Usage of disentangle loss

[zijiebite]

Hi, I'd like to know how can I add disentangle loss into the training process so as to know the true value of disentangling. It seems adding disentangle loss into training requires big adjusts to the code released here. Thanks