92

"<torch.utils.data.dataloader.DataLoader object at 0x7f6885951df0>"

"cuda"

64

12

0.1

200

0.1

0.1

0.0001

0.0005

128

4

10

3

3

4

10

"Discriminator( (first_conv): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(3, 4, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) ) (blocks): ModuleList( (0): DiscriminatorBlock( (convs): ModuleList( (0): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(4, 4, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (2): LayerNorm((4, 64, 64), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (1): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(4, 4, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (2): LayerNorm((4, 64, 64), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (2): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(4, 4, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (2): LayerNorm((4, 64, 64), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) ) (downsample): Conv2d(4, 8, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1)) ) (1): DiscriminatorBlock( (convs): ModuleList( (0): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(8, 8, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (2): LayerNorm((8, 32, 32), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (1): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(8, 8, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (2): LayerNorm((8, 32, 32), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (2): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(8, 8, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (2): LayerNorm((8, 32, 32), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) ) (downsample): Conv2d(8, 16, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1)) ) (2): DiscriminatorBlock( (convs): ModuleList( (0): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(16, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (2): LayerNorm((16, 16, 16), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (1): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(16, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (2): LayerNorm((16, 16, 16), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (2): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(16, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (2): LayerNorm((16, 16, 16), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) ) (downsample): Conv2d(16, 32, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1)) ) (3): DiscriminatorBlock( (convs): ModuleList( (0): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (2): LayerNorm((32, 8, 8), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (1): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (2): LayerNorm((32, 8, 8), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (2): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (2): LayerNorm((32, 8, 8), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) ) (downsample): Conv2d(32, 64, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1)) ) (4): DiscriminatorBlock( (convs): ModuleList( (0): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (2): LayerNorm((64, 4, 4), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (1): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (2): LayerNorm((64, 4, 4), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (2): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (2): LayerNorm((64, 4, 4), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) ) (downsample): Conv2d(64, 128, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1)) ) (5): DiscriminatorBlock( (convs): ModuleList( (0): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (2): LayerNorm((128, 2, 2), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (1): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (2): LayerNorm((128, 2, 2), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (2): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (2): LayerNorm((128, 2, 2), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) ) (downsample): Conv2d(128, 256, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1)) ) ) (classify): Sequential( (0): Conv2d(256, 1, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (1): Flatten(start_dim=1, end_dim=-1) ) )"

