Assertion `is_tensor()' failed. onnx2tensorrt [Bug]

[shuyuan-wang]

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

I used the command below with use of a bin file from nuscenes mini dataset. It shows Asssertion 'is_tensor()' faield when converting from onnx to tensorrt

Reproduction

python tools/deploy.py configs/mmdet3d/voxel-detection/voxel-detection_tensorrt_dynamic-nus-64x4.py ../../wsy/mmdetection3d/configs/pointpillars/hv_pointpillars_fpn_sbn-all_4x8_2x_nus-3d.py ./hv_pointpillars_fpn_sbn-all_4x8_2x_nus-3d_20210826_104936-fca299c1.pth ./n008-2018-08-01-15-16-36-0400__LIDAR_TOP__1533151603597909.pcd.bin --work-dir ./work_dir --show --device cuda:0

Environment

2022-10-21 03:48:01,879 - mmdeploy - INFO - 

2022-10-21 03:48:01,879 - mmdeploy - INFO - **********Environmental information**********
2022-10-21 03:48:02,097 - mmdeploy - INFO - sys.platform: linux
2022-10-21 03:48:02,097 - mmdeploy - INFO - Python: 3.8.13 (default, Mar 28 2022, 11:38:47) [GCC 7.5.0]
2022-10-21 03:48:02,097 - mmdeploy - INFO - CUDA available: True
2022-10-21 03:48:02,097 - mmdeploy - INFO - GPU 0: NVIDIA GeForce RTX 3090
2022-10-21 03:48:02,097 - mmdeploy - INFO - CUDA_HOME: /usr/local/cuda
2022-10-21 03:48:02,097 - mmdeploy - INFO - NVCC: Cuda compilation tools, release 11.6, V11.6.124
2022-10-21 03:48:02,097 - mmdeploy - INFO - GCC: gcc (Ubuntu 9.4.0-1ubuntu1~20.04.1) 9.4.0
2022-10-21 03:48:02,097 - mmdeploy - INFO - PyTorch: 1.10.0
2022-10-21 03:48:02,097 - mmdeploy - INFO - PyTorch compiling details: PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.2 Product Build 20200624 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: AVX2
  - CUDA Runtime 11.3
  - 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_61,code=sm_61;-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;-gencode;arch=compute_37,code=compute_37
  - CuDNN 8.2
  - Magma 2.5.2
  - Build settings: BLAS_INFO=mkl, BUILD_TYPE=Release, CUDA_VERSION=11.3, CUDNN_VERSION=8.2.0, 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, 

2022-10-21 03:48:02,097 - mmdeploy - INFO - TorchVision: 0.11.0
2022-10-21 03:48:02,097 - mmdeploy - INFO - OpenCV: 4.5.4
2022-10-21 03:48:02,097 - mmdeploy - INFO - MMCV: 1.6.0
2022-10-21 03:48:02,097 - mmdeploy - INFO - MMCV Compiler: GCC 9.4
2022-10-21 03:48:02,097 - mmdeploy - INFO - MMCV CUDA Compiler: 11.6
2022-10-21 03:48:02,097 - mmdeploy - INFO - MMDeploy: 0.9.0+b2cb572
2022-10-21 03:48:02,097 - mmdeploy - INFO - 

2022-10-21 03:48:02,097 - mmdeploy - INFO - **********Backend information**********
2022-10-21 03:48:02,527 - mmdeploy - INFO - onnxruntime: 1.8.1  ops_is_avaliable : True
2022-10-21 03:48:02,554 - mmdeploy - INFO - tensorrt: 8.2.4.2   ops_is_avaliable : True
2022-10-21 03:48:02,569 - mmdeploy - INFO - ncnn: None  ops_is_avaliable : False
2022-10-21 03:48:02,569 - mmdeploy - INFO - pplnn_is_avaliable: False
2022-10-21 03:48:02,570 - mmdeploy - INFO - openvino_is_avaliable: False
2022-10-21 03:48:02,583 - mmdeploy - INFO - snpe_is_available: False
2022-10-21 03:48:02,584 - mmdeploy - INFO - ascend_is_available: False
2022-10-21 03:48:02,584 - mmdeploy - INFO - coreml_is_available: False
2022-10-21 03:48:02,584 - mmdeploy - INFO - 

2022-10-21 03:48:02,584 - mmdeploy - INFO - **********Codebase information**********
2022-10-21 03:48:02,587 - mmdeploy - INFO - mmdet:      2.25.2
2022-10-21 03:48:02,587 - mmdeploy - INFO - mmseg:      0.29.0
2022-10-21 03:48:02,587 - mmdeploy - INFO - mmcls:      0.22.1
2022-10-21 03:48:02,587 - mmdeploy - INFO - mmocr:      None
2022-10-21 03:48:02,587 - mmdeploy - INFO - mmedit:     0.15.2
2022-10-21 03:48:02,587 - mmdeploy - INFO - mmdet3d:    1.0.0rc5
2022-10-21 03:48:02,587 - mmdeploy - INFO - mmpose:     0.25.1
2022-10-21 03:48:02,587 - mmdeploy - INFO - mmrotate:   None

Error traceback

python: /root/gpgpu/MachineLearning/myelin/src/compiler/./ir/operand.h:166: myelin::ir::tensor_t*& myelin::ir::operand_t::tensor(): Assertion `is_tensor()' failed.
2022-10-21 03:43:40,509 - mmdeploy - ERROR - `mmdeploy.backend.tensorrt.onnx2tensorrt.onnx2tensorrt` with Call id: 1 failed. exit.

[tpoisonooo]

please check NVIDIA/TensorRT#1541 and try update to trt8.5 with docker.

trt8.4 still has a bug about dynamic shape.

[tpoisonooo]

Also check NVIDIA/TensorRT#2338.

[tpoisonooo]

Or downgrade to cu102+trt8.4, I have tested this combination and passed.

[Capchenxi]

Is there anyway to fix this bug or check which op caused this problem? I am not allowed to update TRT version under my environment. Thanks.

[tpoisonooo]

TRT is actually not an open source project, I think there is nothing I can do.

NVIDIA/TensorRT#1541

[Jason-cs18]

TRT is actually not an open source project, I think there is nothing I can do.

NVIDIA/TensorRT#1541

Hello @tpoisonooo , can you provide a new Dockerfile (trt 8.4 and xxx)? I want to build mmdeploy via docker but I find that existing Dockerfile of master still uses TRT 8.2

[github-actions]

This issue is marked as stale because it has been marked as invalid or awaiting response for 7 days without any further response. It will be closed in 5 days if the stale label is not removed or if there is no further response.

[github-actions]

This issue is closed because it has been stale for 5 days. Please open a new issue if you have similar issues or you have any new updates now.