This program expects me to downgrade to a 2019 version of tensor flow and that would cause everything else to break.

[LinuxBeaver]

Hello, I run a couple of machine learning apps in a miniconda profile (REMBG and Texturize) and I was thinking about adding a third one MAX-IRE to upscale images using AI/Machine learning. But I got this error ModuleNotFoundError: No module named 'tensorflow.contrib and found out it was because MAX-IRE requires an old version of TensorFlow that had the "contrib" library. New versions of TensorFlow do not have said Contrib library.

Can you please get MAX-IRE working without relying on legacy tensorflow?

[kmh4321]

Hi @LinuxBeaver,

Everything the model requires is within it's own Docker container. So it should not cause any issues with your local setup.

[jleaders]

@LinuxBeaver you should be using the docker container which isolate's it naturally through LXC. But if you absolutely must not user docker, user poetry or some other venv manager like conda.