[TRTLLM-6992][feat] Add runtime swap AB for SM100 FP8 blockwise GEMM #6586
Conversation
📝 WalkthroughWalkthroughThe changes update the logic for FP8 GEMM operations in the linear module to handle small batch sizes differently on specific hardware (SM version 100). The quantization utility functions are enhanced to support operand swapping, masked transposition, and improved device handling. A new masked transpose Triton kernel and related function are introduced for efficient tensor manipulation. Weight loading logic in a DeepseekV3 model is modified to adjust how FP8 scale parameters are assigned and transformed, removing a conditional transformation block. Changes
Sequence Diagram(s)sequenceDiagram
participant Caller
participant LinearMethod
participant fp8_utils
participant deep_gemm
Caller->>LinearMethod: apply(input, weight, ...)
alt get_sm_version() == 100
alt batch size < 32
LinearMethod->>fp8_utils: per_token_quant_and_transform(input, swap_ab=True)
LinearMethod->>deep_gemm: fp8_gemm_nt(weight, input, ...)
deep_gemm-->>LinearMethod: padded_output
LinearMethod->>fp8_utils: masked_transpose(padded_output, batch_size)
fp8_utils-->>LinearMethod: output
else batch size >= 32
LinearMethod->>fp8_utils: per_token_quant_and_transform(input)
LinearMethod->>deep_gemm: fp8_gemm_nt(input, weight, ...)
deep_gemm-->>LinearMethod: output
end
else other SM version
LinearMethod->>fp8_utils: per_token_quant_and_transform(input)
LinearMethod->>deep_gemm: fp8_gemm_nt(input, weight, ...)
deep_gemm-->>LinearMethod: output
end
LinearMethod-->>Caller: output
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20 minutes Possibly related PRs
Suggested reviewers
Note ⚡️ Unit Test Generation is now available in beta!Learn more here, or try it out under "Finishing Touches" below. 📜 Recent review detailsConfiguration used: .coderabbit.yaml 📒 Files selected for processing (3)
🚧 Files skipped from review as they are similar to previous changes (3)
⏰ 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)
✨ 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 comments)
Other keywords and placeholders
Documentation and Community
|
Barry-Delaney
changed the title
Barry-Delaney
changed the title
![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)
tensorrt_llm/quantization/utils/fp8_utils.py (1)
457-461: Document the newswap_abparameter.The
swap_abparameter lacks documentation in the function docstring. Please add a description explaining its purpose and effect on the output dimensions.Update the docstring to include:
""" input shape [g, m, k] output shape [g, m, k // 2], dtype fp8 output_scale [g, k // 4, m // 2 // 128], dtype int32 quant_group_size int masked_m shape [g] swap_ab: bool, if True, pads m dimension to multiple of 8 for efficient transpose operations """
🧹 Nitpick comments (1)
tensorrt_llm/_torch/modules/linear.py (1)
577-591: Verify the batch size threshold and document the optimization.The implementation correctly handles small batch optimization by swapping operands and using masked transpose. However:
- The batch size threshold of 32 should be documented or made configurable
- Consider adding a comment explaining why this optimization is beneficial for small batches
Consider making the threshold configurable:
SMALL_BATCH_THRESHOLD = 32 # Empirically determined for SM100 if input.shape[0] < SMALL_BATCH_THRESHOLD:
📜 Review details
Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (2)
tensorrt_llm/_torch/modules/linear.py(1 hunks)tensorrt_llm/quantization/utils/fp8_utils.py(4 hunks)
🧰 Additional context used
📓 Path-based instructions (2)
**/*.py
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
**/*.py: The code developed for TensorRT-LLM should conform to Python 3.8+.
Indent Python code with 4 spaces. Do not use tabs.
Always maintain the namespace when importing in Python, even if only one class or function from a module is used.
Python filenames should use snake_case (e.g., some_file.py).
Python classes should use PascalCase (e.g., class SomeClass).
Python functions and methods should use snake_case (e.g., def my_awesome_function():).
Python local variables should use snake_case. Prefix k for variable names that start with a number (e.g., k_99th_percentile = ...).
Python global variables should use upper snake_case and prefix G (e.g., G_MY_GLOBAL = ...).
Python constants should use upper snake_case (e.g., MY_CONSTANT = ...).
Avoid shadowing variables declared in an outer scope in Python.
Initialize all externally visible members of a Python class in the constructor.
For interfaces that may be used outside a file, prefer docstrings over comments in Python.
Comments in Python should be reserved for code within a function, or interfaces that are local to a file.
Use Google style docstrings for classes and functions in Python, which can be parsed by Sphinx.
Attributes and variables in Python can be documented inline; attribute docstrings will be rendered under the docstring for the class.
Avoid using reflection in Python when functionality can be easily achieved without it.
When using try-except blocks in Python, limit the except to the smallest set of errors possible.
