benz725 edited a comment on issue #19731:
URL:
https://github.com/apache/incubator-mxnet/issues/19731#issuecomment-757115024
> Did you run this successfully:
>
> ```
> # Build ONNX
> pushd .
> echo "Installing ONNX."
> cd 3rdparty/onnx-tensorrt/third_party/onnx
> rm -rf build
> mkdir -p build
> cd build
> cmake -DCMAKE_CXX_FLAGS=-I/usr/include/python${PYVER}
-DBUILD_SHARED_LIBS=ON ..
> make -j$(nproc)
> export LIBRARY_PATH=`pwd`:`pwd`/onnx/:$LIBRARY_PATH
> export CPLUS_INCLUDE_PATH=`pwd`:$CPLUS_INCLUDE_PATH
> export CXXFLAGS=-I`pwd`
>
> popd
>
> # Build ONNX-TensorRT
> export LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:/usr/local/lib
> export
CPLUS_INCLUDE_PATH=${CPLUS_INCLUDE_PATH}:/usr/local/cuda-10.2/targets/x86_64-linux/include/
> pushd .
> cd 3rdparty/onnx-tensorrt/
> mkdir -p build
> cd build
> cmake -DONNX_NAMESPACE=$ONNX_NAMESPACE ..
> make -j$(nproc)
> export LIBRARY_PATH=`pwd`:$LIBRARY_PATH
> popd
>
> mkdir -p /work/mxnet/lib/
> cp 3rdparty/onnx-tensorrt/third_party/onnx/build/*.so
/work/mxnet/lib/
> cp -L 3rdparty/onnx-tensorrt/build/libnvonnxparser.so /work/mxnet/lib/
> ```
>
> If yes, what is the outputs of :
>
> ```
> ls 3rdparty/onnx-tensorrt/third_party/onnx/onnx
> ```
>
> and finally, can you verify if this path is part of `$LIBRARY_PATH`?
I will tell what I have done:
First, I download onnx-tensorrt from "https://github.com/onnx/onnx-tensorrt";
to my /home dir, but in this project, the onnx-tensorrt/third_party/onnx/
directory is empty, so I download the onnx from
"https://github.com/onnx/onnx/tree/553df22c67bee5f0fe6599cff60f1afc6748c635";
and put it to the onnx-tensorrt/third_party/onnx/ dir.
Then I follow the onnx-tensorrt official tutorial, do the compile:
cd onnx-tensorrt
mkdir build && cd build
cmake .. -DTENSORRT_ROOT=<path_to_trt> && make -j
// Ensure that you update your LD_LIBRARY_PATH to pick up the location of
the newly built library:
export LD_LIBRARY_PATH=$PWD:$LD_LIBRARY_PATH
It can compile correctly , but I found nothing like named libonnx.so**. And
when I compile the mxnet with -DTENSORRT=1, It reported error again.
You said I must define some arguments , how many arguments should I have to
define?
Now I have the mxnet-src file in my home directory like :
/home/$USER_NAME/mxnet-src
and onnx-tensortrt in my home dir: /home/$USER_NAME/onnx-tensorrt
and onnx in the onnx-tensorrt directory
:/home/$USER_NAME/onnx-tensorrt/third_party/onnx/
How should I define the eviron arguments? You said I must follow the .sh
file in the docker file, but I'm not sure how many arguments I shoud to
define. Some args like ONNX_NAMESPACE, LIBRARY_PATH, LD_LIBRARY_PATH,
CPLUS_INCLUDE_PATH, CXXFLAGS, should I have to define it manually? And the
popd, pushd, should I have to do that like the .sh file did?
I'm not very good at do compiling and define args, so thanks for your
answering!!
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
