Hi @feranick , thanks for your recommendation. At my company we have been using tensorflow for a long time and just recently upgraded to Python 3.12. The models run fine in the cloud, but I am struggling to locally debug as the model doesn't even start running. I have followed your suggestion and at least now the GPU is listed in the visible devices. However, I am using Keras 3 instead of tf_keras and not getting the desired outcome as you did. Any suggestions?