Which PyTorch/TorchVision build supports CUDA 12.6 on JetPack 6.2 for YOLOv11 training & CSI-camera inference on Jetson Orin Nano Super?

Robotics & Edge Computing Jetson Systems Jetson Nano
, ,
1

Hello, I’m using a Jetson Orin Nano Super Developer Kit with JetPack 6.2 (CUDA 12.6). My goal is to train a YOLO model (e.g. YOLOv11) and deploy it for real-time object detection using a CSI-attached camera.

So far, I have:

  1. Installed JetPack 6.2 and confirmed CUDA 12.6 is active.
  2. Cloned the YOLO repo and downloaded sample weights/datasets.

However, I can’t find a pre-built PyTorch (and corresponding TorchVision) wheel that’s compatible with CUDA 12.6 on this platform. Could someone please advise:

Which PyTorch/TorchVision version or wheel** should I install for CUDA 12.6 on JetPack 6.2?
Are there official L4T-compatible wheels for Orin Nano Super, or must I build from source?
What is the recommended installation method (apt packages, pip wheels, or cross-compile)?
Any example scripts or GitHub repos demonstrating YOLOv5/YOLOv8/YOLOv11 training and CSI-camera inference on Jetson?

Thank you for your help!

3

Hi,

You can find the compatible packages for JetPack 6.2 in the below link:

Thanks.

5

When I click on the site named https://pypi.jetson-ai-lab.dev/jp6/cu126, unfortunately it does not open. How can I access this site?

6

Please use this link:

7

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