[Bug] assert self.num_levels == len(featmap_sizes)

[Hevislee]

Checklist

• I have searched related issues but cannot get the expected help.
• 2. I have read the FAQ documentation but cannot get the expected help.
• 3. The bug has not been fixed in the latest version.

Describe the bug

When I was running "python tools/test.py", an error occurred: assert self.num_levels == len(featmap_sizes)
AssertionError.
This error occurs when I run "python tools/test.py" to test pointpillars model, this does not happen when testing centerpoint.

Reproduction

python tools/test.py /home/mmdeploy/configs/mmdet3d/voxel-detection/voxel-detection_onnxruntime_dynamic.py /home/mmdet3d-v1.1/configs/pointpillars/pointpillars_hv_fpn_sbn-all_8xb4-2x_nus-3d.py --model /home/Pointpillars/end2end.onnx --device cuda:0 --speed-test

Environment

06/20 12:47:45 - mmengine - INFO - 06/20 12:47:45 - mmengine - INFO - **********Environmental information**********06/20 12:47:45 - mmengine - INFO - sys.platform: linux06/20 12:47:45 - mmengine - INFO - Python: 3.8.16 | packaged by conda-forge | (default, Feb  1 2023, 16:01:55) [GCC 11.3.0]
06/20 12:47:45 - mmengine - INFO - CUDA available: True
06/20 12:47:45 - mmengine - INFO - numpy_random_seed: 2147483648
06/20 12:47:45 - mmengine - INFO - GPU 0,1,2,3: NVIDIA GeForce RTX 3090
06/20 12:47:45 - mmengine - INFO - CUDA_HOME: /usr/local/cuda-11.1
06/20 12:47:45 - mmengine - INFO - NVCC: Cuda compilation tools, release 11.1, V11.1.105
06/20 12:47:45 - mmengine - INFO - GCC: gcc (Ubuntu 9.4.0-1ubuntu1~20.04.1) 9.4.0
06/20 12:47:45 - mmengine - INFO - PyTorch: 1.10.0+cu111
06/20 12:47:45 - mmengine - INFO - PyTorch compiling details: PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v2.2.3 (Git Hash 7336ca9f055cf1bfa13efb658fe15dc9b41f0740)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - LAPACK is enabled (usually provided by MKL)
  - NNPACK is enabled
  - CPU capability usage: AVX512
  - CUDA Runtime 11.1
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75;-gencode;arch=compute_80,code=sm_80;-gencode;arch=compute_86,code=sm_86
  - CuDNN 8.0.5
  - Magma 2.5.2
  - Build settings: BLAS_INFO=mkl, BUILD_TYPE=Release, CUDA_VERSION=11.1, CUDNN_VERSION=8.0.5, CXX_COMPILER=/opt/rh/devtoolset-7/root/usr/bin/c++, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_KINETO -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DSYMBOLICATE_MOBILE_DEBUG_HANDLE -DEDGE_PROFILER_USE_KINETO -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, LAPACK_INFO=mkl, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, TORCH_VERSION=1.10.0, USE_CUDA=ON, USE_CUDNN=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

06/20 12:47:45 - mmengine - INFO - TorchVision: 0.11.0+cu111
06/20 12:47:45 - mmengine - INFO - OpenCV: 4.7.0
06/20 12:47:45 - mmengine - INFO - MMEngine: 0.7.3
06/20 12:47:45 - mmengine - INFO - MMCV: 2.0.0
06/20 12:47:45 - mmengine - INFO - MMCV Compiler: GCC 7.3
06/20 12:47:45 - mmengine - INFO - MMCV CUDA Compiler: 11.1
06/20 12:47:45 - mmengine - INFO - MMDeploy: 1.1.0+439f88b
06/20 12:47:45 - mmengine - INFO - 

06/20 12:47:45 - mmengine - INFO - **********Backend information**********
06/20 12:47:45 - mmengine - INFO - tensorrt:    8.4.2.4
06/20 12:47:45 - mmengine - INFO - tensorrt custom ops: Available
06/20 12:47:45 - mmengine - INFO - ONNXRuntime: None
06/20 12:47:45 - mmengine - INFO - ONNXRuntime-gpu:     1.8.1
06/20 12:47:45 - mmengine - INFO - ONNXRuntime custom ops:      Available
06/20 12:47:45 - mmengine - INFO - pplnn:       None
06/20 12:47:45 - mmengine - INFO - ncnn:        None
06/20 12:47:45 - mmengine - INFO - snpe:        None
06/20 12:47:45 - mmengine - INFO - openvino:    None
06/20 12:47:45 - mmengine - INFO - torchscript: 1.10.0+cu111
06/20 12:47:45 - mmengine - INFO - torchscript custom ops:      NotAvailable
06/20 12:47:45 - mmengine - INFO - rknn-toolkit:        None
06/20 12:47:45 - mmengine - INFO - rknn-toolkit2:       None
06/20 12:47:45 - mmengine - INFO - ascend:      None
06/20 12:47:45 - mmengine - INFO - coreml:      None
06/20 12:47:45 - mmengine - INFO - tvm: None
06/20 12:47:45 - mmengine - INFO - vacc:        None
06/20 12:47:45 - mmengine - INFO - 

