def __init__(self,
num_features=64,
dim_z=128,
bottom_width=6,
activation=F.relu,
num_classes=0,
bottleneck=False):
super(ResNetGenerator48, self).__init__()
self.num_features = num_features
self.dim_z = dim_z
self.bottom_width = bottom_width
self.activation = activation
self.num_classes = num_classes
self.l1 = nn.Linear(dim_z, 8 * num_features * bottom_width**2)
self.block2 = GenBlock(num_features * 8,
num_features * 4,
activation=activation,
upsample=True,
num_classes=num_classes,
bottleneck=bottleneck)
self.block3 = GenBlock(num_features * 4,
num_features * 2,
activation=activation,
upsample=True,
num_classes=num_classes,
bottleneck=bottleneck)
self.block4 = GenBlock(num_features * 2,
num_features,
activation=activation,
upsample=True,
num_classes=num_classes,
bottleneck=bottleneck)
self.b5 = nn.BatchNorm2d(num_features)
self.conv5 = nn.Conv2d(num_features, 3, 1, 1)
def __init__(self,
num_features=256,
dim_z=128,
channel=3,
bottom_width=4,
activation=F.relu,
num_classes=0,
distribution='normal'):
super(ResNetGenerator32, self).__init__()
self.num_features = num_features
self.dim_z = dim_z
self.bottom_width = bottom_width
self.activation = activation
self.num_classes = num_classes
self.distribution = distribution
self.l1 = nn.Linear(dim_z, num_features * bottom_width**2)
self.block2 = GenBlock(num_features,
num_features,
activation=activation,
upsample=True,
num_classes=num_classes)
self.block3 = GenBlock(num_features,
num_features,
activation=activation,
upsample=True,
num_classes=num_classes)
self.block4 = GenBlock(num_features,
num_features,
activation=activation,
upsample=True,
num_classes=num_classes)
self.b5 = nn.BatchNorm2d(num_features)
self.conv5 = nn.Conv2d(num_features, channel, 1, 1)
def __init__(self,
num_features=64,
dim_z=128,
bottom_width=4,
activation=F.relu,
num_classes=0,
distribution='normal'):
super(ResNetGenerator64, self).__init__()
self.num_features = num_features
self.dim_z = dim_z
self.bottom_width = bottom_width
self.activation = activation
self.num_classes = num_classes
self.distribution = distribution
self.l1 = nn.Linear(dim_z, 16 * num_features * bottom_width**2)
self.block2 = GenBlock(num_features * 16,
num_features * 8,
activation=activation,
upsample=True,
num_classes=num_classes)
self.block3 = GenBlock(num_features * 8,
num_features * 4,
activation=activation,
upsample=True,
num_classes=num_classes)
self.block4 = GenBlock(num_features * 4,
num_features * 2,
activation=activation,
upsample=True,
num_classes=num_classes)
self.block5 = GenBlock(num_features * 2,
num_features,
activation=activation,
upsample=True,
num_classes=num_classes)
self.b6 = nn.BatchNorm2d(num_features)
self.conv6 = nn.Conv2d(num_features, 3, 1, 1)
self._initialize()