"StyleGAN( (mapping): MappingNetwork( (norm): LayerNorm((12,), eps=1e-05, elementwise_affine=True) (layers): ModuleList( (0): Sequential( (0): Linear(in_features=12, out_features=12, bias=True) (1): LayerNorm((12,), eps=1e-05, elementwise_affine=True) (2): LeakyReLU(negative_slope=0.01) ) (1): Sequential( (0): Linear(in_features=12, out_features=12, bias=True) (1): LayerNorm((12,), eps=1e-05, elementwise_affine=True) (2): LeakyReLU(negative_slope=0.01) ) (2): Sequential( (0): Linear(in_features=12, out_features=12, bias=True) (1): LayerNorm((12,), eps=1e-05, elementwise_affine=True) (2): LeakyReLU(negative_slope=0.01) ) (3): Sequential( (0): Linear(in_features=12, out_features=12, bias=True) (1): LayerNorm((12,), eps=1e-05, elementwise_affine=True) (2): LeakyReLU(negative_slope=0.01) ) ) (out): Linear(in_features=12, out_features=12, bias=True) ) (synthesis): SynthesisNetwork( (blocks): ModuleList( (0): SynthesisBlock( (layers): ModuleList( (0): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (2): LayerNorm((128, 2, 2), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (1): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (2): LayerNorm((128, 2, 2), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (2): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (2): LayerNorm((128, 2, 2), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) ) (ada_in): AdaIN() (A1): Linear(in_features=12, out_features=256, bias=True) (A2): Linear(in_features=12, out_features=256, bias=True) (B1): Conv2d(10, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (B2): Conv2d(10, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) ) (1): SynthesisBlock( (upsample): ConvTranspose2d(128, 64, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1)) (conv): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (2): LayerNorm((64, 4, 4), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (layers): ModuleList( (0): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (2): LayerNorm((64, 4, 4), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (1): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (2): LayerNorm((64, 4, 4), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (2): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (2): LayerNorm((64, 4, 4), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) ) (ada_in): AdaIN() (A1): Linear(in_features=12, out_features=128, bias=True) (A2): Linear(in_features=12, out_features=128, bias=True) (B1): Conv2d(10, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (B2): Conv2d(10, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) ) (2): SynthesisBlock( (upsample): ConvTranspose2d(64, 32, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1)) (conv): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (2): LayerNorm((32, 8, 8), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (layers): ModuleList( (0): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (2): LayerNorm((32, 8, 8), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (1): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (2): LayerNorm((32, 8, 8), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (2): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (2): LayerNorm((32, 8, 8), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) ) (ada_in): AdaIN() (A1): Linear(in_features=12, out_features=64, bias=True) (A2): Linear(in_features=12, out_features=64, bias=True) (B1): Conv2d(10, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (B2): Conv2d(10, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) ) (3): SynthesisBlock( (upsample): ConvTranspose2d(32, 16, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1)) (conv): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(16, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (2): LayerNorm((16, 16, 16), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (layers): ModuleList( (0): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(16, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (2): LayerNorm((16, 16, 16), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (1): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(16, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (2): LayerNorm((16, 16, 16), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (2): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(16, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (2): LayerNorm((16, 16, 16), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) ) (ada_in): AdaIN() (A1): Linear(in_features=12, out_features=32, bias=True) (A2): Linear(in_features=12, out_features=32, bias=True) (B1): Conv2d(10, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (B2): Conv2d(10, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) ) (4): SynthesisBlock( (upsample): ConvTranspose2d(16, 8, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1)) (conv): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(8, 8, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (2): LayerNorm((8, 32, 32), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (layers): ModuleList( (0): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(8, 8, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (2): LayerNorm((8, 32, 32), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (1): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(8, 8, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (2): LayerNorm((8, 32, 32), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (2): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(8, 8, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (2): LayerNorm((8, 32, 32), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) ) (ada_in): AdaIN() (A1): Linear(in_features=12, out_features=16, bias=True) (A2): Linear(in_features=12, out_features=16, bias=True) (B1): Conv2d(10, 8, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (B2): Conv2d(10, 8, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) ) (5): SynthesisBlock( (upsample): ConvTranspose2d(8, 4, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1)) (conv): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(4, 4, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (2): LayerNorm((4, 64, 64), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (layers): ModuleList( (0): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(4, 4, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (2): LayerNorm((4, 64, 64), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (1): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(4, 4, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (2): LayerNorm((4, 64, 64), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) (2): Sequential( (0): Dropout(p=0.1, inplace=False) (1): Conv2d(4, 4, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (2): LayerNorm((4, 64, 64), eps=1e-05, elementwise_affine=True) (3): LeakyReLU(negative_slope=0.01) ) ) (ada_in): AdaIN() (A1): Linear(in_features=12, out_features=8, bias=True) (A2): Linear(in_features=12, out_features=8, bias=True) (B1): Conv2d(10, 4, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (B2): Conv2d(10, 4, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) ) ) (to_rgb): Conv2d(4, 3, kernel_size=(1, 1), stride=(1, 1)) ) )"

"Adam ( Parameter Group 0 amsgrad: False betas: (0.5, 0.99) eps: 1e-08 initial_lr: 0.0001 lr: 0.0001 weight_decay: 0 )"

"Adam ( Parameter Group 0 amsgrad: False betas: (0.5, 0.5) eps: 1e-08 initial_lr: 0.0005 lr: 0.0005 weight_decay: 0 )"

0.9

0.9

0

"<torch.optim.lr_scheduler.MultiStepLR object at 0x7f688faeb580>"

"<torch.optim.lr_scheduler.MultiStepLR object at 0x7f6885951b20>"

0.5

0.5

0.09847034513950348

0.21063511818647385

0.3227998673915863

0.9264719724655152

4.604772424697876

4.604772901535034

0.8757079005241394

0.00000002551406517171

0.00000118532167334706