How to save the latent variables (a particular sample) from Big Sleep?

[conorheins]

Hi @lucidrains,

Is there a quick or simple way to save the latent variables of the generator network (or more specifically, a particular sample/realization of these latent variables) that are passed through BigGAN to create a particular image?

In other words, is there a way we can extract the variable cond_vector (a concatenation of the latent 128-dimensional z vector and a vector of the embedded class probabilities, e.g. self.embeddings(class_label)) from the BigGAN model and e.g. save it to disk for later use?

Thanks for offering the world this amazing tool!

Best wishes,
Conor