[None][autodeploy] Add group attention pattern that supports attention masks #7054
Conversation
Signed-off-by: Frida Hou <201670829+Fridah-nv@users.noreply.github.com>
📝 WalkthroughWalkthroughAdds a new grouped-attention match/rewrite (pattern_5 → replacement_5) that repeats K/V by n_rep, matches SDPA with attn_mask, and replaces it with grouped SDPA. Registers dummy_args_3 and scalar_workaround for n_rep. Emits a warning if no grouped-attention patterns are found. No public API changes. Changes
Sequence Diagram(s)sequenceDiagram
autonumber
participant MC as Model Code
participant GT as GroupedAttention Transformer
participant PM as Pattern Matcher (patterns 1..5)
participant RP as Replacement Builder
participant GSD as torch_attention_grouped_sdpa
MC->>GT: attention(q, k1, v1, n_rep, attn_mask)
GT->>PM: attempt match (pattern_1..pattern_5)
PM-->>GT: match pattern_5 (repeat_kv + SDPA with attn_mask)
GT->>RP: construct replacement_5 (dummy_args_3, scalar_workaround(n_rep))
RP->>GSD: default(q, k, v, attn_mask)
GSD-->>MC: grouped SDPA result
Note over GT,PM: other patterns (1–4) remain registered/unchanged
Estimated code review effort🎯 2 (Simple) | ⏱️ ~10 minutes 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. 📜 Recent review detailsConfiguration used: .coderabbit.yaml 💡 Knowledge Base configuration:
You can enable these sources in your CodeRabbit configuration. 📒 Files selected for processing (1)
🚧 Files skipped from review as they are similar to previous changes (1)
⏰ 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/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: 0
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
tensorrt_llm/_torch/auto_deploy/transform/library/attention.py (1)
1-1: Missing NVIDIA copyright headerPer the repository guidelines, prepend the current-year NVIDIA copyright header.
Example header to add at the very top:
# Copyright (c) 2025, NVIDIA CORPORATION. All rights reserved.
🧹 Nitpick comments (4)
tensorrt_llm/_torch/auto_deploy/transform/library/attention.py (4)
327-331: Use keyword argument for attn_mask in SDPA call to avoid schema/positional driftPrefer passing attn_mask as a keyword to align with the rest of the file and avoid accidental breakage if the op schema or arg ordering changes.
Apply this diff:
- return torch.ops.auto_deploy.torch_attention_sdpa.default(q, k, v, attn_mask) + return torch.ops.auto_deploy.torch_attention_sdpa.default(q, k, v, attn_mask=attn_mask)
333-335: Use keyword argument for attn_mask in grouped SDPA replacementSame nit as above: use a keyword for clarity and resilience to schema changes.
- return torch.ops.auto_deploy.torch_attention_grouped_sdpa.default(q, k, v, attn_mask) + return torch.ops.auto_deploy.torch_attention_grouped_sdpa.default(q, k, v, attn_mask=attn_mask)
436-436: Heads mismatch risk in dummy_args_3 (q=8 heads vs. repeated k/v to 7): hard error if executedq has 8 heads while k1/v1 have 1 head and you repeat with n_rep=7 → k/v end up with 7 heads. If the pattern function is ever executed instead of traced symbolically, SDPA will see q_heads != kv_heads and crash. Even if it works today via FX tracing, tightening this avoids fragile behavior.
Consider computing n_rep from q and k1 to keep heads consistent for this pattern only:
- dummy_args_3 = [q, k1, v1, n_rep, attn_mask] + dummy_args_3 = [q, k1, v1, q.shape[1] // k1.shape[1], attn_mask]If you prefer to keep n_rep as a shared constant for the earlier patterns, this keeps them unchanged while hardening just pattern_5.
472-478: Make pattern_5 registration more robust; also align scalar_workaround with computed n_rep
- Add op_ignore_types to tolerate common dtype casts around masks in the wild (mirrors how SFDP patterns ignore aten.to(dtype)).
- If you adopt the computed n_rep in dummy_args_3 above, mirror that here for scalar_workaround to ensure a stable scalar bind.
register_ad_pattern( search_fn=_grouped_attn_pattern_5, replace_fn=_grouped_attn_replacement_5, patterns=patterns, - dummy_args=dummy_args_3, - scalar_workaround={"n_rep": n_rep}, + dummy_args=dummy_args_3, + scalar_workaround={"n_rep": q.shape[1] // k1.shape[1]}, + op_ignore_types={torch.ops.aten.to.dtype: (torch.dtype,)}, )If you’d like, I can follow up with a lightweight unit test that constructs a minimal FX graph exhibiting this pattern and verifies the rewrite hits.
📜 Review details
Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
💡 Knowledge Base configuration:
- MCP integration is disabled by default for public repositories
- Jira integration is disabled by default for public repositories
- Linear integration is disabled by default for public repositories
You can enable these sources in your CodeRabbit configuration.
📒 Files selected for processing (1)
tensorrt_llm/_torch/auto_deploy/transform/library/attention.py(3 hunks)
🧰 Additional context used
📓 Path-based instructions (2)
**/*.py
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
**/*.py: Python code must target Python 3.8+
Python indentation: 4 spaces, no tabs
Maintain module namespace in imports (from package.subpackage import foo; then use foo.SomeClass())
Python file names use snake_case
Python class names use PascalCase
Python functions/methods and local variables use snake_case; variables starting with a number get k_ prefix (e.g., k_99th_percentile)
Global variables use G_ prefixed UPPER_SNAKE_CASE (e.g., G_MY_GLOBAL)
Constants use UPPER_SNAKE_CASE in Python
Avoid shadowing variables from outer scopes in Python
Initialize all externally visible members of a Python class in init
Prefer docstrings for interfaces used outside a file; comments for local code
Use Google-style docstrings for classes and functions (Sphinx-parsable)
Document attributes/variables inline with short docstrings
Avoid reflection when simple alternatives exist (e.g., prefer explicit parameters over dict(**locals()))
In try/except, catch the narrowest exceptions possible
For duck-typing with try/except, keep try body minimal and put logic in else
Files:
tensorrt_llm/_torch/auto_deploy/transform/library/attention.py
**/*.{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:
tensorrt_llm/_torch/auto_deploy/transform/library/attention.py
Signed-off-by: Frida Hou <201670829+Fridah-nv@users.noreply.github.com>
|
/bot run |
|
PR_Github #15812 [ run ] triggered by Bot |
suyoggupta
changed the title
|
PR_Github #15812 [ run ] completed with state |
4 participants
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.
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.