
Files:
cpp/tensorrt_llm/common/attentionOp.cpp
**/*.{cpp,cxx,cc,cu,h,hpp,hxx,hh,cuh,py}
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
Prepend NVIDIA copyright header (current year) to all source files
Files:
cpp/tensorrt_llm/common/attentionOp.cpp
peaceh-nv
force-pushed
the
fix-deepseek-lite-acc
branch
from
29172af to
bff138f
Compare
… context MLA Signed-off-by: peaceh <103117813+peaceh-nv@users.noreply.github.com>
peaceh-nv
force-pushed
the
fix-deepseek-lite-acc
branch
from
bff138f to
d19422e
Compare
peaceh-nv
changed the title
|
/bot run |
|
PR_Github #15260 [ run ] triggered by Bot |
There was a problem hiding this comment.
Actionable comments posted: 2
🧹 Nitpick comments (2)
cpp/tensorrt_llm/kernels/mlaKernels.cu (2)
235-260: Const-correctness and constant naming per guidelines
- Make local temporaries const.
- Rename kLog2e to kLOG2E to follow kPREFIXED_UPPER_SNAKE style for constants (coding_guidelines).
Apply:
- if (cache_type == KvCacheDataType::FP8) + if (cache_type == KvCacheDataType::FP8) { - float dequant_scale_q_val = dequant_scale_q ? dequant_scale_q[0] : 1.f; - float dequant_scale_kv_val = dequant_scale_kv ? dequant_scale_kv[0] : 1.f; - float quant_scale_o_val = quant_scale_o ? quant_scale_o[0] : 1.f; + float const dequant_scale_q_val = dequant_scale_q ? dequant_scale_q[0] : 1.f; + float const dequant_scale_kv_val = dequant_scale_kv ? dequant_scale_kv[0] : 1.f; + float const quant_scale_o_val = quant_scale_o ? quant_scale_o[0] : 1.f; if (bmm1_scale) { // The scale prepared for log2 optimization. - constexpr float kLog2e = 1.4426950408889634074f; + constexpr float kLOG2E = 1.4426950408889634074f; // The scale after fmha bmm1. - float bmm1_scale_val = dequant_scale_q_val * dequant_scale_kv_val * host_bmm1_scale; + float const bmm1_scale_val = dequant_scale_q_val * dequant_scale_kv_val * host_bmm1_scale; bmm1_scale[0] = bmm1_scale_val; - bmm1_scale[1] = bmm1_scale_val * kLog2e; + bmm1_scale[1] = bmm1_scale_val * kLOG2E; } if (bmm2_scale) { // The scale after fmha bmm2. - bmm2_scale[0] = quant_scale_o_val * dequant_scale_kv_val; + bmm2_scale[0] = quant_scale_o_val * dequant_scale_kv_val; } }
1-15: Update copyright yearPer coding guidelines, prepend NVIDIA copyright header (current year). The file shows 2019-2023; please update to include 2025 (e.g., “2019-2025”).
📜 Review details
Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (2)
cpp/tensorrt_llm/common/attentionOp.cpp(1 hunks)cpp/tensorrt_llm/kernels/mlaKernels.cu(3 hunks)
🚧 Files skipped from review as they are similar to previous changes (1)
- cpp/tensorrt_llm/common/attentionOp.cpp
🧰 Additional context used
📓 Path-based instructions (4)
**/*.{cpp,cxx,cc,cu,h,hpp,hxx,hh,cuh}
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
**/*.{cpp,cxx,cc,cu,h,hpp,hxx,hh,cuh}: In C++, close namespaces with a comment naming the namespace (e.g., } // namespace foo)
Prefer const/constexpr variables over #define for constants
Declare variables const if not modified after initialization
Use Allman brace style in C++
C++ filenames use lowerCamelCase and must be case-insensitively unique within a build target
C++ type names use UpperCamelCase
Local variables, methods, and namespaces use lowerCamelCase
Global non-static variables not in anonymous namespace use gPrefix lowerCamelCase (e.g., gExample)
Static globals or globals in anonymous namespaces use sPrefix lowerCamelCase
Locally visible static variables start with 's' (e.g., static std::once_flag sFlag;)
Member variables use mPrefix lowerCamelCase; public members may omit but are encouraged to use 'm'
Constants (enums, global/static/function-scope magic numbers) use kPREFIXED_UPPER_SNAKE (e.g., kDIGIT_NUM)
If macros are unavoidable, use UPPER_SNAKE_CASE (prefer constants over #define)
Constructor parameter that conflicts with a public member name gets trailing underscore (foo_)
Literal suffixes should be uppercase (e.g., 1234L not 1234l)
C++: use spaces only; indent 4 spaces
Run clang-format (LLVM style) before submitting; wrap lines at 120 characters
If formatting must be bypassed, use // clang-format off/on around the section
Prefer smart pointers; use unique_ptr for sole ownership, shared_ptr for shared; weak_ptr only in exceptional cases
Do not use deprecated pre-C++11 smart pointers
Use C++ style comments; avoid C comments except special inline cases; prefer // single-line
Capitalize and punctuate full-sentence comments
Follow Doxygen rules: use //! for comments and //!< for members in C++
Disable code with #if/#endif and mnemonic conditions; avoid commented-out code; avoid dead code
Do not throw exceptions across library boundaries
Use least-forceful casts; avoid removing const/volatile; avoid C-style and functional casts (except constructors); p...