06/20 12:47:45 - mmengine - INFO - **********Codebase information**********
06/20 12:47:45 - mmengine - INFO - mmdet:       3.0.0
06/20 12:47:45 - mmengine - INFO - mmseg:       None
06/20 12:47:45 - mmengine - INFO - mmpretrain:  None
06/20 12:47:45 - mmengine - INFO - mmocr:       None
06/20 12:47:45 - mmengine - INFO - mmagic:      None
06/20 12:47:45 - mmengine - INFO - mmdet3d:     1.1.0
06/20 12:47:45 - mmengine - INFO - mmpose:      None
06/20 12:47:45 - mmengine - INFO - mmrotate:    None
06/20 12:47:45 - mmengine - INFO - mmaction:    None
06/20 12:47:45 - mmengine - INFO - mmrazor:     None

Error traceback

/home/mmdetection3d/mmdet3d/models/dense_heads/anchor3d_head.py:92: UserWarning: dir_offset and dir_limit_offset will be depressed and be incorporated into box coder in the future
  warnings.warn(
Traceback (most recent call last):
  File "tools/test.py", line 159, in <module>
    main()
  File "tools/test.py", line 150, in main
    runner.test()
  File "/home/anaconda3/envs/mmdet3d/lib/python3.8/site-packages/mmengine/runner/runner.py", line 1767, in test
    metrics = self.test_loop.run()  # type: ignore
  File "/home/anaconda3/envs/mmdet3d/lib/python3.8/site-packages/mmengine/runner/loops.py", line 435, in run
    self.run_iter(idx, data_batch)
  File "/home/anaconda3/envs/mmdet3d/lib/python3.8/site-packages/torch/autograd/grad_mode.py", line 28, in decorate_context
    return func(*args, **kwargs)
  File "/home/anaconda3/envs/mmdet3d/lib/python3.8/site-packages/mmengine/runner/loops.py", line 454, in run_iter
    outputs = self.runner.model.test_step(data_batch)
  File "/home/anaconda3/envs/mmdet3d/lib/python3.8/site-packages/mmengine/model/base_model/base_model.py", line 145, in test_step
    return self._run_forward(data, mode='predict')  # type: ignore
  File "/home/anaconda3/envs/mmdet3d/lib/python3.8/site-packages/mmengine/model/base_model/base_model.py", line 340, in _run_forward
    results = self(**data, mode=mode)
  File "/home/anaconda3/envs/mmdet3d/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
    return forward_call(*input, **kwargs)
  File "/home/anaconda3/envs/mmdet3d/lib/python3.8/site-packages/mmdeploy/codebase/mmdet3d/deploy/voxel_detection_model.py", line 97, in forward
    prediction = VoxelDetectionModel.postprocess(
  File "/home/anaconda3/envs/mmdet3d/lib/python3.8/site-packages/mmdeploy/codebase/mmdet3d/deploy/voxel_detection_model.py", line 245, in postprocess
    data_instances_3d = head.predict_by_feat(
  File "/home/mmdetection3d/mmdet3d/models/dense_heads/base_3d_dense_head.py", line 232, in predict_by_feat
    mlvl_priors = self.prior_generator.grid_anchors(
  File "/home/mmdetection3d/mmdet3d/models/task_modules/anchor/anchor_3d_generator.py", line 105, in grid_anchors
    assert self.num_levels == len(featmap_sizes)
AssertionError

[RunningLeon]

hi, are you using the official model config and ckpt from mmdet3d?

[Hevislee]

hi, are you using the official model config and ckpt from mmdet3d?

yes, i used the official model config and ckpt from mmdet3d.

[RunningLeon]

@Hevislee hi, there's a bug for fpn models. Here is a simple workaround and pls. change the three files according to the following codes diff or use my temporary branch.

diff --git a/configs/mmdet3d/voxel-detection/voxel-detection_static.py b/configs/mmdet3d/voxel-detection/voxel-detection_static.py
index bba7e819..7f44d78c 100644
--- a/configs/mmdet3d/voxel-detection/voxel-detection_static.py
+++ b/configs/mmdet3d/voxel-detection/voxel-detection_static.py
@@ -3,4 +3,7 @@ codebase_config = dict(
     type='mmdet3d', task='VoxelDetection', model_type='end2end')
 onnx_config = dict(
     input_names=['voxels', 'num_points', 'coors'],
-    output_names=['cls_score', 'bbox_pred', 'dir_cls_pred'])
+    output_names=['cls_score0', 'bbox_pred0', 'dir_cls_pred0',
+                  'cls_score1', 'bbox_pred1', 'dir_cls_pred1',
+                  'cls_score2', 'bbox_pred2', 'dir_cls_pred2',
+                  ])
diff --git a/mmdeploy/codebase/mmdet3d/deploy/voxel_detection_model.py b/mmdeploy/codebase/mmdet3d/deploy/voxel_detection_model.py
index 949f2902..bbbf65f7 100644
--- a/mmdeploy/codebase/mmdet3d/deploy/voxel_detection_model.py
+++ b/mmdeploy/codebase/mmdet3d/deploy/voxel_detection_model.py
@@ -90,7 +90,10 @@ class VoxelDetectionModel(BaseBackendModel):
         }
 
