gt4sd.frameworks.torch.vae module¶

pytorch utils for VAEs.

Summary¶

Functions:

reparameterize

Applies reparametrization trick to obtain sample from latent space.

Reference¶

reparameterize(mu, logvar)[source]¶

Applies reparametrization trick to obtain sample from latent space.

Parameters

mu (Tensor) – the latent means of shape batch_size x latent_size.
logvar (Tensor) – latent log variances, shape batch_size x latent_size.

Returns

sampled Z from the latent distribution.

Return type

torch.Tensor