Releases: lucidrains/lightweight-gan
Releases · lucidrains/lightweight-gan
0.12.2
add truncated normals, default truncation at thresholds -1.5 and 1.5
0.12.1
add patch from @cafeal for faster multi-gpu training
0.12.0
0.12.0
0.11.3
fix spatial sle for 1024
0.11.2
autoset augmentation probability for user
0.11.1
amp will only work with torch 1.6 or greater
0.11.0
add mixed precision support, leading to 33% increase in speed and 40%…
0.10.0
add a small discriminator for heavily downsized image
0.9.2
fix bug
0.9.1
validate attention resolution layers inputs, update readme