[Bug]TypeError: class `MMArchitectureQuant` in mmrazor/models/algorithms/quantization/mm_architecture.py: _fuse_fx() got an unexpected keyword argument 'graph_module'

[PancakeAwesome]

Describe the bug

A clear and concise description of what the bug is.

Traceback (most recent call last): File "/opt/conda/envs/torch2/lib/python3.8/site-packages/mmengine/registry/build_functions.py", line 122, in build_from_cfg obj = obj_cls(**args) # type: ignore File "/mnt/jingcheng/mmrazor/mmrazor/models/algorithms/quantization/mm_architecture.py", line 90, in init self.qmodels = self._build_qmodels(self.architecture) File "/mnt/jingcheng/mmrazor/mmrazor/models/algorithms/quantization/mm_architecture.py", line 297, in _build_qmodels observed_module = self.quantizer.prepare( File "/mnt/jingcheng/mmrazor/mmrazor/models/quantizers/native_quantizer.py", line 238, in prepare graph_module = _fuse_fx( TypeError: _fuse_fx() got an unexpected keyword argument 'graph_module'

During handling of the above exception, another exception occurred:

Traceback (most recent call last): File "tools/ptq.py", line 73, in main() File "tools/ptq.py", line 66, in main runner = Runner.from_cfg(cfg) File "/opt/conda/envs/torch2/lib/python3.8/site-packages/mmengine/runner/runner.py", line 439, in from_cfg runner = cls( File "/opt/conda/envs/torch2/lib/python3.8/site-packages/mmengine/runner/runner.py", line 406, in init self.model = self.build_model(model) File "/opt/conda/envs/torch2/lib/python3.8/site-packages/mmengine/runner/runner.py", line 813, in build_model model = MODELS.build(model) File "/opt/conda/envs/torch2/lib/python3.8/site-packages/mmengine/registry/registry.py", line 548, in build return self.build_func(cfg, *args, **kwargs, registry=self) File "/opt/conda/envs/torch2/lib/python3.8/site-packages/mmengine/registry/build_functions.py", line 250, in build_model_from_cfg return build_from_cfg(cfg, registry, default_args) File "/opt/conda/envs/torch2/lib/python3.8/site-packages/mmengine/registry/build_functions.py", line 144, in build_from_cfg raise type(e)( TypeError: class MMArchitectureQuant in mmrazor/models/algorithms/quantization/mm_architecture.py: _fuse_fx() got an unexpected keyword argument 'graph_module'

To Reproduce

The command you executed.

python tools/ptq.py my_configs/ptq_openvino_yolox_s_1xb32-400e_coco_calib32xb32.py

Post related information

1. The output of pip list | grep "mmcv\|mmrazor\|^torch"

06/13 15:10:00 - mmengine - INFO - PyTorch: 2.0.0+cu117 06/13 14:41:19 - mmengine - INFO - TorchVision: 0.15.1+cu117 06/13 14:41:19 - mmengine - INFO - OpenCV: 4.7.0 06/13 14:41:19 - mmengine - INFO - MMEngine: 0.7.3 06/13 14:41:19 - mmengine - INFO - MMCV: 2.0.0 06/13 14:41:19 - mmengine - INFO - MMCV Compiler: GCC 9.3 06/13 14:41:19 - mmengine - INFO - MMCV CUDA Compiler: 11.7 06/13 14:41:19 - mmengine - INFO - MMDeploy: 1.1.0+ 06/13 14:41:19 - mmengine - INFO -

06/13 14:41:19 - mmengine - INFO - Backend information 06/13 14:41:19 - mmengine - INFO - tensorrt: 8.2.3.0 06/13 14:41:19 - mmengine - INFO - tensorrt custom ops: Available 06/13 14:41:19 - mmengine - INFO - ONNXRuntime: 1.14.1 06/13 14:41:19 - mmengine - INFO - ONNXRuntime-gpu: 1.8.1 06/13 14:41:19 - mmengine - INFO - ONNXRuntime custom ops: NotAvailable 06/13 14:41:19 - mmengine - INFO - pplnn: None 06/13 14:41:19 - mmengine - INFO - ncnn: 1.0.20230517 06/13 14:41:19 - mmengine - INFO - ncnn custom ops: NotAvailable 06/13 14:41:19 - mmengine - INFO - snpe: None 06/13 14:41:19 - mmengine - INFO - openvino: 2023.0.0 06/13 14:41:19 - mmengine - INFO - torchscript: 2.0.0 06/13 14:41:19 - mmengine - INFO - torchscript custom ops: NotAvailable 06/13 14:41:19 - mmengine - INFO - rknn-toolkit: None 06/13 14:41:19 - mmengine - INFO - rknn-toolkit2: None 06/13 14:41:19 - mmengine - INFO - ascend: None 06/13 14:41:19 - mmengine - INFO - coreml: None 06/13 14:41:19 - mmengine - INFO - tvm: None 06/13 14:41:19 - mmengine - INFO - vacc: None 06/13 14:41:19 - mmengine - INFO -

06/13 14:41:19 - mmengine - INFO - Codebase information 06/13 14:41:19 - mmengine - INFO - mmdet: 3.0.0 06/13 14:41:19 - mmengine - INFO - mmseg: None 06/13 14:41:19 - mmengine - INFO - mmpretrain: 1.0.0rc8 06/13 14:41:19 - mmengine - INFO - mmocr: 1.0.0 06/13 14:41:19 - mmengine - INFO - mmagic: None 06/13 14:41:19 - mmengine - INFO - mmdet3d: None 06/13 14:41:19 - mmengine - INFO - mmpose: None 06/13 14:41:19 - mmengine - INFO - mmrotate: None 06/13 14:41:19 - mmengine - INFO - mmaction: None 06/13 14:41:19 - mmengine - INFO - mmrazor: 1.0.0 3. Your config file if you modified it or created a new one.

[here]

4. Your train log file if you meet the problem during training. [here]
5. Other code you modified in the mmrazor folder. [here]

Additional context

Add any other context about the problem here.

[here]

[pppppM]

I apologize for the inconvenience, but I was unable to reproduce the issue.

If the #554 is unresolved, it should not reach the _fuse_fx step. Perhaps you have made some other modifications?

You can try following the solution in #554 first and see if you still encounter this problem.

[PancakeAwesome]

I apologize for the inconvenience, but I was unable to reproduce the issue.

If the #554 is unresolved, it should not reach the _fuse_fx step. Perhaps you have made some other modifications?

You can try following the solution in #554 first and see if you still encounter this problem.

ok thanks.

[saeedadeeb103]

@PancakeAwesome i faced this Error before its beacuse ur using torch 2.0 and they updated _fuse_fx from

def _fuse_fx( graph_module: GraphModule, is_qat: bool, fuse_custom_config: Union[FuseCustomConfig, Dict[str, Any], None] = None, backend_config: Union[BackendConfig, Dict[str, Any], None] = None, ) -> GraphModule:

to ->

def _fuse_fx( model: GraphModule, is_qat: bool, fuse_custom_config: Union[FuseCustomConfig, Dict[str, Any], None] = None, backend_config: Union[BackendConfig, Dict[str, Any], None] = None, ) -> GraphModule:

so just change it from inside native_quantizer.py line 238 and it should work