[#11318][infra] AutoDeploy: Add fused rope kernel - triton_rope_on_interleaved_qk_inputs#11327
Conversation
|
@CodeRabbit summary |
✅ Actions performedSummary regeneration triggered. |
📝 WalkthroughWalkthroughAdds a Triton JIT kernel and public op to apply RoPE on interleaved Q/K (position lookup + de-interleave + rotation), integrates an optimizer path to replace FX nodes with the new op, and includes kernel, wrapper, tests, and README entry. Changes
Sequence Diagram(s)mermaid Estimated code review effort🎯 4 (Complex) | ⏱️ ~45 minutes 🚥 Pre-merge checks | ✅ 2 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (2 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing touches
🧪 Generate unit tests (beta)
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 |
bmarimuthu-nv
force-pushed
the
bala/triton_rope_on_interleaved_qk_inputs
branch
from
6dc88b4 to
1d2475f
Compare
|
/bot run |
|
PR_Github #35032 [ run ] triggered by Bot. Commit: |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 4
🤖 Fix all issues with AI agents
In `@tensorrt_llm/_torch/auto_deploy/custom_ops/triton_kernels/rope.py`:
- Around line 131-146: The kernel assumes contiguous [B,S,H,D] layout by
computing q_out_stride_b, q_out_stride_s, q_out_stride_h and then
q_out_offsets_first/second and uses tl.store into q_out_ptr; this breaks if the
output tensor is non-contiguous. Fix by adding explicit output-stride kernel
parameters (e.g., q_out_stride_b, q_out_stride_s, q_out_stride_h or a
q_out_strides tuple) and replace the hardcoded stride calculations with those
passed-in strides when computing q_out_offsets_first and q_out_offsets_second,
then use tl.store(q_out_ptr + q_out_offsets_..., ...) as before; alternatively
ensure the caller allocates a contiguous output and document/validate that in
the kernel entry (e.g., assert contiguous) if you prefer the caller-side fix.
- Around line 5-188: Add the missing NVIDIA copyright header to the top of this
file: insert the standard NVIDIA header block (with the correct year of latest
meaningful modification) as the first lines of
tensorrt_llm/_torch/auto_deploy/custom_ops/triton_kernels/rope.py so the file
that defines rope_fwd_interleaved_kernel contains the required header; ensure
the header format matches other TensorRT-LLM source files and precedes the
definition of rope_fwd_interleaved_kernel and any imports or Triton kernel code.
In `@tensorrt_llm/_torch/auto_deploy/custom_ops/triton_rope.py`:
- Around line 143-228: Add the required NVIDIA copyright header to the top of
tensorrt_llm/_torch/auto_deploy/custom_ops/triton_rope.py (above the
apply_rope_on_interleaved_qk_inputs definition and any imports) so the file
includes the standard NVIDIA header block used across the repo (with correct
year and organization attribution and any SPDX/license identifier), preserving
existing file contents and formatting; ensure the header appears before symbols
like apply_rope_on_interleaved_qk_inputs and any triton/torch imports.
- Around line 173-196: Add strict input validation before grid computation in
triton_rope.py: assert position_ids.shape == (B, S) to ensure position_ids
matches q/k batch and sequence dims; assert k.shape[:2] == q.shape[:2] so Q and
K share B and S; and either assert H_Q >= H_K with a clear error (e.g., "H_Q
must be >= H_K") or change the grid calculation to use max(H_Q, H_K) (replace
triton.cdiv(H_Q, BLOCK_SIZE_H) with triton.cdiv(max(H_Q, H_K), BLOCK_SIZE_H)) so
heads in K beyond H_Q are processed; also keep the existing cos/sin cache shape
assertion and raise informative messages on failure.
