Documentation clarification of forward_kld

[adam2392]

https://vincentstimper.github.io/normalizing-flows/references/#normflows.flows does not have an entry for y in the forward-kid function even though it is an input argument for forward_kld.

I was wondering what the description of y should be in this method? In log_prob(x, y), it specifies that y is the class. Is this the case? So given say a dataset of (X, y), we feed in batches of X and class labels y and want to minimize the KL divergence between the decoded \hat{X} and input X over each class label distribution?

[VincentStimper]

Hi @adam2392,

The documentation is indeed not clear there, thanks for pointing this out. In this case, y is the class in case the model is class-conditional. I clarified this in this commit.

Best regards,
Vincent