Normalization of model inputs

[ando-shah]

The training code in training/SurfaceFootNet_multinode_training.py seems to have code for z-score norming of the inputs, but is unused (e.g. here.

There is however, a recurring comment on what the mean values should be:

    # [ 3.77901167e-02 -7.10877708e-02  1.24484683e+00  2.56569862e+02
    #   9.80964342e+02  2.82531180e+02  2.88608260e+02  1.18414726e+00
    #   7.51264114e+00  9.86283611e-01  1.14933095e-02  8.46494663e+02
    #   2.86330624e+02  1.02993263e+02]

If you could clarify:

1. Are these the values that were used for normalization? If so what were the means and standard deviations used, per variable, during training?
2. Why is only the gp being z-score normed, and not the other input values?
3. No norming is being applied in the example code notebooks, such as footnet/ExampleSurfaceFootNet.ipynb and corresponding column notebook.
   If you could clarify what normalization was used during training and therefore what should be used during inference, that would be great!