I have access to a Jetson Orin NX computer with Jetpack 6.1. I am not able to update the OS installation or anything, and I need to run a python script that makes use of keras 2.15.X + tensorflow 2.15.X. However, according to this index, which I found in this tutorial, for Jetpack 6.1, the only available version of tensorflow is 2.16.1, which, thanks to tensorflow and keras being the absolute versioning nightmare they are, is only compatible with keras > 3.X.
Thus, is there a way to install a prior version of tensorflow in this Jetson device? If not, what would you do about this? Is there a way to make keras 2.15.X compatible with tensorflow 2.16.1? Or maybe there is a way to load models from that version of keras into keras 3.10.0?
Hi AastaLLL, thank you very much for the quick reply.
I am trying to follow step by step this TensorFlow tutorial to build from source the version 2.15.1 into a venv. But - clearly because I’m completely new to this world - I am finding issue after issue, and never getting to the end. Searching for a while over the internet, and talking with ChatGPT, it seems like maybe the tutorial is for newer versions, and older versions of the code are missing some config files - again, I have never built any library using bazel.
Is there someone I can contact that can help me? Or is there a tutorial to build specifically that version?
The file is generated for the Jetson environment so it is much closer to your target.
You can also try to install the 2.16 TensorFlow to see if everything works well then downgrade to the 2.15.