0.17.6

fix pl_mean initialized value

0.17.5

add dataset augmentation

0.17.4

add a feedforward after the self-attention

0.17.3

add a feedforward after the self-attention

0.17.2

add ability to save individual frames of interpolation

0.17.1

remove transparency when generating interpolation

0.17.0

Cap number of feature maps in generator and discriminator, like original paper

0.16.2

make sure contrastive loss regularization and transparent images are …

0.16.1

make sure contrastive loss reg still works

0.16.0

Add feature to augment images going into discriminator, for low training data