0.4.7

given RQ transformer design, number of codes is not an issue. figure …

0.4.6

prepare for text embeds to be cached during training, for text condit…

0.4.5

fix typo

0.4.4

patch trainer

0.4.3

fix eos handling, thanks to @qixuema https://github.com/lucidrains/me…

0.4.2

dry a bit more stuff unrelated to research

0.4.1

move adam optimizer factory to another repo

0.4.0

closely align the layer dimensions in the autoencoder to be as in the…

0.3.20

allow one to add only attention layers to autoencoder in encoder or d…

0.3.19

stop gap measure for supporting classifier free guidance condition sc…