         outputs = self.wrapper(input_dict)
-
+        new_outputs = dict(cls_score=[outputs[f'cls_score{i}'] for i in range(3)],
+                           bbox_pred=[outputs[f'bbox_pred{i}'] for i in range(3)],
+                           dir_cls_pred=[outputs[f'dir_cls_pred{i}'] for i in range(3)])
+        outputs = new_outputs
         if data_samples is None:
             return outputs
 
@@ -239,9 +242,9 @@ class VoxelDetectionModel(BaseBackendModel):
 
         if not hasattr(head, 'task_heads'):
             data_instances_3d = head.predict_by_feat(
-                cls_scores=[cls_score],
-                bbox_preds=[bbox_pred],
-                dir_cls_preds=[dir_cls_pred],
+                cls_scores=cls_score,
+                bbox_preds=bbox_pred,
+                dir_cls_preds=dir_cls_pred,
                 batch_input_metas=batch_input_metas,
                 cfg=cfg)
 
diff --git a/mmdeploy/codebase/mmdet3d/models/mvx_two_stage.py b/mmdeploy/codebase/mmdet3d/models/mvx_two_stage.py
index 12df74ff..632242ed 100644
--- a/mmdeploy/codebase/mmdet3d/models/mvx_two_stage.py
+++ b/mmdeploy/codebase/mmdet3d/models/mvx_two_stage.py
@@ -82,5 +82,8 @@ def mvxtwostagedetector__forward(self, inputs: list, **kwargs):
         dir_scores = torch.cat(dir_scores, dim=1)
         return scores, bbox_preds, dir_scores
     else:
-        cls_score, bbox_pred, dir_cls_pred = outs[0][0], outs[1][0], outs[2][0]
-        return cls_score, bbox_pred, dir_cls_pred
+        # cls_score, bbox_pred, dir_cls_pred = outs[0][0], outs[1][0], outs[2][0]
+        preds = []
+        for i in range(len(outs[0])):
+            preds += [outs[0][i], outs[1][i], outs[2][i]]
+        return tuple(preds)

[Hevislee]

Thanks for reply!
I change the three files as above , but another error occurred：
Traceback (most recent call last):
File "tools/test.py", line 159, in
main()
File "tools/test.py", line 150, in main
runner.test()
File "/home/anaconda3/envs/mmdet3d/lib/python3.8/site-packages/mmengine/runner/runner.py", line 1767, in test
metrics = self.test_loop.run() # type: ignore
File "/home/anaconda3/envs/mmdet3d/lib/python3.8/site-packages/mmengine/runner/loops.py", line 435, in run
self.run_iter(idx, data_batch)
File "/home/anaconda3/envs/mmdet3d/lib/python3.8/site-packages/torch/autograd/grad_mode.py", line 28, in decorate_context
return func(*args, **kwargs)
File "/home/anaconda3/envs/mmdet3d/lib/python3.8/site-packages/mmengine/runner/loops.py", line 454, in run_iter
outputs = self.runner.model.test_step(data_batch)
File "/home/anaconda3/envs/mmdet3d/lib/python3.8/site-packages/mmengine/model/base_model/base_model.py", line 145, in test_step
return self._run_forward(data, mode='predict') # type: ignore
File "/home/anaconda3/envs/mmdet3d/lib/python3.8/site-packages/mmengine/model/base_model/base_model.py", line 340, in _run_forward
results = self(**data, mode=mode)
File "/home/anaconda3/envs/mmdet3d/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/home/mmdeploy/mmdeploy/codebase/mmdet3d/deploy/voxel_detection_model.py", line 92, in forward
outputs = self.wrapper(input_dict)
File "/home/anaconda3/envs/mmdet3d/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/home/mmdeploy/mmdeploy/backend/onnxruntime/wrapper.py", line 97, in forward
self.__ort_execute(self.io_binding)
File "/home/mmdeploy/mmdeploy/utils/timer.py", line 67, in fun
result = func(*args, **kwargs)
File "/home/mmdeploy/mmdeploy/backend/onnxruntime/wrapper.py", line 113, in __ort_execute
self.sess.run_with_iobinding(io_binding)
File "/home/anaconda3/envs/mmdet3d/lib/python3.8/site-packages/onnxruntime/capi/onnxruntime_inference_collection.py", line 229, in run_with_iobinding
self._sess.run_with_iobinding(iobinding._iobinding, run_options)
RuntimeError: Error in execution: Invalid Output Name:cls_score0

[RunningLeon]

you should rerun tools/deploy.py to do torch2onnx again.

[Hevislee]

you should rerun tools/deploy.py to do torch2onnx again.

Suddenly understand！Thanks a lot！