Jetson Orin NX SDK Version Conflict Issue

Robotics & Edge Computing Jetson Systems Jetson Orin NX
1

My Orin NX is running JetPack 6.2 with CUDA version 12.6 and Python 3.10. The problem is that I need to use an SDK that only works with Python 3.8. However, after searching for many PyTorch versions compatible with JetPack 6.2 (for example: https://github.com/ultralytics/assets/releases/download/v0.0.0/torch-2.5.0a0+872d972e41.nv24.08-cp310-cp310-linux_aarch64.whl), I still can’t find a version suitable for Python 3.8 (which must support CUDA). I would like to ask if there is a way to solve this problem without reflashing the system to downgrade to JetPack 5.

2

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

I understand your concern. You’re running JetPack 6.2 on your Jetson Orin NX with CUDA version 11.6 and Python 3.10, but you need to use an SDK that only works with Python 3.8. You’ve searched for PyTorch versions compatible with JetPack 6.2, but couldn’t find one that supports Python 3.8.

Unfortunately, downgrading Python from 3.10 to 3.8 on JetPack 6.2 is not a straightforward process. JetPack 6.2 is designed to work with Python 3.10, and downgrading Python would likely cause compatibility issues with other packages and libraries.

However, I can offer a few potential solutions:

  1. Use a virtual environment: You can create a virtual environment using tools like virtualenv or conda that uses Python 3.8. This would allow you to isolate the Python 3.8 environment from the system Python 3.10 environment. You can then install the required SDK and dependencies within the virtual environment.
  2. Use a Docker container: You can create a Docker container that uses a base image with Python 3.8 and install the required SDK and dependencies within the container. This would allow you to run the SDK in a isolated environment without affecting the system Python environment.
  3. Compile PyTorch from source: If you’re comfortable with compiling software from source, you can try compiling PyTorch from source with Python 3.8 support. This would require you to modify the PyTorch build configuration to use Python 3.8 instead of Python 3.10.
  4. Use a different SDK: If the SDK you’re trying to use is not compatible with Python 3.10, you may want to explore alternative SDKs that are compatible with Python 3.10.

In comparison to reflashing the system to downgrade to JetPack 5, the above solutions are more feasible and less invasive. Reflashing the system would require you to reinstall all your software and configurations, which can be time-consuming and may lead to data loss.

Before attempting any of these solutions, I recommend verifying that the SDK you’re trying to use is not compatible with Python 3.10. You can try contacting the SDK developers or checking their documentation to see if they have any plans to support Python 3.10 in the future.

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

5

Hi,

To build the pytorch with specific python version.

Please refer to below topics build from source section

Thanks

6

I’m sorry to bother you again. In that case, I would like to know: if I create a virtual environment for my JetPack 6.2 system and configure Python 3.8, and considering that the PyTorch versions compatible with Python 3.8 mentioned in your reply are only for JetPack 5, can I try installing that PyTorch version on my system? If not, could you please tell me if there are other solutions, or if the only option is to downgrade the JetPack version? Thank you very much for your reply.

7

Hi,

Sorry, we dont support it.

Please try below two options:

  • Downgrade to 5.x
  • Build pytorch from source.

Thanks