Running class notebooks on M3 MacBook Pro

[jeffsnell]

I am running the Jupyter Notebook code in Jupyter Lab on a 2023 M3 Max MacBook Pro.
I have 36 GB RAM, running Sonoma 14.4.1
I am using Python v3.10, TensorFlow v2.14, Apple Metal v1.1.0

When I got to the "circles" models (binary classification) the models would not train like Colab (I also tried that) or the video.
Changing the hidden layer activation function from RELU to ELU made the "circles" models build pretty much identical to the results in the video (training was a bit faster, but accuracy/loss plots are quite similar) - also leaky_relu worked almost as well.

For morę information on alternate activation functions see:
https://towardsdatascience.com/7-popular-activation-functions-you-should-know-in-deep-learning-and-how-to-use-them-with-keras-and-27b4d838dfe6