[TRTLLM-10673][feat] Improved layer classification for sharding #10718
Conversation
|
/bot run |
📝 WalkthroughWalkthroughThe changes refactor sharding layer detection from attention-name-based heuristics to a subgraph-driven classification approach. This includes introducing a new Changes
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~22 minutes 🚥 Pre-merge checks | ✅ 3 | ❌ 2❌ Failed checks (2 warnings)
✅ Passed checks (3 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing touches
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: 1
🤖 Fix all issues with AI agents
In `@tensorrt_llm/_torch/auto_deploy/transform/library/sharding.py`:
- Around line 2039-2055: The branch handling when len(layer_subgraph) != 1 is
unclear: ensure the control flow explicitly skips processing for non-matched
linear nodes and that the warning message matches the actual behavior. Update
the else block around the layer_subgraph lookup so that when lin_node is in
unprocessed_linear_nodes you set layer_type = LayerType.UNKNOWN and continue,
and when not in unprocessed_linear_nodes you call ad_logger.warning(...) and
then continue; i.e., move or add the continue so both paths unambiguously skip
further processing of that lin_node (referencing lin_node, layer_subgraph,
unprocessed_linear_nodes, LayerType.UNKNOWN, and ad_logger).
🧹 Nitpick comments (1)
tensorrt_llm/_torch/auto_deploy/utils/node_utils.py (1)
309-315: Consider adding a docstring for this public helper.The function logic is clear, but since this is a new public helper (as noted in the AI summary), adding a brief docstring would improve discoverability and maintainability.
📝 Suggested docstring
def is_parametrized_op(node: Node) -> bool: + """Check if the node is a parametrized operation with multi-dimensional weights. + + Returns True if the node has an associated weight node with shape dimension > 1. + This filters out nodes with only 1D parameters like biases. + """ # check if the node has a weight argument if (w := extract_weight_node(node)) is not None: if len(shape(w)) > 1: return True return False
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (2)
tensorrt_llm/_torch/auto_deploy/transform/library/sharding.pytensorrt_llm/_torch/auto_deploy/utils/node_utils.py
🧰 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 Python modules, 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, with prefixkfor variable names that start with a number (e.g.,k_99th_percentile)
Python global variables should use upper snake_case with prefixG(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 Python interfaces that may be used outside a file, prefer docstrings over comments
Use comments in Python for code within a function, or interfaces that are local to a file
Use Google-style docstrings for Python classes and functions, which can be parsed by Sphinx
Python attributes and variables can be documented inline with the format"""<type>: Description"""
Avoid using reflection in Python when functionality can be easily achieved without reflection
When using try-except blocks in Python, limit the except clause to the smallest set of errors possible
When using try-except blocks in Python to handle multiple possible variable types (duck-typing), keep the body of the try as small as possible and use the else block for the main logic
Files:
tensorrt_llm/_torch/auto_deploy/transform/library/sharding.pytensorrt_llm/_torch/auto_deploy/utils/node_utils.py
**/*.{cpp,cc,cxx,h,hpp,hxx,cu,cuh,py}
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
All TensorRT-LLM source files (.cpp, .h, .cu, .py, and other source files) should contain an NVIDIA copyright header with the year of latest meaningful modification
Files:
tensorrt_llm/_torch/auto_deploy/transform/library/sharding.pytensorrt_llm/_torch/auto_deploy/utils/node_utils.py
🧠 Learnings (1)
📚 Learning: 2025-10-20T16:54:09.824Z
Learnt from: nvchenghaoz
Repo: NVIDIA/TensorRT-LLM PR: 8469
File: tensorrt_llm/_torch/auto_deploy/custom_ops/rms_norm.py:6-6
Timestamp: 2025-10-20T16:54:09.824Z
Learning: In tensorrt_llm/_torch/auto_deploy/custom_ops/rms_norm.py, the import `from ...modules.mamba.layernorm_gated import _layer_norm_fwd` is correct and should not be changed to modules.fla.layernorm_gated. The _layer_norm_fwd function exists in both modules/mamba/layernorm_gated.py and modules/fla/layernorm_gated.py, but the mamba version is the intended implementation for this use case.
