No module named 'onnx_tensorrt'

AI & Data Science Deep Learning (Training & Inference) TensorRT
1

import onnx
import onnx_tensorrt.backend as backend
import numpy as np

model = onnx.load(“./New/Data/model.onnx”)
engine = backend.prepare(model, device=‘CUDA:0’)
input_data = np.random.random(size=(1, 3, 112, 112)).astype(np.float32)
output_data = engine.run(input_data)[0]
print(output_data)
print(output_data.shape)

Error___________________

ModuleNotFoundError Traceback (most recent call last)
in
1 import onnx
----> 2 import onnx_tensorrt.backend as backend
3 import numpy as np
4
5 model = onnx.load(“/home/souvik/Documents/TNPL/New/Convert_Mxnet_to_Onnx/model.onnx”)

ModuleNotFoundError: No module named ‘onnx_tensorrt’

2

Hi,
Did you find out what the error is?

3

You need to install onnx_tensorrt:
[url]https://github.com/onnx/onnx-tensorrt[/url]

Terveisin, Markus

4

You need to install onnx_tensorrt:
[url]https://github.com/onnx/onnx-tensorrt[/url]

Terveisin, Markus

5

Hi,
I did that part but still, I’m getting that error. Do I need to set something in .bashrc?

6

Hi,
What should i give in cmake … -DTENSORRT_ROOT=

ritesh@mach-1:~$ sudo find / -name tensorrt 2> /dev/null
/usr/share/doc/tensorrt
/usr/lib/python3.6/dist-packages/tensorrt
/usr/src/tensorrt
/home/ritesh/.local/lib/python3.6/site-packages/tensorflow/python/compiler/tensorrt
/home/ritesh/.local/lib/python3.6/site-packages/tensorflow/contrib/tensorrt

ritesh@mach-1:~$ ls -ld /usr/src/tensorrt/
drwxr-xr-x 5 root root 4096 Aug 24 19:37 /usr/src/tensorrt/

ritesh@mach-1:~$ ls /usr/src/tensorrt/
bin data samples

These are the result I’m getting. Can you what you wrote in DTENSORRT_ROOT path?

7

Hei,
Below is stuff I gathered from various sources and got it working in Jetson Nano.
I’m using sd card image from

https://courses.nvidia.com/courses/course-v1:DLI+C-RX-02+V1/about

Not sure you need all this.

pip3 install --user -r requirements_onnx.txt

requirements_onnx.txt:

numpy
onnx==1.4.1
pycuda==2019.1
wget>=3.2
Pillow>=5.2.0
bistiming
eyewitness==1.0.8
scipy==1.2.1
celery==4.3.0
gevent
line-bot-sdk

mkdir ~/git
cd ~/git
git clone --recursive https://github.com/onnx/onnx-tensorrt.git
export CUDACXX="/usr/local/cuda-10.0/bin/nvcc"
cd onnx-tensorrt
git checkout v5.0   # to get the rigth api for tensorRT 5.06
mkdir build; cd build
cmake .. -DTENSORRT_ROOT=/usr/src/tensorrt -DGPU_ARCHS="53" -DCUDA_INCLUDE_DIRS=/usr/local/cuda-10.0/include
make -j8
sudo make install
in onnx-tensorrt/setup.py change to

EXTRA_COMPILE_ARGS =  [
    '-std=c++11',
    '-DUNIX',
    '-D__UNIX',
    '-fPIC',
    '-O3',
    '-ffast-math',
    '-flto',
    '-march=armv8-a+crypto',
    '-mcpu=cortex-a57+crypto',
    '-w',
    '-fmessage-length=0',
    '-fno-strict-aliasing',
    '-D_FORTIFY_SOURCE=2',
    '-fstack-protector',
    '--param=ssp-buffer-size=4',
    '-Wformat',
    '-Werror=format-security',
    '-DNDEBUG',
    '-fwrapv',
    '-Wall',
    '-DSWIG',
]

and

dlinano@jetson-nano:~/git/onnx-tensorrt$ pip3 install --user .

You may also need to install python modules:
https://github.com/onnx/onnx-tensorrt#python-modules

Terveisin, Markus