0.1.25

gradient accumulate matching aware loss, and then project management

0.1.24

final auxiliary loss wired up in gigagan trainer

0.1.23

add gradient penalty for vision aided discriminator

0.1.22

format

0.1.21

add preliminary support for vision aided discriminator as part of tra…

0.1.19

stick with convention of self attention before cross attention in ups…

0.1.18

patch

0.1.17

also use bilinear upsample for rgb path in unet upsampler, following …

0.1.16

default back to bilinear upsample for base generator as in stylegan

0.1.15

for conditional training, complete the auxiliary clip contrastive loss