DoppelGANger: How to tell whether training is ok

[ldv1]

Hi,

How to tell whether the training of the DoppelGANger is ok, or to compare different models when the hyperparameters are varied? The obvious method would be to synthetize samples (for each model) and inspect them visually, but this is time consuming, especially if we try a lot of hyperparameters. What else could be done?
Thanks for the help.