ModuleNotFoundError: No module named 'tensorrt'

Autonomous Machines Jetson & Embedded Systems Jetson AGX Xavier
,
1

Description

On my Jetson Xavier AGX, configured with Jetpack 4.4.1 I can not import tensorrt in Python3.6.
I have successfully installed the following packages, but tensorrt does not appear in the Python lib directories.

  • nvidia-container-csv-tensorrt_7.1.3.0-1+cuda10.2_arm64.deb
  • tensorrt_7.1.3.0-1+cuda10.2_arm64.deb

Environment

System: Jetson Xavier AGX
Jetpack Version: 4.4.1
TensorRT Version: 7.1.3.0 arm64
CUDA Version: 10.2
CUDNN Version: 8.0
Operating System + Version: Ubuntu 18.04
Python Version (if applicable): 3.6

Steps To Reproduce

From the host machine I have copied the aforementioned files to the Jetso Xavier via ssh.

scp ~/Downloads/nvidia/sdkm_downloads/nvidia-container-csv-tensorrt_7.1.3.0-1+cuda10.2_arm64.deb xavier@192.168.1.110:/home/xavier/
scp ~/Downloads/nvidia/sdkm_downloads/tensorrt_7.1.3.0-1+cuda10.2_arm64.deb xavier@192.168.1.110:/home/xavier/

On the Jetson Xavier I have installed them the common way:

sudo dpkg -i nvidia-container-csv-tensorrt_7.1.3.0-1+cuda10.2_arm64.deb 
sudo dpkg -i tensorrt_7.1.3.0-1+cuda10.2_arm64.deb
2

Hi @patrick.nagelm8t1h,
I suggest you to raise the concern in Jetson Xavier forum to get better assistance.

Thanks!

3

Done! Thanks!

6

Hi,

Please use JetPack to handle the installation for you.
If you want to do it manually, you will need all the following packages to make sure the functionality.

tensorrt_7.2.1.6-1+cuda10.2_arm64.deb
libnvinfer7_7.2.1-1+cuda10.2_arm64.deb
libnvinfer-dev_7.2.1-1+cuda10.2_arm64.deb
libnvinfer-bin_7.2.1-1+cuda10.2_arm64.deb
libnvinfer-doc_7.2.1-1+cuda10.2_all.deb
libnvinfer-samples_7.2.1-1+cuda10.2_all.deb
libnvinfer-plugin7_7.2.1-1+cuda10.2_arm64.deb
libnvinfer-plugin-dev_7.2.1-1+cuda10.2_arm64.deb
libnvparsers7_7.2.1-1+cuda10.2_arm64.deb
libnvparsers-dev_7.2.1-1+cuda10.2_arm64.deb
libnvonnxparsers7_7.2.1-1+cuda10.2_arm64.deb
libnvonnxparsers-dev_7.2.1-1+cuda10.2_arm64.deb

uff-converter-tf_7.2.1-1+cuda10.2_arm64.deb
graphsurgeon-tf_7.2.1-1+cuda10.2_arm64.deb
python-libnvinfer_7.2.1-1+cuda10.2_arm64.deb
python-libnvinfer-dev_7.2.1-1+cuda10.2_arm64.deb
python3-libnvinfer_7.2.1-1+cuda10.2_arm64.deb
python3-libnvinfer-dev_7.2.1-1+cuda10.2_arm64.deb

More, TensorRT requires CUDA and cuDNN pre-installed.
So if JetPack is an optional for you, it’s recommended to use JetPack for convience.

Thanks.