Onnx-tensorrt: failed to build

[Stephenfang51]

My command is
cmake .. DTENSORRT_ROOT=/input/TensorRT-6.0.1.5 -DGPU_ARCHS="61"

Generated: /input/onnx-tensorrt/build/third_party/onnx/onnx/onnx_onnx2trt_onnx-ml.proto
Generated: /input/onnx-tensorrt/build/third_party/onnx/onnx/onnx-operators_onnx2trt_onnx-ml.proto
--
-- ******** Summary ********
--   CMake version         : 3.15.5
--   CMake command         : /usr/local/bin/cmake
--   System                : Linux
--   C++ compiler          : /usr/bin/c++
--   C++ compiler version  : 5.4.0
--   CXX flags             :  -Wall -Wno-deprecated-declarations -Wno-unused-function -Wnon-virtual-dtor
--   Build type            : Release
--   Compile definitions   : ONNX_NAMESPACE=onnx2trt_onnx
--   CMAKE_PREFIX_PATH     :
--   CMAKE_INSTALL_PREFIX  : /usr/local
--   CMAKE_MODULE_PATH     :
--
--   ONNX version          : 1.6.0
--   ONNX NAMESPACE        : onnx2trt_onnx
--   ONNX_BUILD_TESTS      : OFF
--   ONNX_BUILD_BENCHMARKS : OFF
--   ONNX_USE_LITE_PROTO   : OFF
--   ONNXIFI_DUMMY_BACKEND : OFF
--   ONNXIFI_ENABLE_EXT    : OFF
--
--   Protobuf compiler     : /usr/bin/protoc
--   Protobuf includes     : /usr/include
--   Protobuf libraries    : /usr/lib/x86_64-linux-gnu/libprotobuf.so;-lpthread
--   BUILD_ONNX_PYTHON     : OFF
-- Found TensorRT headers at /input/TensorRT-6.0.1.5/include
-- Find TensorRT libs at /input/TensorRT-6.0.1.5/lib/libnvinfer.so;/input/TensorRT-6.0.1.5/lib/libnvinfer_plugin.so;TENSORRT_LIBRARY_MYELIN-NOTFOUND
-- Could NOT find TENSORRT (missing: TENSORRT_LIBRARY)
ERRORCannot find TensorRT library.
CMake Error: The following variables are used in this project, but they are set to NOTFOUND.
Please set them or make sure they are set and tested correctly in the CMake files:
TENSORRT_LIBRARY_MYELIN
    linked by target "nvonnxparser_static" in directory /input/onnx-tensorrt
    linked by target "nvonnxparser" in directory /input/onnx-tensorrt

-- Configuring incomplete, errors occurred!
See also "/input/onnx-tensorrt/build/CMakeFiles/CMakeOutput.log".
See also "/input/onnx-tensorrt/build/CMakeFiles/CMakeError.log".

Any solution for this problem ?

Thanks

[Semihal]

@Stephenfang51 , hi!

Try it:

cd onnx-tensorrt
cmake . -DCUDA_INCLUDE_DIRS=/usr/local/cuda/include -DTENSORRT_ROOT=/usr/src/tensorrt -DCMAKE_CUDA_COMPILER=/usr/local/cuda/bin/nvcc
sudo make install

[Stephenfang51]

