Jetson Xavier NX torch and torchvision compability

Hi guys,

I’m working on Xavier NX. My project needs torch version >2.0.0 , so I installed 2.0.0 version, but I couldn’t find which version of torchvision is compatible with torch 2.0.0. Here is my versions and the screen output.

The links I registered below.

Hi,

Based on our PyTorch container, the compatible TorchVision version of PyTorch 2.0 is 0.14.1.

Thanks.

Hi,

I downloaded the "torchvision-0.14.1-cp38-cp38-manylinux2014_aarch64.whl" file from "https://download.pytorch.org/whl/torchvision/"and installed it. However, when I try to import torchvision, I encounter the following warning:

/home/hamed/.local/lib/python3.8/site-packages/torchvision/io/image.py:13: UserWarning: Failed to load the image Python extension: 
  warn(f"Failed to load the image Python extension: {e}")

I attempted to resolve the problem by running the command sudo apt-get install libjpeg-dev libpng-dev, but it did not resolve the issue. Is there a comprehensive tutorial available that explains how to install torchvision? While the installation of PyTorch is straightforward, torchvision seems to be more challenging.

Regarding my JetPack version, it is nvidia-jetpack 5.1.1-b56 for the arm64 architecture. This version corresponds to the NVIDIA Jetpack Meta Package.

Hi @hamed.t.g, to build torchvision with CUDA support see the Installation section of this post:

There should be an image.so under your torchvision install path (currently /home/hamed/.local/lib/python3.8/site-packages/torchvision)

Thanks a lot. It worked for me.

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