Onnx for Jetson Orin with Jetpack 6.1

Robotics & Edge Computing Jetson Systems Jetson Orin Nano
1

Hi, i m new to Nvidia Jetson Orin. I m trying to use the foundation stereo on my jetson and followed the instructions at the github code for the jetson version.
After some installation issues, and trying different options i had hit a deadblock. I m not able to install onnx as I dont see a version released for TensorRT 10.3. Is there any other way to overcome this?

Jet pack version : output of the command :

head -n 1 /etc/nv_tegra_release

R36 (release), REVISION: 4.4, GCID: 41062509, BOARD: generic, EABI: aarch64, DATE: Mon Jun 16 16:07:13 UTC 2025

TensorRT 10.3

Best Regards,
Ravi Sama

2

*** Please note that this reply is generated by LLM automatically ***

I’d be happy to help you with your question, Ravi!

It seems like you’re trying to install ONNX on your Jetson Orin Nano, but you’re facing issues due to the lack of a released version for TensorRT 10.3. I’ll provide some context and possible solutions to help you overcome this hurdle.

TensorRT and ONNX

TensorRT is a software development kit (SDK) for high-performance deep learning inference. It’s optimized for NVIDIA GPUs, including the Jetson series. ONNX (Open Neural Network Exchange) is an open format for representing deep learning models, allowing them to be transferred between different frameworks and platforms.

JetPack and TensorRT versions

You mentioned that your JetPack version is R36, which includes TensorRT 10.3. However, it appears that there is no official ONNX release for TensorRT 10.3.

Possible solutions

  1. Check for updates: Ensure that you’re running the latest version of JetPack and TensorRT. You can check for updates using the sudo apt-get update and sudo apt-get upgrade commands.
  2. Use a different ONNX version: If you’re not tied to a specific ONNX version, you can try installing an earlier or later version that is compatible with TensorRT 10.3. You can check the ONNX GitHub repository for available versions.
  3. Build ONNX from source: If you’re comfortable with building software from source, you can try compiling ONNX for your Jetson Orin Nano. This will require you to install the necessary dependencies and build tools.
  4. Use a different deep learning framework: If you’re not tied to ONNX, you can explore other deep learning frameworks like TensorFlow, PyTorch, or Caffe, which may have better support for TensorRT 10.3.
  5. Contact NVIDIA support: If none of the above solutions work, you can reach out to NVIDIA support for further assistance. They may be able to provide additional guidance or a custom solution for your specific use case.

Comparison with other Jetson devices

The Jetson Orin Nano is a relatively new device, and it’s possible that the ONNX support is still evolving. If you’re experiencing issues with the Jetson Orin Nano, you may want to explore other Jetson devices, like the Jetson Xavier NX or Jetson AGX Orin, which may have better support for ONNX and TensorRT.

In summary, Ravi, I hope these suggestions help you overcome the issues you’re facing with installing ONNX on your Jetson Orin Nano. If you have any further questions or concerns, feel free to ask!

*** Please give the thumbs up if you get answers you like. Or provide feedback to help us improve the answer. ***

5

Hi,

ONNX can be installed via pip3 and it should be compatible with TensorRT 10.3.
Do you meet any error when using it?

Thanks.