0.0.18

0.0.18

0.0.17

prepare for cascading diffusion in unet, save the full progressive up…

0.0.16

fix bug with @jihoonerd

0.0.15

use t5 relative positional bias in prior network causal transformer, …

0.0.14

allow for decoder conditioning with the text encodings from CLIP, if …

0.0.12

use the eval decorator

0.0.11

use cross attention for conditioning unet based on image embedding to…

0.0.10

start using swish glu everywhere, given success of PaLM

0.0.9

fix bug with classifier free guidance for prior network, even though …

0.0.8

offer continuously parameterized time embedding for diffusion prior n…