| td/vq-resnet/4 |
cd0d878 |
0.8784 |
Single parallel VQ layer with gradient skipping |
| td/vq-resnet/5 |
cd0d878 |
0.8837 |
Single parallel VQ layer without skipping |
| td/vq-resnet/6 |
cd0d878 |
0.9504 |
Single parallel VQ layer with stop_gradient |
| td/vq-resnet/11 |
81bf80c |
1.1817 |
Single parallel VQ layer trained with min-coulomb loss |
| fp/adversarial-net/7 |
- |
1.7681 |
Unmodified, as-is ALP-trained ResNet from the models repo |
| td/vq-lesci/8 |
3cf175e |
1.3123 |
First submission with LESCI |
| td/adversarial-net/1 |
- |
1.9406 |
Just like "florianpfisterer/adversarial-net/7" but slightly adjusted preprocessing (i.e. scaling) |