Installing library that requires python>=3.9 on AGX Orin with Jetpack 5.1.2


I’m trying to install DSPy (GitHub - stanfordnlp/dspy: Stanford DSPy: The framework for programming—not prompting—foundation models) which requires python>=3.9.
I am on a Jetson AGX Orin with Jetpack 5.1.2. When I look at PyTorch for Jetson I see that PyTorch with CUDA enabled wheels are only supported with Python 3.10 in Jetpack 6.
Is there any way I can install DSPy with python>=3.9 (presumably python==3.10.0) along with the suitable version of PyTorch?

I think this would require upgrading Jetpack to 6, is that possible for the AGX Orin? I tried looking for steps on that and I did not find definitive results. Could I just run the dustynv/l4t-pytorch:r36.2.0 docker container? I’m a bit confused as to how docker containers work with L4T but to my understanding I would first need to upgrade L4T to 36 (if such a thing is possible) and then I could load the container which should be using Jetpack 6 and therefore Python 3.10 and have a CUDA-enbled PyTorch built. Please let me know if you have any advice.



Yes, please flash your device into r36.2 and use the dustynv/l4t-pytorch:r36.2.0 container.

1 Like

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