1.8.2

1.8.2

1.8.1

fix vector quantization

1.8.0

use layernorm for attention

1.7.0

release dual contrastive loss in 1.7.0

1.6.0

rewrite linear attention with some small tweaks

1.5.14

calculate FID faster by making sure real images are not recomputed ea…

1.5.12

fix an issue with new version of pytorch fid

1.5.11

1.5.11

1.5.10

auto-set augmentation probability if number of training samples is low

1.5.9

fix generating images with ModelLoader