def __init__(self, in_channels: int, num_classes: int) -> None:
super().__init__()
self.head = nn.Sequential(
nn.AdaptiveAvgPool2d(1),
tnn.Reshape(in_channels),
)
self.emb = nn.Embedding(num_classes, in_channels)
self.discr = nn.Linear(in_channels, 1)
def __init__(self, feat_extractor, feature_size, num_classes):
super(ProjectionDiscr, self).__init__()
self.bone = feat_extractor
self.head = nn.Sequential(
nn.AdaptiveMaxPool2d(1),
tnn.Reshape(feature_size),
)
self.emb = nn.Embedding(num_classes, feature_size)
self.discr = nn.Linear(feature_size, 1)
def __init__(self, feat_extractor, feature_size, num_classes):
super(Classifier1, self).__init__()
self.bone = feat_extractor
self.head = nn.Sequential(
nn.AdaptiveMaxPool2d(1),
tnn.Reshape(feature_size),
#nn.Dropout(0.5),
kaiming(nn.Linear(feature_size, num_classes)),
nn.BatchNorm1d(num_classes))
def set_pool_size(self, size: int) -> 'ClassificationHead':
"""
Average pool to spatial size :code:`size` rather than 1. Recreate the
first Linear to accomodate the change.
"""
self.pool = nn.AdaptiveAvgPool2d(size)
self.reshape = tnn.Reshape(size * size * self.in_channels)
self.linear1 = kaiming(
nn.Linear(size * size * self.in_channels,
self.linear1.out_features))
return self
def __init__(self, feat_extractor, feature_size, num_classes):
super(Classifier, self).__init__()
self.bone = feat_extractor
self.head = nn.Sequential(
nn.AdaptiveMaxPool2d(1),
tnn.Reshape(feature_size),
kaiming(nn.Linear(feature_size, feature_size)),
nn.ReLU(inplace=True),
xavier(nn.Linear(feature_size, num_classes)),
)
def __init__(self, in_ch, num_classes, B=0):
def ch(ch):
return int(ch * 1.1**B) // 8 * 8
def n_layers(d):
return int(math.ceil(d * 1.2**B))
def r():
return int(224 * 1.15**B)
super(EfficientNet, self).__init__(
# Stage 1
# nn.UpsamplingBilinear2d(size=(r(), r())),
tu.kaiming(tnn.Conv3x3(in_ch, ch(32), stride=2, bias=False)),
nn.BatchNorm2d(ch(32)),
tnn.HardSwish(),
# Stage 2
MBConv(ch(32), ch(16), 3, mul_factor=1),
*[
MBConv(ch(16), ch(16), 3, mul_factor=1)
for _ in range(n_layers(1) - 1)
],
# Stage 3
MBConv(ch(16), ch(24), 3, stride=2),
*[MBConv(ch(24), ch(24), 3) for _ in range(n_layers(2) - 1)],
# Stage 4
MBConv(ch(24), ch(40), 5, stride=2),
*[MBConv(ch(40), ch(40), 5) for _ in range(n_layers(2) - 1)],
# Stage 5
MBConv(ch(40), ch(80), 3, stride=2),
*[MBConv(ch(80), ch(80), 3) for _ in range(n_layers(3) - 1)],
# Stage 6
MBConv(ch(80), ch(112), 5),
*[MBConv(ch(112), ch(112), 5) for _ in range(n_layers(3) - 1)],
# Stage 7
MBConv(ch(112), ch(192), 5, stride=2),
*[MBConv(ch(192), ch(192), 5) for _ in range(n_layers(4) - 1)],
# Stage 8
MBConv(ch(192), ch(320), 3),
*[MBConv(ch(320), ch(320), 3) for _ in range(n_layers(1) - 1)],
tu.kaiming(tnn.Conv1x1(ch(320), ch(1280), bias=False)),
nn.BatchNorm2d(ch(1280)),
tnn.HardSwish(),
nn.AdaptiveAvgPool2d(1),
tnn.Reshape(-1),
tu.xavier(nn.Linear(ch(1280), num_classes)))
def to_two_layers(self, hidden_channels: int) -> 'ClassificationHead':
"""
Set the classifier architecture to avgpool-flatten-linear1-relu-linear2.