🧹 Nitpick comments (1)
tensorrt_llm/_torch/auto_deploy/custom_ops/triton_kernels/rope.py (1)
116-129: RoPE computation not promoted to fp32, unlike sibling kernels.The existing
rope_fwd_kernel(line 250) andrope_fwd_flattened_kernel(line 337) explicitly cast loaded values totl.float32before performing RoPE math. This new kernel operates in the input dtype (e.g., bf16/fp16), which may reduce numerical accuracy for the multiply-subtract/add operations.Consider promoting
q_a,q_b,cos, andsintotl.float32before computingq_y1/q_y2(and similarly for K), then casting back when storing.
|
@CodeRabbit summary |
✅ Actions performedSummary regeneration triggered. |
|
/bot run |
1 similar comment
|
/bot run |
|
PR_Github #35052 [ run ] triggered by Bot. Commit: |
|
PR_Github #35052 [ run ] completed with state |
bmarimuthu-nv
force-pushed
the
bala/triton_rope_on_interleaved_qk_inputs
branch
from
0b3f736 to
3799161
Compare
|
/bot run |
|
PR_Github #35135 [ run ] triggered by Bot. Commit: |
|
PR_Github #35135 [ run ] completed with state
|
|
/bot run |
|
PR_Github #35158 [ run ] triggered by Bot. Commit: |
|
PR_Github #35158 [ run ] completed with state
|
bmarimuthu-nv
force-pushed
the
bala/triton_rope_on_interleaved_qk_inputs
branch
from
426157c to
dee5f09
Compare
|
/bot run |
|
PR_Github #35365 [ run ] triggered by Bot. Commit: |
bmarimuthu-nv
force-pushed
the
bala/triton_rope_on_interleaved_qk_inputs
branch
from
dee5f09 to
c08e4b7
Compare
|
/bot run |
|
PR_Github #35543 [ run ] triggered by Bot. Commit: |
|
PR_Github #35543 [ run ] completed with state
|
|
/bot run --extra-stage "DGX_B200-4_GPUs-AutoDeploy-1, DGX_H100-4_GPUs-AutoDeploy-1" |
|
PR_Github #35966 [ run ] triggered by Bot. Commit: |
|
PR_Github #35966 [ run ] completed with state
|
Signed-off-by: Balamurugan Marimuthu <246387390+bmarimuthu-nv@users.noreply.github.com>
Signed-off-by: Balamurugan Marimuthu <246387390+bmarimuthu-nv@users.noreply.github.com>
Signed-off-by: Balamurugan Marimuthu <246387390+bmarimuthu-nv@users.noreply.github.com>
Signed-off-by: Balamurugan Marimuthu <246387390+bmarimuthu-nv@users.noreply.github.com>
Signed-off-by: Suyog Gupta <41447211+suyoggupta@users.noreply.github.com>
suyoggupta
force-pushed
the
bala/triton_rope_on_interleaved_qk_inputs
branch
from
c08e4b7 to
83751b6
Compare
|
/bot run --extra-stage "DGX_B200-4_GPUs-AutoDeploy-1, DGX_H100-4_GPUs-AutoDeploy-1" |
|
PR_Github #36062 [ run ] triggered by Bot. Commit: |
|
PR_Github #36062 [ run ] completed with state
|
|
/bot run --extra-stage "DGX_B200-4_GPUs-AutoDeploy-1, DGX_H100-4_GPUs-AutoDeploy-1" |
1 similar comment
|
/bot run --extra-stage "DGX_B200-4_GPUs-AutoDeploy-1, DGX_H100-4_GPUs-AutoDeploy-1" |
|
PR_Github #36066 [ run ] triggered by Bot. Commit: |
|
PR_Github #36066 [ run ] completed with state |
|
/bot run --extra-stage "DGX_B200-4_GPUs-AutoDeploy-1, DGX_H100-4_GPUs-AutoDeploy-1" |
|
PR_Github #36067 [ run ] triggered by Bot. Commit: |
|
PR_Github #36067 [ run ] completed with state |
|
/bot run --extra-stage "DGX_B200-4_GPUs-AutoDeploy-1, DGX_H100-4_GPUs-AutoDeploy-1" |
|
PR_Github #36080 [ run ] triggered by Bot. Commit: |
|
PR_Github #36080 [ run ] completed with state |
4 participants
Summary by CodeRabbit
New Features
Tests
Documentation
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
Update tava architecture diagram if there is a significant design change in PR.
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.
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.