Applied to files:
tensorrt_llm/_torch/auto_deploy/utils/node_utils.py
🧬 Code graph analysis (1)
tensorrt_llm/_torch/auto_deploy/transform/library/sharding.py (2)
tensorrt_llm/_torch/auto_deploy/utils/node_utils.py (3)
get_all_layer_subgraphs(468-519)extract_weight_node(132-174)LayerType(34-42)tensorrt_llm/_torch/auto_deploy/transform/library/fused_moe.py (1)
target(513-514)
⏰ 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 (4)
tensorrt_llm/_torch/auto_deploy/utils/node_utils.py (3)
132-132: DefensiveNonehandling forextract_weight_node.The early return when
weight_node is Noneprevents potential issues when passingNonetofind_get_attr_node. The return type annotation correctly reflects that this function can returnNone.Note: Callers like
extract_param_names_from_node(line 191) already assert that the result is non-None, so existing call sites appear compatible with this change.Also applies to: 171-174
858-869: LGTM - Key fix for issue#10673.This correctly handles the case where subgraph detection "spills over" layer boundaries (e.g., when unexpected RMSNorm appears between projections). Marking these as
UNKNOWNensures they fall back to conservative simple-sharding rather than applying potentially incorrect column-row sharding.The comment explaining the MoLE scenario is helpful for future maintainers.
902-928: More conservative layer type classification - addresses the core issue.The stricter checks are well-designed:
- SSM/attention layers must have exactly one corresponding op, otherwise
UNKNOWN- MLA detection requires the specific pattern of 2 intermediate linear + 1 attention node
- The
intermediate_weight_nodescheck (lines 926-928) catches cases like RMSNorm where unexpected weighted operations appear in the subgraph interiorThis directly addresses the MiniMax model crash from issue
#10673where RMSNorm weights (non-sharded) caused shape mismatches when column-row sharding was applied.tensorrt_llm/_torch/auto_deploy/transform/library/sharding.py (1)
2069-2073: LGTM - Extends config support for MLA layers.Adding
"mla"to the condition ensures MLA configurations fromtp_plantrigger the proper layer subgraph detection, consistent with theLayerType.MLAenum value and the_process_mla_shardingfunction already in the codebase.
✏️ Tip: You can disable this entire section by setting review_details to false in your review settings.
|
PR_Github #32164 [ run ] triggered by Bot. Commit: |
|
PR_Github #32164 [ run ] completed with state
|
|
/bot run --reuse-test |
|
PR_Github #32290 [ run ] triggered by Bot. Commit: |
|
/bot run |
|
PR_Github #32293 [ run ] triggered by Bot. Commit: |
|
PR_Github #32290 [ run ] completed with state |
|
PR_Github #32293 [ run ] completed with state |
|
/bot run --reuse-test |
|
PR_Github #32318 [ run ] triggered by Bot. Commit: |
|
PR_Github #32318 [ run ] completed with state
|
|
/bot run --reuse-test |
|
PR_Github #32331 [ run ] triggered by Bot. Commit: |
|
PR_Github #32331 [ run ] completed with state |
|
/bot run --reuse-test |
|
PR_Github #32408 [ run ] triggered by Bot. Commit: |
|
PR_Github #32408 [ run ] completed with state
|
|
/bot run --reuse-test |
|
PR_Github #32410 [ run ] triggered by Bot. Commit: |
|
PR_Github #32410 [ run ] completed with state |
|
/bot run |
|
PR_Github #32419 [ run ] triggered by Bot. Commit: |
|
PR_Github #32419 [ run ] completed with state |
|
/bot run |
|
PR_Github #32436 [ run ] triggered by Bot. Commit: |
greg-kwasniewski1
removed request for
a team,
Naveassaf,
Shixiaowei02,
Tabrizian,
chuangz0,
hyukn,
laikhtewari,
suyoggupta,
syuoni,
taylor-yb-lee,
yuxianq,
zeroepoch and
zheyuf
Signed-off-by: greg-kwasniewski1 <213329731+greg-kwasniewski1@users.noreply.github.com>
greg-kwasniewski1
force-pushed
the
gk/better_layer_classification
branch
from
b784fe9 to
4139524
Compare
|
/bot run |
Signed-off-by: greg-kwasniewski1 <213329731+greg-kwasniewski1@users.noreply.github.com>
|
PR_Github #33208 [ run ] triggered by Bot. Commit: |
|
/bot run |
|
PR_Github #33210 [ run ] triggered by Bot. Commit: |
|
PR_Github #33210 [ run ] completed with state
|
|
/bot run --extra-stage "DGX_B200-4_GPUs-AutoDeploy-1, DGX_H100-4_GPUs-AutoDeploy-1" |
2 participants
Fixes #10673
Summary by CodeRabbit
✏️ Tip: You can customize this high-level summary in your review settings.
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.