[#10170][fix] Add export patch for GraniteMoe MoE models to enable torch.export compatibility #10169
Conversation
📝 WalkthroughWalkthroughA new patch module is introduced that rewrites Changes
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20–25 minutes
Pre-merge checks and finishing touches✅ Passed checks (3 passed)
✨ 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 |
![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
🧹 Nitpick comments (2)
tensorrt_llm/_torch/auto_deploy/models/patches/granitemoe.py (2)
19-19: Add return type hint.The function signature is missing a return type hint. Based on line 62, the function returns a tuple of tensors:
-def _forward_moe(self, layer_input: torch.Tensor): +def _forward_moe(self, layer_input: torch.Tensor) -> tuple[torch.Tensor, torch.Tensor]:
35-35: Consider usingdim=-1for clarity.While
dim=1is functionally correct for the 2D tensortop_k_logitswith shape[B*S, top_k], usingdim=-1would be clearer and more robust (always refers to the last dimension regardless of tensor rank).- routing_weights = F.softmax(top_k_logits, dim=1, dtype=torch.float) + routing_weights = F.softmax(top_k_logits, dim=-1, dtype=torch.float)
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (1)
tensorrt_llm/_torch/auto_deploy/models/patches/granitemoe.py(1 hunks)
🧰 Additional context used
📓 Path-based instructions (2)
**/*.py
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
**/*.py: 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 files should use snake_case naming:some_file.py
Python classes should use PascalCase naming:class SomeClass
Python functions and methods should use snake_case naming:def my_awesome_function():
Python local variables should use snake_case naming:my_variable = ...
Python variable names that start with a number should be prefixed with 'k':k_99th_percentile = ...
Python global variables should use upper snake_case with prefix 'G':G_MY_GLOBAL = ...
Python constants should use upper snake_case naming: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
Python comments should be reserved for code within a function, or interfaces that are local to a file
Use Google style docstrings in Python for classes and functions, which can be parsed by Sphinx
Python attributes and variables can be documented inline with type and description
Avoid using reflection in Python when functionality can be easily achieved without reflection
When using try-except blocks in Python, limit the except 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, using the else block for logic
Files:
tensorrt_llm/_torch/auto_deploy/models/patches/granitemoe.py
**/*.{cpp,h,cu,cuh,py}
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
All TensorRT-LLM Open Source Software code should contain an NVIDIA copyright header that includes the year of its latest meaningful modification
Files:
tensorrt_llm/_torch/auto_deploy/models/patches/granitemoe.py
🧠 Learnings (1)
📓 Common learnings
Learnt from: djns99
Repo: NVIDIA/TensorRT-LLM PR: 6915
File: cpp/tensorrt_llm/kernels/cutlass_kernels/moe_gemm/moe_kernels.cu:4010-4012
Timestamp: 2025-08-14T23:23:27.449Z
Learning: For MOE (Mixture of Experts) code reviews in TensorRT-LLM, avoid repeatedly suggesting finalize fusion validation checks and safety assertions. The user djns99 has indicated these suggestions are repetitive and unwanted across multiple MOE-related changes.
Learnt from: jhaotingc
Repo: NVIDIA/TensorRT-LLM PR: 7856
File: cpp/tensorrt_llm/thop/fp8BlockScaleMoe.cpp:159-166
Timestamp: 2025-09-19T21:28:13.751Z
Learning: In TensorRT-LLM blockScaleMoe routing (cpp/tensorrt_llm/kernels/trtllmGenKernels/blockScaleMoe/runner.cu), the DeepSeek routing method performs reinterpret_cast<float*>(routingLogits) at line 89, which could cause issues if routing_logits are BF16. However, Qwen3-FP8 models use RenormalizeNaive routing method and are not affected by this dtype casting issue.
