[TRTLLM-8714][fix] update create_input_processor to handle custom checkpoint format #7811
Conversation
|
/bot run |
|
PR_Github #19014 [ run ] triggered by Bot |
|
PR_Github #19014 [ run ] completed with state |
Funatiq
force-pushed
the
fix/custom_loading
branch
from
e284e4f to
e92f6c7
Compare
|
/bot run |
|
PR_Github #19166 [ run ] triggered by Bot |
|
PR_Github #19166 [ run ] completed with state |
📝 WalkthroughWalkthroughIntroduces an optional checkpoint_format parameter to create_input_processor, altering control flow to conditionally skip HF config loading for non-HF formats. Updates LLM Torch backend to read checkpoint_format from args and pass it to input processor creation. Defaults preserve previous behavior when checkpoint_format is None or "HF". Changes
Sequence Diagram(s)sequenceDiagram
autonumber
actor Client
participant LLM as LLM (Torch backend)
participant Inputs as inputs.create_input_processor
participant HF as ModelConfig.from_pretrained
Client->>LLM: build model
LLM->>LLM: read args.checkpoint_format
LLM->>Inputs: create_input_processor(model_dir, tokenizer, checkpoint_format)
alt checkpoint_format is None or "HF"
Inputs->>HF: try load ModelConfig
alt load succeeds
HF-->>Inputs: model_config
Inputs->>Inputs: select processor by model type
else load fails
HF-->>Inputs: error
Inputs->>Inputs: fallback to DefaultInputProcessor
end
else non-HF format
Inputs->>Inputs: skip HF config load
Inputs->>Inputs: fallback/resolve appropriate processor (may default)
end
Inputs-->>LLM: InputProcessor instance
LLM-->>Client: proceed with preprocessing
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Pre-merge checks and finishing touches❌ Failed checks (2 warnings)
✅ Passed checks (1 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: 0
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (3)
tensorrt_llm/inputs/registry.py (1)
1-1: Add NVIDIA Apache-2.0 header (2025).Per coding guidelines, prepend the NVIDIA Apache-2.0 copyright header with current year to all .py files.
+# Copyright (c) 2025, NVIDIA CORPORATION. All rights reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License.tensorrt_llm/llmapi/llm.py (2)
1-1: Add NVIDIA Apache-2.0 header (2025).Apply the standard header at top of file.
+# Copyright (c) 2025, NVIDIA CORPORATION. All rights reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License.
960-978: Addcheckpoint_formatto allcreate_input_processorcalls
Intensorrt_llm/llmapi/llm.py(lines 807–811) andtensorrt_llm/llmapi/mm_encoder.py(lines 54–56), include thecheckpoint_formatargument (e.g.getattr(self.args, "checkpoint_format", None)) when invokingcreate_input_processor.
🧹 Nitpick comments (4)
tensorrt_llm/inputs/registry.py (3)
411-418: Polish API: add return type, expand docstring, and follow module-namespace import style.
- Add return type hint (InputProcessor).
- Expand docstring with Args/Returns and default behavior.
- Import modules, not symbols, per guidelines.
-def create_input_processor(model_path_or_dir: str, - tokenizer, - checkpoint_format: Optional[str] = None): - """Create an input processor for a specific model. - - If checkpoint_format is not "HF", fall back to DefaultInputProcessor. - """ - from tensorrt_llm._torch.model_config import ModelConfig - from tensorrt_llm._torch.models import get_model_architecture +def create_input_processor(model_path_or_dir: str, + tokenizer, + checkpoint_format: Optional[str] = None) -> InputProcessor: + """Create an input processor for a specific model. + + Args: + model_path_or_dir: Path or repo id used to locate pretrained config/tokenizer. + tokenizer: Tokenizer instance. + checkpoint_format: Checkpoint format identifier. "HF" uses Hugging Face-style + config loading; any other value skips HF config loading. None is treated as "HF". + + Returns: + An InputProcessor implementation (model-specific if registered; otherwise DefaultInputProcessor). + """ + from tensorrt_llm._torch import model_config as tllm_model_config + from tensorrt_llm._torch import models as tllm_models
421-445: Make format check case-insensitive, log the branch, broaden exception, and pass through model_path on fallback.
- Case-insensitive "HF" check.
- Debug logs to clarify behavior.
- Catch OSError (alias of EnvironmentError) to be future-proof; drop unused local.
