def __init__(self, features_dim: int):
"""
Args:
features_dim: size of the output tensor
"""
super(SimpleConv, self).__init__()
self._net = nn.Sequential(
nn.Conv2d(1, 32, 3, 1),
nn.ReLU(),
nn.Conv2d(32, 64, 3, 1),
nn.ReLU(),
nn.MaxPool2d(2),
Flatten(),
nn.Linear(9216, 128),
nn.ReLU(),
nn.Linear(128, features_dim),
Normalize(),
)
def __init__(self, out_features: int, normalize: bool = True):
"""
Args:
out_features: size of the output tensor
"""
super().__init__()
layers = [
nn.Conv2d(1, 32, 3, 1),
nn.ReLU(),
nn.Conv2d(32, 64, 3, 1),
nn.ReLU(),
nn.MaxPool2d(2),
Flatten(),
nn.Linear(9216, 128),
nn.ReLU(),
nn.Linear(128, out_features),
]
if normalize:
layers.append(Normalize())
self._net = nn.Sequential(*layers)
def __init__(self, num_hidden1=128, num_hidden2=64):
"""
Args:
num_hidden1: size of the first hidden representation
num_hidden2: size of the second hidden representation
"""
super().__init__()
self.conv_net = nn.Sequential(
nn.Conv2d(1, 32, 3, 1),
nn.ReLU(),
nn.Conv2d(32, 64, 3, 1),
nn.ReLU(),
nn.MaxPool2d(2),
Flatten(),
)
self.linear_net = nn.Sequential(
nn.Linear(9216, num_hidden1),
nn.ReLU(),
nn.Linear(num_hidden1, num_hidden2),
Normalize(),
)
self._net = nn.Sequential(self.conv_net, self.linear_net)