Files:
cpp/tensorrt_llm/kernels/mlaKernels.cu
**/*.{cpp,cxx,cc,cu}
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
**/*.{cpp,cxx,cc,cu}: Avoid literal values except for 0, nullptr, true, false; use named constexpr for other literals
Place semicolon of empty for/while loop on a new line
Always use brace-delimited bodies for switch/while/do-for/if/else
Use inline C comments in argument lists when parameter meaning is unclear (e.g., /* checkForErrors = */ false)
Do not use assignment in subexpressions (e.g., if (x = y) ... is forbidden)
Switch on enums should enumerate all values and omit default to catch new values at compile time
Structure switch statements; prohibit fallthrough except between empty cases; each case ends with break or throw; return at end of case not allowed; put break inside braces for compound case
Prefer anonymous namespaces over static for internal linkage of functions
Every defined function must be called at least once (no unused methods)
Files:
cpp/tensorrt_llm/kernels/mlaKernels.cu
**/*.{h,hpp,hxx,hh,cuh,cpp,cxx,cc,cu}
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
Parameter names must be consistent between declarations and definitions
Files:
cpp/tensorrt_llm/kernels/mlaKernels.cu
**/*.{cpp,cxx,cc,cu,h,hpp,hxx,hh,cuh,py}
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
Prepend NVIDIA copyright header (current year) to all source files
Files:
cpp/tensorrt_llm/kernels/mlaKernels.cu
⏰ 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)
cpp/tensorrt_llm/kernels/mlaKernels.cu (2)
949-953: Host invocation updated correctlyThe new arguments order matches the kernel signature and types. No issues spotted with grid/block dimensions or parameter wiring.
949-953: Verified — kernel call sites updated and bmm scale buffers sized correctly
Summary: I checked the repo — the kernel is called only with the new signature and bmm1_scale/bmm2_scale workspace sizes allocate 2 and 1 floats respectively; host_bmm1_scale is present.
Evidence (key locations):
- Kernel signature: cpp/tensorrt_llm/kernels/mlaKernels.cu:207–213 (expects float* bmm1_scale, float* bmm2_scale, float host_bmm1_scale).
- Kernel call sites: cpp/tensorrt_llm/kernels/mlaKernels.cu:949 and cpp/tensorrt_llm/kernels/mlaKernels.cu:1024 (both pass params.bmm1_scale, params.bmm2_scale, params.host_bmm1_scale).
- bmm1_scale writes two values (indices 0 and 1): cpp/tensorrt_llm/kernels/mlaKernels.cu:249–251 and 430–432.
- Workspaces allocate correct sizes: cpp/tensorrt_llm/kernels/decoderMaskedMultiheadAttention/decoderXQAImplCommon.h:276–281 (bmm1_scale_size = sizeof(float) * 2; bmm2_scale_size = sizeof(float)); attentionOp.cpp also sets/assigns mla_bmm1_scale_ptr and mla_bmm2_scale_ptr (see the workspace/pointer assignments around cpp/tensorrt_llm/common/attentionOp.cpp:957–972 and usage at 1081–1084).
Conclusion: No lingering call sites or sizing issues found — no changes required.
|
/bot run |
|
PR_Github #15357 [ run ] triggered by Bot |
|
/bot run |
|
PR_Github #15362 [ run ] triggered by Bot |
|
PR_Github #15357 [ run ] completed with state |
|
PR_Github #15362 [ run ] completed with state |
… context MLA (NVIDIA#6881) Signed-off-by: peaceh <103117813+peaceh-nv@users.noreply.github.com>
… context MLA (NVIDIA#6881) Signed-off-by: peaceh <103117813+peaceh-nv@users.noreply.github.com> Signed-off-by: Wangshanshan <30051912+dominicshanshan@users.noreply.github.com>
… context MLA (NVIDIA#6881) Signed-off-by: peaceh <103117813+peaceh-nv@users.noreply.github.com> Signed-off-by: Wangshanshan <30051912+dominicshanshan@users.noreply.github.com>
… context MLA (NVIDIA#6881) Signed-off-by: peaceh <103117813+peaceh-nv@users.noreply.github.com>
… context MLA (NVIDIA#6881) Signed-off-by: peaceh <103117813+peaceh-nv@users.noreply.github.com> Signed-off-by: Wangshanshan <30051912+dominicshanshan@users.noreply.github.com>
… context MLA (NVIDIA#6881) Signed-off-by: peaceh <103117813+peaceh-nv@users.noreply.github.com>
… context MLA (NVIDIA#6881) Signed-off-by: peaceh <103117813+peaceh-nv@users.noreply.github.com> Signed-off-by: Wangshanshan <30051912+dominicshanshan@users.noreply.github.com>
4 participants
softmax[scale(Q*K)]and its value is1 / (q_scaling * sqrt(qk_head_dim))for any dtypeq_scale_quant_orig * kv_scale_quant_orig, and bmm1_scale = q_scale_quant_orig * kv_scale_quant_orig * (the scale mentioned in 1)Summary by CodeRabbit
Description
Test Coverage
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.
Details
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.