No "tensorrt" module in python3.8 ?
Akiautiste opened this issue · 10 comments
dear author,i can not import python package:"tensorrt",is there not include tensorrt pybind in this ubuntu20.04 image?
Unfortunately, it seems rather difficult to change the Python binding of TensorRT. I've been working on it for a while now, but no luck so far. TensorRT is an integral part of your JetPack and is not easy to install 'separately'. First of all, I don't see a solution to make TensorRT work with Python 3.8. If anyone has any suggestions, let me know!
Unfortunately, it seems rather difficult to change the Python binding of TensorRT. I've been working on it for a while now, but no luck so far. TensorRT is an integral part of your JetPack and is not easy to install 'separately'. First of all, I don't see a solution to make TensorRT work with Python 3.8. If anyone has any suggestions, let me know!
haha, i solved it...
I extract ubuntu18.04 jetson nano image file,can find tensorrt packages in path:/usr/lib/python3.6/dist-packages/tensorrt ,
Just copy tensorrt*/ to ubuntu20.04,of course ,need build&install python3.6 in ubuntu20.04.
Glad to hear you succeeded!
Hi,
got same issue and of course manage also to make it work with python 3.6.
But we need a way to have tensorrt with 3.8.
I see perhaps a solution with tiny tensorrt… really not sure !
I was wondering why I get the following error message:
I see perhaps a solution with tiny tensorrt… really not sure !
Does anyone perhaps have a solution?
Unfortunately, it seems rather difficult to change the Python binding of TensorRT. I've been working on it for a while now, but no luck so far. TensorRT is an integral part of your JetPack and is not easy to install 'separately'. First of all, I don't see a solution to make TensorRT work with Python 3.8. If anyone has any suggestions, let me know!
haha, i solved it... I extract ubuntu18.04 jetson nano image file,can find tensorrt packages in path:/usr/lib/python3.6/dist-packages/tensorrt , Just copy tensorrt*/ to ubuntu20.04,of course ,need build&install python3.6 in ubuntu20.04.
I copied tensorrt*, but no, do you change anything else?
Unfortunately, it seems rather difficult to change the Python binding of TensorRT. I've been working on it for a while now, but no luck so far. TensorRT is an integral part of your JetPack and is not easy to install 'separately'. First of all, I don't see a solution to make TensorRT work with Python 3.8. If anyone has any suggestions, let me know!
haha, i solved it... I extract ubuntu18.04 jetson nano image file,can find tensorrt packages in path:/usr/lib/python3.6/dist-packages/tensorrt , Just copy tensorrt*/ to ubuntu20.04,of course ,need build&install python3.6 in ubuntu20.04.
I copied tensorrt*, but no, do you change anything else?
He managed with python 3.6...
Unfortunately, it seems rather difficult to change the Python binding of TensorRT. I've been working on it for a while now, but no luck so far. TensorRT is an integral part of your JetPack and is not easy to install 'separately'. First of all, I don't see a solution to make TensorRT work with Python 3.8. If anyone has any suggestions, let me know!
haha, i solved it... I extract ubuntu18.04 jetson nano image file,can find tensorrt packages in path:/usr/lib/python3.6/dist-packages/tensorrt , Just copy tensorrt*/ to ubuntu20.04,of course ,need build&install python3.6 in ubuntu20.04.
I copied tensorrt*, but no, do you change anything else?
He managed with python 3.6...
Me too, but I can't use it
Found that you can pretty easily compile tensorrt from sources. I used the python3.8.10 shipped with the Image provided by Qengineering and followed these instructions to build tensorrt:
https://github.com/NVIDIA/TensorRT/tree/8.0.1#building-tensorrt-oss
cd $TRT_OSSPATH
mkdir -p build && cd build
cmake .. -DTRT_LIB_DIR=$TRT_LIBPATH -DTRT_OUT_DIR=`pwd`/out -DTRT_PLATFORM_ID=aarch64 -DCUDA_VERSION=10.2
make -j$(nproc)
sudo make install
I also created python bindings by following the instructions here:
https://github.com/NVIDIA/TensorRT/tree/8.0.1/python
Hope this helps @Akiautiste @liutaiyuan1 @Qengineering
Please find the TensorRT python wheel at the main page.