"""
self._modules = OrderedDict([
('pool', nn.AdaptiveAvgPool2d(1)),
('reshape', tnn.Reshape(self.in_channels)),
('linear1', kaiming(nn.Linear(self.in_channels, hidden_channels))),
('relu1', nn.ReLU(True)),
('linear2', kaiming(nn.Linear(hidden_channels, self.num_classes))),
])
return self
def to_resnet_style(self) -> 'ClassificationHead':
"""
Set the classifier architecture to avgpool-flatten-linear.
"""
self._modules = OrderedDict([
('pool', nn.AdaptiveAvgPool2d(1)),
('reshape', tnn.Reshape(self.in_channels)),
('linear1', kaiming(nn.Linear(self.in_channels,
self.num_classes))),
])
return self
def VggGeneratorDebug(in_noise=32, out_ch=3):
"""
A not so small Vgg net image GAN generator for testing purposes
Args:
in_noise (int): dimension of the input noise
out_ch (int): number of output channels (3 for RGB images)
Returns:
a VGG instance
"""
return nn.Sequential(
kaiming(nn.Linear(in_noise, 128 * 16)), tnn.Reshape(128, 4, 4),
nn.LeakyReLU(0.2, inplace=True),
VggBNBone([128, 'U', 64, 'U', 32, 'U', 16], in_ch=128),
xavier(tnn.Conv1x1(16, 1)), nn.Sigmoid())
def __init__(self, in_noise, out_ch, side_ch=1):
super(VggImg2ImgGeneratorDebug, self).__init__()
def make_block(in_ch, out_ch, **kwargs):
return tnn.Conv2dNormReLU(
in_ch,
out_ch,
norm=lambda out: tnn.Spade2d(out, side_ch, 64),
**kwargs)
self.net = tnn.CondSeq(
kaiming(nn.Linear(in_noise, 128 * 16)), tnn.Reshape(128, 4, 4),
nn.LeakyReLU(0.2, inplace=True),
VggBNBone([128, 'U', 64, 'U', 32, 'U', 16],
in_ch=128,
block=make_block), xavier(tnn.Conv1x1(16, out_ch)),
nn.Sigmoid())
def __init__(self, in_noise, out_ch, num_classes):
super(VggClassCondGeneratorDebug, self).__init__()
def make_block(in_ch, out_ch, **kwargs):
return tnn.Conv2dNormReLU(
in_ch,
out_ch,
norm=lambda out: tnn.ConditionalBN2d(out, 64),
**kwargs)
self.emb = nn.Embedding(num_classes, 64)
self.net = tnn.CondSeq(
kaiming(nn.Linear(in_noise, 128 * 16)), tnn.Reshape(128, 4, 4),
nn.LeakyReLU(0.2, inplace=True),
VggBNBone([128, 'U', 64, 'U', 32, 'U', 16],
in_ch=128,
block=make_block), xavier(tnn.Conv1x1(16, out_ch)),
nn.Sigmoid())
def to_vgg_style(self, hidden_channels: int) -> 'ClassificationHead':
"""
Set the classifier architecture to
avgpool-flatten-linear1-relu-dropout-linear2-relu-dropout-linear3, like
initially done with VGG.
"""
self._modules = OrderedDict([
('pool', nn.AdaptiveAvgPool2d(7)),
('reshape', tnn.Reshape(7 * 7 * self.in_channels)),
('linear1',
kaiming(nn.Linear(7 * 7 * self.in_channels, hidden_channels))),
('relu1', nn.ReLU(True)),
('dropout1', nn.Dropout(0.5)),
('linear2', kaiming(nn.Linear(hidden_channels, hidden_channels))),
('relu2', nn.ReLU(True)),
('dropout2', nn.Dropout(0.5)),
('linear3', kaiming(nn.Linear(hidden_channels, self.num_classes))),
])
return self
