I am a sysadmin that started as MLOPs, so please bear with me... This could be wrong... But I managed to install it using

• Python 3.11 (brew vent)
• a requirements.txt file to fix the following packages versions:
• Tensorflow fixed to version 2.17 (2.17.1)
• tensorflow-metal to 1.1.0 (latest, sep 2023...)
• tf_keras 2.17.0

I do not use Conda. My company does not use it. But the packages version should be pretty similar a part of some dependencies I read people speaking about Conda (Conda install apple) (https://medium.com/@jarondlk/installing-tensorflow-metal-on-apple-silicon-macos-with-miniconda-f43121fe3054 as ONE example)