🧬 Code graph analysis (1)
tensorrt_llm/_torch/auto_deploy/models/patches/granitemoe.py (2)
tensorrt_llm/_torch/auto_deploy/export/interface.py (2)
BaseExportPatch(47-139)ExportPatchRegistry(186-233)tensorrt_llm/_torch/auto_deploy/custom_ops/fused_moe/torch_moe.py (1)
torch_moe(90-201)
⏰ 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)
tensorrt_llm/_torch/auto_deploy/models/patches/granitemoe.py (2)
27-62: LGTM! The export-compatible forward pass is correctly implemented.The implementation properly:
- Avoids the problematic
.tolist()by directly accessingself.router.layerfor router logits computation- Performs top-k selection and softmax in an export-compatible manner
- Correctly reorders weights from GraniteMoe's
[gate, up]format to torch_moe's expected[w3, w1]stacked format- Invokes
torch_moewith the correct stacked-tensor format (single-element lists forw1_weightandw2_weight, empty list forw3_weight)- Returns both
layer_outputandrouter_logitsto match the original forward signatureThe PR objectives confirm that numerical accuracy was verified against the original forward pass.
65-89: LGTM! The patch mechanism is correctly implemented.The patch class properly:
- Registers with the export patch registry under the key
"hf_granitemoe_moe"- Implements the required
_apply_patchand_revert_patchmethods fromBaseExportPatch- Safely stores the original
forwardmethod before patching and restores it on revert- Uses method monkey-patching to replace
GraniteMoeMoE.forwardat the class levelThe implementation follows the export patch framework contract and ensures clean patch application/removal.
| """A patch for GraniteMoe to make it compatible with torch.export. | ||
|
|
||
| The main issue is that GraniteMoeTopKGating calls `.tolist()` on a tensor which is: | ||
| 1. Incompatible with meta tensors (no data to access) | ||
| 2. Incompatible with torch.compile/export (data-dependent operation) | ||
|
|
||
| This patch rewrites the MoE forward to use the torch_moe custom op instead. | ||
| """ | ||
|
|
||
| import torch | ||
| import torch.nn.functional as F | ||
|
|
||
| # Try to import GraniteMoe classes from transformers | ||
| from transformers.models.granitemoe.modeling_granitemoe import GraniteMoeMoE | ||
|
|
||
| from ...export.interface import BaseExportPatch, ExportPatchRegistry | ||
|
|
There was a problem hiding this comment.
Add required copyright header and fix import style.
Two coding guideline violations:
-
Missing copyright header: All TensorRT-LLM code must include an NVIDIA copyright header with the year of last modification.
-
Import style (line 14): The coding guidelines require maintaining namespace when importing. Instead of importing
GraniteMoeMoEdirectly, import the module:from transformers.models.granitemoe import modeling_granitemoe
Then reference as
modeling_granitemoe.GraniteMoeMoEin the code (lines 14, 81, 84, 89).
As per coding guidelines, Python files should contain copyright headers and maintain namespace on imports.
🤖 Prompt for AI Agents
tensorrt_llm/_torch/auto_deploy/models/patches/granitemoe.py lines 1-17: add the
required NVIDIA copyright header with the last modification year (2025) at the
top of the file, and replace the direct class import on line 14 with a
namespaced module import: "from transformers.models.granitemoe import
modeling_granitemoe", then update all uses of GraniteMoeMoE (lines 14, 81, 84,
89) to reference modeling_granitemoe.GraniteMoeMoE so the namespace is preserved
per coding guidelines.
karthikvetrivel
changed the title
|
/bot run |
|
PR_Github #29191 [ run ] triggered by Bot. Commit: |
|
PR_Github #29191 [ run ] completed with state
|
|
/bot run |
|
PR_Github #29198 [ run ] triggered by Bot. Commit: |
|
PR_Github #29198 [ run ] completed with state
|
…patibility Signed-off-by: Karthik Vetrivel <[email protected]>
karthikvetrivel
force-pushed
the
add-support-graniteMOE
branch
from
cf27ce0 to
ede0dcd
Compare
|
/bot run |
3 participants
Description
Fixes #10170 AutoDeploy compilation failures for IBM Granite MoE models (e.g.,
ibm-granite/granite-3.1-3b-a800m-instruct) with the error:Root Cause:
GraniteMoeTopKGatingcalls.tolist()on a tensor, which is:Solution: Adds a new export patch (
hf_granitemoe_moe) that rewrites the MoE forward pass.Test Coverage
torch.exportof the full model, which was previously failing with the "Cannot copy out of meta tensor" error.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.
Summary by CodeRabbit
✏️ Tip: You can customize this high-level summary in your review settings.