Unable to install Tensorrt bindings and libs (8.6.1.6) on Jetson AGX Orin

Robotics & Edge Computing Jetson & Embedded Systems Jetson AGX Orin
,
1

Hi,

Firstly I wanted to say how amazing this community is!

I am trying to install Tensorrt 8.6.1.6 on my Jetson AGX Orin.

System Specs -
Ubuntu 20.04
CUDA 12.4
Python 3.9.19
Torch 2.0.1

I installed Tensorrt 8.6.1 GA following the instructions provided in Log in | NVIDIA Developer. However, I am unable to install tensorrt-bindings==8.6.1 and tensorrt-libs==8.6.1. I tried using pip but the versions are not available.

These are the commands I tried
pip install tensorrt-bindings==8.6.1
python -m pip install --extra-index-url https://pypi.nvidia.com tensorrt_libs

I have been stuck at the issue for a few days and would appreciate any guidance on how to move forward.

Thank you!

3

Hi,

We don’t have a prebuild for Python 3.9 on Jetson.
But it’s easy to build the binging from the source.

Please check the GitHub below:

Thanks.

4

Hi AastaLLL,

Thanks so much for the quick reply!

So installing (or building, sorry for not using the right term) the bindings would solve the missing tensorrt-bindings and tensorrt-libs package issues?

5

Hi,

Sorry that I double-checked your environment and the pybinging might not fix your issue.

Do you use JetPack 5 since the OS is 20.04?
We don’t have TensorRT 8.6 for JetPack 5 environment but it’s available once you upgrade the Orin with JetPack 6.
Is this an option for you?

Thanks.

6

Hi AastaLLL,

I think I can upgrade the version to Jetpack6. Let me try that and update you with what I get.

Thanks once again for your help!

8

Is this still an issue to support? Any result can be shared?

9

Hi
Still working on this issue. Will update here once i get some results.

11

Hi AastaLLL,

I have upgraded the version to Jetpack6.

image
image702×224 29 KB

What would be the next steps to installing Tensorrt bindings and libs (8.6.1.6) on the Jetson Orin?

Thanks once again for your time and help!

12

Hi,

The default version in JetPack 6 should be already meet you requirement:

$ python3
Python 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import tensorrt as trt
>>> trt.__version__
'8.6.2'

Thanks.

13

Hi AastaLLL,

Thanks for the info. I found this post (TensorRT 8.6 not running properly on Orin NX with Jetpack 6) which linked the docker container http://nvcr.io/nvidia/l4t-tensorrt:r8.6.2-runtime.

i ran the command sudo docker run --rm -it --runtime=nvidia --gpus all nvcr.io/nvidia/l4t-tensorrt:r8.6.2-runtime /bin/bash

when trying to import tensorrt I get this error -

image
image837×217 14.6 KB

running pip list I can see that tensorrt 8.6.2 is present. However, I do not see tensorrt-libs and tensorrt-bindings. I appreciate your help with this.

Thank you.

14

Hi,

For container usage, you will need to manually install the nvidia-l4t-dla-compiler_36.3.1-20240516220919_arm64.deb package in the below link:

https://repo.download.nvidia.com/jetson

Thanks.

15

Thanks,

I will update with the results here. (Also, happy 8-year anniversary! :3)

16

Thanks.
Look forward to the update.

17

Is this still an issue to support? Any result can be shared?

18

Hi

so i found that this docker container - nvcr.io/nvidia/l4t-jetpack:r36.3.0
has tensorrt installed and working. so i am using this currently. regarding the shared file issue, i still have not gotten that to work

19

Hi,

You will need to install the nvidia-l4t-dla-compiler natively on the device.

Since the package is added to the /etc/nvidia-container-runtime/host-files-for-container.d/drivers.csv, it can be accessed inside the container.

Thanks.

21

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.