When using try-except blocks to handle multiple possible variable types in Python, keep the body of the try as small as possible, using the else block to implement the logic.
Files:
tensorrt_llm/_torch/modules/linear.pytensorrt_llm/quantization/utils/fp8_utils.py
**/*.{cpp,h,hpp,cc,cxx,cu,py}
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
All TensorRT-LLM Open Source Software code should contain an NVIDIA copyright header that includes the current year. This includes .cpp, .h, .cu, .py, and any other source files which are compiled or interpreted.
Files:
tensorrt_llm/_torch/modules/linear.pytensorrt_llm/quantization/utils/fp8_utils.py
🔇 Additional comments (4)
tensorrt_llm/quantization/utils/fp8_utils.py (3)
339-341: Good improvements for performance and device handling.The changes improve efficiency by avoiding unnecessary zero-initialization and ensure correct device placement by using the input tensor's device.
481-495: Correct implementation of conditional padding logic.The implementation properly handles the
swap_abflag to pad themdimension when needed, and consistently usesm_paddedthroughout the function. The device handling improvements are also good.
537-555: Well-implemented transpose kernel.The Triton kernel correctly implements a masked transpose operation with proper boundary handling and flexible stride support.
tensorrt_llm/_torch/modules/linear.py (1)
582-590: Verify error handling for deep_gemm.fp8_gemm_nt failuresThe calls to
deep_gemm.fp8_gemm_ntin both
tensorrt_llm/_torch/modules/linear.pytensorrt_llm/_torch/modules/fused_moe/fused_moe_deepgemm.pyassume the underlying C-extension will either succeed or raise a Python exception. If you need to detect timeout, out-of-memory, or other low-level failures and either recover or emit a clearer error message, wrap each
deep_gemm.fp8_gemm_nt(...)invocation in atry/exceptblock.• Confirm what errors (if any)
deep_gemm.fp8_gemm_ntmay raise on failure
• Decide on fallback behavior or more descriptive logging/user feedback
• Update the two call sites to catch and handle those exceptions appropriately
|
/bot run |
|
PR_Github #13921 [ run ] triggered by Bot |
|
PR_Github #13921 [ run ] completed with state |
Barry-Delaney
force-pushed
the
user/barry/swap_ab
branch
from
5eac9b4 to
76ca12d
Compare
|
/bot run |
|
PR_Github #13937 [ run ] triggered by Bot |
Barry-Delaney
changed the title
|
/bot kill |
|
/bot run |
|
PR_Github #13943 [ kill ] triggered by Bot |
|
PR_Github #13937 [ run ] completed with state |
|
PR_Github #13945 [ run ] triggered by Bot |
|
PR_Github #13943 [ kill ] completed with state |
|
/bot kill |
|
PR_Github #13963 [ kill ] triggered by Bot |
|
PR_Github #13945 [ run ] completed with state |
|
PR_Github #13963 [ kill ] completed with state |
Barry-Delaney
force-pushed
the
user/barry/swap_ab
branch
from
e16a15f to
0171d1b
Compare
|
PR_Github #14939 [ run ] completed with state |
Barry-Delaney
force-pushed
the
user/barry/swap_ab
branch
from
70fe9c4 to
14101fd
Compare
|
/bot run |
|
/bot kill |
Barry-Delaney
force-pushed
the
user/barry/swap_ab
branch
2 times, most recently
from
c368be1 to
e82f206
Compare
|
/bot run |
|
PR_Github #15065 [ run ] triggered by Bot |
|
PR_Github #15065 [ run ] completed with state |
Barry-Delaney
force-pushed
the
user/barry/swap_ab
branch
2 times, most recently
from
f83e882 to
0160ff0
Compare
|
/bot run |
|
PR_Github #15100 [ run ] triggered by Bot |
|
PR_Github #15100 [ run ] completed with state |
|
/bot run |
|
PR_Github #15149 [ run ] triggered by Bot |
|
PR_Github #15149 [ run ] completed with state |
Barry-Delaney
force-pushed
the
user/barry/swap_ab
branch
from
0160ff0 to
4db1500
Compare
|
/bot run |
|
PR_Github #15215 [ run ] triggered by Bot |
|
PR_Github #15215 [ run ] completed with state |
|
/bot run |
|
PR_Github #15228 [ run ] triggered by Bot |
Barry-Delaney
force-pushed
the
user/barry/swap_ab
branch
from
75315be to
d184326
Compare
|
/bot run |
|
PR_Github #15361 [ run ] triggered by Bot |
|
PR_Github #15361 [ run ] completed with state |
Signed-off-by: Barry Kang <[email protected]>
Signed-off-by: Barry Kang <[email protected]>
Barry-Delaney
force-pushed
the
user/barry/swap_ab
branch
from
d184326 to
c99ab7b
Compare
Summary by CodeRabbit
New Features
Bug Fixes
Refactor
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.
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.