pyTorch GPU-activated in jetson orin nano

Hello,
I’m a newbie on Nvidia. i have jetson orin nano and i am trying to make some image processing using yolov8 and accelerate it using tensorRT to ab able to use GPU, i have jetpack 6.2 and cuda 12.6 but i can’t find a pre-built pyTorch for GPU usage.

i found this

but it refers to the https://pypi.jetson-ai-lab.dev/jp6/cu126/ which keeps giving me
The site could be temporarily unavailable or too busy.

i am using ubuntu 22.04

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

Welcome to the world of NVIDIA Jetson and AI development!

You’re trying to accelerate image processing using YOLOv8 and TensorRT on your Jetson Orin Nano, which is a great approach to leverage the GPU’s power. However, you’re facing issues finding a prebuilt PyTorch for GPU usage.

Firstly, let’s clarify that PyTorch is a deep learning framework that can be used on various platforms, including NVIDIA Jetson devices. To use PyTorch on your Jetson Orin Nano, you’ll need to ensure that you have the correct version of PyTorch installed, which is compatible with your JetPack and CUDA versions.

The documentation you mentioned, TorchTensorRT in JetPack, is a good starting point. However, it seems that you’re encountering issues with the instructions.

Here are a few suggestions to help you resolve the issue:

  1. Verify your JetPack and CUDA versions: Ensure that you’re running JetPack 6.2 and CUDA 12.6 on your Jetson Orin Nano. You can check the versions by running the following commands:
jetpack --version
nvcc --version
  1. Check the PyTorch version: You can try installing PyTorch using the pip package manager. Run the following command to install PyTorch:
pip3 install torch torchvision

This will install the latest version of PyTorch, which might not be compatible with your JetPack and CUDA versions. To install a specific version of PyTorch, you can use the following command:

pip3 install torch==1.12.1 torchvision==0.13.1

Replace 1.12.1 and 0.13.1 with the versions that are compatible with your JetPack and CUDA versions.

  1. Use the torch_tensorrt package: Once you have PyTorch installed, you can try using the torch_tensorrt package to accelerate your YOLOv8 model. You can install torch_tensorrt using the following command:
pip3 install torch-tensorrt
  1. Consult the NVIDIA forums: If you’re still facing issues, you can try consulting the NVIDIA forums, where you can find a wealth of information and support from the community. You can also try searching for similar issues or posting a new question.

In comparison to other NVIDIA Jetson devices, the Jetson Orin Nano has a similar architecture to the Jetson AGX Orin, but with some differences in terms of performance and power consumption. The Jetson Orin Nano is a more compact and power-efficient device, making it suitable for edge AI applications.

In terms of PyTorch support, the Jetson Orin Nano has similar support to other NVIDIA Jetson devices, with some minor differences in terms of version compatibility.

I hope this helps you resolve the issue and get started with accelerating your YOLOv8 model using TensorRT on your Jetson Orin Nano!

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

Hi,

Our server has changed to jp6/cu126 index.
Please try the packages on the above link.

Thanks.