@Semihal Dear Sir

I have tried your command and it shows

-- The CXX compiler identification is GNU 5.4.0
-- The C compiler identification is GNU 5.4.0
-- Check for working CXX compiler: /usr/bin/c++
-- Check for working CXX compiler: /usr/bin/c++ -- works
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Check for working C compiler: /usr/bin/cc
-- Check for working C compiler: /usr/bin/cc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Found Protobuf: /usr/lib/x86_64-linux-gnu/libprotobuf.so;-lpthread (found version "2.6.1")
-- Build type not set - defaulting to Release
Generated: /input/onnx-tensorrt/third_party/onnx/onnx/onnx_onnx2trt_onnx-ml.proto
Generated: /input/onnx-tensorrt/third_party/onnx/onnx/onnx-operators_onnx2trt_onnx-ml.proto
--
-- ******** Summary ********
--   CMake version         : 3.15.5
--   CMake command         : /usr/local/bin/cmake
--   System                : Linux
--   C++ compiler          : /usr/bin/c++
--   C++ compiler version  : 5.4.0
--   CXX flags             :  -Wall -Wno-deprecated-declarations -Wno-unused-function -Wnon-virtual-dtor
--   Build type            : Release
--   Compile definitions   : ONNX_NAMESPACE=onnx2trt_onnx
--   CMAKE_PREFIX_PATH     :
--   CMAKE_INSTALL_PREFIX  : /usr/local
--   CMAKE_MODULE_PATH     :
--
--   ONNX version          : 1.6.0
--   ONNX NAMESPACE        : onnx2trt_onnx
--   ONNX_BUILD_TESTS      : OFF
--   ONNX_BUILD_BENCHMARKS : OFF
--   ONNX_USE_LITE_PROTO   : OFF
--   ONNXIFI_DUMMY_BACKEND : OFF
--   ONNXIFI_ENABLE_EXT    : OFF
--
--   Protobuf compiler     : /usr/bin/protoc
--   Protobuf includes     : /usr/include
--   Protobuf libraries    : /usr/lib/x86_64-linux-gnu/libprotobuf.so;-lpthread
--   BUILD_ONNX_PYTHON     : OFF
-- Found TensorRT headers at TENSORRT_INCLUDE_DIR-NOTFOUND
-- Find TensorRT libs at TENSORRT_LIBRARY_INFER-NOTFOUND;TENSORRT_LIBRARY_INFER_PLUGIN-NOTFOUND;TENSORRT_LIBRARY_MYELIN-NOTFOUND
-- Could NOT find TENSORRT (missing: TENSORRT_INCLUDE_DIR TENSORRT_LIBRARY)
ERRORCannot find TensorRT library.
CMake Error: The following variables are used in this project, but they are set to NOTFOUND.
Please set them or make sure they are set and tested correctly in the CMake files:
/input/onnx-tensorrt/TENSORRT_INCLUDE_DIR
   used as include directory in directory /input/onnx-tensorrt
   used as include directory in directory /input/onnx-tensorrt
TENSORRT_LIBRARY_INFER
    linked by target "nvonnxparser_static" in directory /input/onnx-tensorrt
    linked by target "nvonnxparser" in directory /input/onnx-tensorrt
TENSORRT_LIBRARY_INFER_PLUGIN
    linked by target "nvonnxparser_static" in directory /input/onnx-tensorrt
    linked by target "nvonnxparser" in directory /input/onnx-tensorrt
TENSORRT_LIBRARY_MYELIN
    linked by target "nvonnxparser_static" in directory /input/onnx-tensorrt
    linked by target "nvonnxparser" in directory /input/onnx-tensorrt

-- Configuring incomplete, errors occurred!

[Semihal]

Please print the result: dpkg -l | grep Tensor

[Stephenfang51]

dpkg -l | grep Tensor

Well, it printed nothing ....

I can import tensorrt in my python shell which means tensorRT installed correctly right?

here is my bashrc

# added by Miniconda3 installer
export PATH="/usr/local/miniconda3/bin:$PATH"
alias sa="source activate"
alias tb="setsid setsid /usr/local/miniconda3/envs/dl/bin/python /usr/local/miniconda3/envs/dl/bin/tensorboard --logdir=/output --port=28007"
export LC_ALL=zh_CN.UTF-8
export LANG=zh_CN.UTF-8
export LANGUAGE=zh_CN.UTF-8
export PATH="/usr/local/cuda/bin:$PATH"
export LD_LIBRARY_PATH="/usr/local/miniconda3/envs/dl/lib:/usr/local/cuda/extras/CUPTI/lib64:$LD_LIBRARY_PATH"
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/input/TensorRT-6.0.1.5/lib

I can give you some clues, I installed tensorRT in conda env, maybe that's the reason can't find the TensorRT library ??

[Semihal]

Oh, sorry, I didn't work with anaconda.

[Stephenfang51]

Oh, sorry, I didn't work with anaconda.

That's okay, Thanks anyway

[1LOVESJohnny]

@Stephenfang51 Hi, have you solved the problem? I met the same error. Expect for your reply.

[1LOVESJohnny]

@Stephenfang51 Hi, I've figured it out. This problem is caused by unmatched TRT version. I uninstalled the original TRT6.0 and installed TRT7.0. The problem is solved.

[Stephenfang51]

@Stephenfang51 Hi, I've figured it out. This problem is caused by unmatched TRT version. I uninstalled the original TRT6.0 and installed TRT7.0. The problem is solved.

Wow it doesn't make sense....
all of your error messsge are same with mine ?

[rmccorm4]

Hi @Stephenfang51,

The fact that it's looking for this TENSORRT_LIBRARY_MYELIN means this is specific to TensorRT 7.

Similarly, I believe only TRT7 supports ONNX 1.6: ONNX version : 1.6.0.

I haven't tried building this myself, but you'll likely need to either install TRT7 to build master / 7.0 branch, or keep TRT6 and build from 6.0 / 6.0-full-dims branch: https://github.com/onnx/onnx-tensorrt/tree/6.0-full-dims

[Stephenfang51]

@1LOVESJohnny @rmccorm4 Yes, problem solved. Thank you guys so much !