def __init__(self, num_classes, bottom_width=4, nz=128, ngf=256, **kwargs):
super().__init__(nz=nz,
ngf=ngf,
bottom_width=bottom_width,
num_classes=num_classes,
**kwargs)
# Build the layers
self.l1 = nn.Linear(self.nz, (self.bottom_width**2) * self.ngf)
self.block2 = GBlock(self.ngf,
self.ngf,
upsample=True,
num_classes=self.num_classes)
self.block3 = GBlock(self.ngf,
self.ngf,
upsample=True,
num_classes=self.num_classes)
self.block4 = GBlock(self.ngf,
self.ngf,
upsample=True,
num_classes=self.num_classes)
self.b5 = nn.BatchNorm2d(self.ngf)
self.c5 = nn.Conv2d(self.ngf, 3, 3, 1, padding=1)
self.activation = nn.ReLU(True)
# Initialise the weights
nn.init.xavier_uniform_(self.l1.weight.data, 1.0)
nn.init.xavier_uniform_(self.c5.weight.data, 1.0)
def __init__(self,
num_classes,
nz=128,
ngf=1024,
bottom_width=4,
**kwargs):
super().__init__(nz=nz,
ngf=ngf,
bottom_width=bottom_width,
num_classes=num_classes,
**kwargs)
# Build the layers
self.l1 = SNLinear(self.nz, (self.bottom_width**2) * self.ngf)
self.block2 = GBlock(self.ngf,
self.ngf,
upsample=True,
num_classes=self.num_classes,
spectral_norm=True)
self.block3 = GBlock(self.ngf,
self.ngf >> 1,
upsample=True,
num_classes=self.num_classes,
spectral_norm=True)
self.block4 = GBlock(self.ngf >> 1,
self.ngf >> 2,
upsample=True,
num_classes=self.num_classes,
spectral_norm=True)
self.block5 = GBlock(self.ngf >> 2,
self.ngf >> 3,
upsample=True,
num_classes=self.num_classes,
spectral_norm=True)
self.block6 = GBlock(self.ngf >> 3,
self.ngf >> 4,
upsample=True,
num_classes=self.num_classes,
spectral_norm=True)
self.b7 = nn.BatchNorm2d(self.ngf >> 4)
self.c7 = SNConv2d(self.ngf >> 4, 3, 3, 1, padding=1)
self.activation = nn.ReLU(True)
# SA block
self.attn_block = SelfAttention(self.ngf >> 2, spectral_norm=True)
# Initialise the weights
nn.init.xavier_uniform_(self.l1.weight.data, 1.0)
nn.init.xavier_uniform_(self.c7.weight.data, 1.0)
def __init__(self, nz=128, ngf=1024, bottom_width=4, **kwargs):
super().__init__(nz=nz, ngf=ngf, bottom_width=bottom_width, **kwargs)
# Build the layers
self.l1 = nn.Linear(self.nz, (self.bottom_width**2) * self.ngf)
self.block2 = GBlock(self.ngf, self.ngf >> 1, upsample=True)
self.block3 = GBlock(self.ngf >> 1, self.ngf >> 2, upsample=True)
self.block4 = GBlock(self.ngf >> 2, self.ngf >> 3, upsample=True)
self.block5 = GBlock(self.ngf >> 3, self.ngf >> 4, upsample=True)
self.b6 = nn.BatchNorm2d(self.ngf >> 4)
self.c6 = nn.Conv2d(self.ngf >> 4, 3, 3, 1, padding=1)
self.activation = nn.ReLU(True)
# Initialise the weights
nn.init.xavier_uniform_(self.l1.weight.data, 1.0)
nn.init.xavier_uniform_(self.c6.weight.data, 1.0)
def __init__(self,
nz=128,
ngf=256,
bottom_width=4,
use_nfl=True,
double_layers=False,
**kwargs):
super().__init__(nz=nz, ngf=ngf, bottom_width=bottom_width, **kwargs)
self.use_nfl = use_nfl
if use_nfl:
print('using nfl!')
#from networks.cosgrove.attentive_densenet import AttentiveDensenet
self.ad = AttentiveDensenet(layer_channels=[
self.ngf, self.ngf, self.ngf, self.ngf, self.ngf, self.ngf,
self.ngf, self.ngf
],
key_size=32,
val_size=32,
n_heads=4)
# Build the layers
self.l1 = nn.Linear(self.nz, (self.bottom_width**2) * self.ngf)
self.block2 = GBlock(self.ngf, self.ngf, upsample=True)
self.block3 = GBlock(self.ngf, self.ngf, upsample=True)
self.block4 = GBlock(self.ngf, self.ngf, upsample=True)
#self.b5 = nn.BatchNorm2d(self.ngf)
#self.c5 = nn.Conv2d(self.ngf, 3, 3, 1, padding=1)
self.l1_b = nn.Linear(self.nz, (self.bottom_width**2) * self.ngf)
self.block2_b = GBlock(self.ngf, self.ngf, upsample=True)
self.block3_b = GBlock(self.ngf, self.ngf, upsample=True)
self.block4_b = GBlock(self.ngf, self.ngf, upsample=True)
self.b5_b = nn.BatchNorm2d(self.ngf)
self.c5_b = nn.Conv2d(self.ngf, 3, 3, 1, padding=1)
self.double_layers = double_layers
if double_layers:
self.block2_bd = GBlock(self.ngf, self.ngf, upsample=False)
self.block3_bd = GBlock(self.ngf, self.ngf, upsample=False)
self.block4_bd = GBlock(self.ngf, self.ngf, upsample=False)
self.activation = nn.ReLU(True)
# Initialise the weights
nn.init.xavier_uniform_(self.l1.weight.data, 1.0)
nn.init.xavier_uniform_(self.l1_b.weight.data, 1.0)
nn.init.xavier_uniform_(self.c5_b.weight.data, 1.0)