How can I install TensorRT on my TX2?

I have a Jetson TX2 with Ubuntu 16.04 64-bit. It was flashed with Jetpack 3.3. I am using Python 3.5.2.

Nvidia’s documentation says that TensorRT comes with Jetpack 3.3, but TensorRT is nowhere to be found on the TX2.

I already checked the TensorRT Installation Guide found here:
https://docs.nvidia.com/deeplearning/sdk/tensorrt-install-guide/index.html

From the TensorRT installation guide, I tried the two install methods for Linux:

  1. Debian method - this did not work, as the debian package is amd64, and my system on the TX2 is arm64
  2. Tar method - this did not work. I followed the instructions and entered ‘sudo pip3 install tensorrt-4.0.1.6-cp35-cp35m-linux_x86_64.whl’ and the reply I received was ‘tensorrt-4.0.1.6-cp35-cp35m-linux_x86_64.whl is not a supported wheel on this platform.’

How can I install TensorRT on my TX2?

Hi ofiesh_dimitri,

I would suggest moving to the latest version if possible.

Using NVIDIA SDK Manager, you can see TensorRT on Target will be installed as below.

vickyy, thanks for your response. I used the NVIDIA SDK Manager to install the latest Jetpack (4.2) as you suggested. My host computer has Ubuntu 16.04, but somehow it flashed 18.04 to the Jetson. My Jetson does have TensorRT now, but I need Ubuntu 16.04, not 18.04. Can I use the SDK Manager to flash 16.04 from my host to the Jetson?

If you must stay with ubuntu 16.04, you need to use JetPack 3.3 with older TensorRT (4.0).