Tensorflow and PyTorch GPU on Jetson Orin Nano Dev Kit

Autonomous Machines Jetson & Embedded Systems Jetson Orin Nano
, ,
1

Trying to work on image semantic processing on the Orin Nano 8GB Dev Kit. Since it’s image processing, I wanted to ensure the GPU is enabled for both PyTorch and

I’ve tried several times to install Tensorflow and PyTorch with GPU enabled. I can get Tensorflow installed with GPU. I can get PyTorch installed with GPU. The issue is I can’t get both installed at the same time with GPU. The versions don’t seem compatible due to package dependencies.

Has anyone successfully installed both? If so, can you please post your pip list with the package versions you’re using? Any assistance is greatly appreciated.

3

Hi,

It’s recommended to use our prebuilt libraries.
We provide packages of both that are built on the same jetPack environment.

Thanks.

4

I did, but please take a look at the documentation. Ex. PyTorch says numpy==1.19.4, tensorflow needs 1.22. The aws package just breaks everything. Several packages are incompatible versions. I’ll try to deconstruct the requirements and see if I can de-conflict them.

5

Got it to work! :) Thank you! :)

The trick was to use both wheels from Index of /compute/redist/jp/v512 simultaneously and NOT use the instructions!!

Many thanks!

8

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