- Pass model_path_or_dir to DefaultInputProcessor for context.
- model_config = None - - if checkpoint_format is None or checkpoint_format == "HF": - try: - config = ModelConfig.from_pretrained(model_path_or_dir, - trust_remote_code=True) - model_config = config.pretrained_config - except (ValueError, EnvironmentError): - config = None + model_config = None + fmt_is_hf = (checkpoint_format or "HF").upper() == "HF" + if not fmt_is_hf: + logger.debug(f"checkpoint_format={checkpoint_format!r}; skipping HF config load.") + if fmt_is_hf: + try: + config = tllm_model_config.ModelConfig.from_pretrained( + model_path_or_dir, trust_remote_code=True) + model_config = config.pretrained_config + except (ValueError, EnvironmentError, OSError) as e: + logger.debug(f"Unable to load HF config from {model_path_or_dir!r}: {e}. Falling back.") if model_config is not None: try: - model_cls, _ = get_model_architecture(model_config) + model_cls, _ = tllm_models.get_model_architecture(model_config) input_processor_cls = INPUT_PROCESSOR_REGISTRY._input_processors_cls_by_model_type \ .get(model_cls) except RuntimeError: # unregistered model logger.info("Unregistered model, using DefaultInputProcessor") input_processor_cls = None if input_processor_cls is not None: return input_processor_cls(model_path_or_dir, model_config, tokenizer, trust_remote_code=True) - return DefaultInputProcessor(None, None, tokenizer) + return DefaultInputProcessor(model_path_or_dir, None, tokenizer)
411-446: Add basic unit tests to lock HF vs non‑HF behavior.Recommend covering:
- checkpoint_format None/"HF" loads HF config and returns registered processor when available.
- checkpoint_format "custom" skips HF and returns DefaultInputProcessor.
- Graceful fallback when HF config load fails.
Would you like me to scaffold pytest tests for these paths?
tensorrt_llm/llmapi/llm.py (1)
809-812: Parity: pass checkpoint_format in TRT backend too to avoid unnecessary HF config attempts.This keeps behavior consistent with PyTorch backend and skips HF config load when using custom formats.
- self.input_processor = create_input_processor(self._hf_model_dir, - self.tokenizer) + checkpoint_format = getattr(self.args, "checkpoint_format", None) + self.input_processor = create_input_processor(self._hf_model_dir, + self.tokenizer, + checkpoint_format)
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (2)
tensorrt_llm/inputs/registry.py(1 hunks)tensorrt_llm/llmapi/llm.py(1 hunks)
🧰 Additional context used
📓 Path-based instructions (3)
**/*.{h,hpp,hh,hxx,cpp,cxx,cc,cu,cuh,py}
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
Use only spaces, no tabs; indent with 4 spaces.
Files:
tensorrt_llm/llmapi/llm.pytensorrt_llm/inputs/registry.py
**/*.py
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
**/*.py: Python code must target Python 3.8+.
Indent Python code with 4 spaces; do not use tabs.
Maintain module namespace when importing; prefer 'from package.subpackage import foo' then 'foo.SomeClass()' instead of importing the class directly.
Python filenames should be snake_case (e.g., some_file.py).
Python classes use PascalCase names.
Functions and methods use snake_case names.
Local variables use snake_case; prefix 'k' for variables that start with a number (e.g., k_99th_percentile).
Global variables use upper SNAKE_CASE prefixed with 'G' (e.g., G_MY_GLOBAL).
Constants use upper SNAKE_CASE (e.g., MY_CONSTANT).
Avoid shadowing variables from an outer scope.
Initialize all externally visible members of a class in the constructor.
Prefer docstrings for interfaces that may be used outside a file; comments for in-function or file-local interfaces.
Use Google-style docstrings for classes and functions (Sphinx-parsable).
Document attributes and variables inline so they render under the class/function docstring.
Avoid reflection when a simpler, explicit approach suffices (e.g., avoid dict(**locals()) patterns).
In try/except, catch the most specific exceptions possible.
For duck-typing try/except, keep the try body minimal and use else for the main logic.
Files:
tensorrt_llm/llmapi/llm.pytensorrt_llm/inputs/registry.py
**/*.{cpp,cxx,cc,h,hpp,hh,hxx,cu,cuh,py}
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
Prepend the NVIDIA Apache-2.0 copyright header with current year to the top of all source files (e.g., .cpp, .h, .cu, .py).
Files:
tensorrt_llm/llmapi/llm.pytensorrt_llm/inputs/registry.py
🧬 Code graph analysis (2)
tensorrt_llm/llmapi/llm.py (3)
tensorrt_llm/_torch/models/checkpoints/base_checkpoint_loader.py (1)
checkpoint_format(50-51)tensorrt_llm/_torch/models/checkpoints/hf/checkpoint_loader.py (1)
checkpoint_format(74-75)tensorrt_llm/inputs/registry.py (1)
create_input_processor(411-445)
tensorrt_llm/inputs/registry.py (1)
tensorrt_llm/_torch/models/modeling_utils.py (1)
get_model_architecture(708-720)
🔇 Additional comments (2)
tensorrt_llm/inputs/registry.py (1)
411-446: All create_input_processor callsites pass at most three positional arguments; the new optional checkpoint_format parameter won’t break existing calls.tensorrt_llm/llmapi/llm.py (1)
974-977: LGTM: checkpoint_format default present
TorchLlmArgs definescheckpoint_formatwithdefault=Noneand a post-validator that sets it to"HF"when unset, so threading it intocreate_input_processoris safe.
|
/bot run |
|
PR_Github #20894 [ run ] triggered by Bot |
|
PR_Github #20894 [ run ] completed with state |
|
/bot run |
|
PR_Github #20906 [ run ] triggered by Bot |
|
PR_Github #20906 [ run ] completed with state |
|
/bot run --disable-fail-fast |
|
PR_Github #20912 [ run ] triggered by Bot |
|
PR_Github #20912 [ run ] completed with state |
Funatiq
force-pushed
the
fix/custom_loading
branch
from
d95478f to
ecb5914
Compare
|
/bot run |
|
PR_Github #21029 [ run ] triggered by Bot |
|
PR_Github #21029 [ run ] completed with state |
|
/bot run --disable-fail-fast |
|
PR_Github #21035 [ run ] triggered by Bot |
|
PR_Github #21035 [ run ] completed with state |
|
PR_Github #21699 [ run ] completed with state |
|
/bot run --disable-fail-fast |
|
PR_Github #21897 [ run ] triggered by Bot. Commit: |
|
/bot run |
|
PR_Github #21907 [ run ] triggered by Bot. Commit: |
|
PR_Github #21897 [ run ] completed with state |
|
PR_Github #21907 [ run ] completed with state |
|
/bot run |
|
PR_Github #22032 [ run ] triggered by Bot. Commit: |
… format - Modified the create_input_processor function to accept a checkpoint_format parameter, defaulting to "HF". - Add detailed parameter descriptions and return type clarification. - The function now conditionally attempts to load the model configuration based on the specified format. Signed-off-by: Robin Kobus <[email protected]>
- Added debug logging for exceptions when loading the HF model configuration. - Included a fallback message when skipping the HF config load based on checkpoint format. Signed-off-by: Robin Kobus <[email protected]>
…ading - Get checkpoint_format in MultimodalEncoder and pass it to create_input_processor. Signed-off-by: Robin Kobus <[email protected]>
…tion in LLMArgs Signed-off-by: Robin Kobus <[email protected]>
Funatiq
force-pushed
the
fix/custom_loading
branch
from
dad1982 to
f288868
Compare
|
/bot run |
|
PR_Github #22050 [ run ] triggered by Bot. Commit: |
|
PR_Github #22032 [ run ] completed with state |
|
PR_Github #22050 [ run ] completed with state |
|
/bot run |
|
PR_Github #22066 [ run ] triggered by Bot. Commit: |
|
PR_Github #22066 [ run ] completed with state |
hchings
left a comment
hchings
left a comment
There was a problem hiding this comment.
Left a nit. Otherwise LGTM.
…ckpoint format (NVIDIA#7811) Signed-off-by: Robin Kobus <[email protected]> Signed-off-by: yufeiwu-nv <[email protected]>
…ckpoint format (NVIDIA#7811) Signed-off-by: Robin Kobus <[email protected]>
…ckpoint format (NVIDIA#7811) Signed-off-by: Robin Kobus <[email protected]>
…ckpoint format (NVIDIA#7811) Signed-off-by: Robin Kobus <[email protected]>
…ckpoint format (NVIDIA#7811) Signed-off-by: Robin Kobus <[email protected]>
4 participants
Summary by CodeRabbit
New Features
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
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.
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.