def __init__(self):
super(CIFAR_Net, self).__init__()
self.conv1 = nn.Conv2d(3, 64, 3, 1)
self.conv1b = nn.Conv2d(64, 64, 3, 1)
self.conv2 = nn.Conv2d(64, 128, 3, 1)
self.conv2b = nn.Conv2d(128, 128, 3, 1)
self.fc1 = nn.Linear(128 * 5 * 5, 512)
self.fc2 = nn.Linear(512, 128)
self.fc3 = nn.Linear(128, 10)
def __init__(self, vocab_size: int):
super().__init__()
# Embedding dimension: vocab_size + <unk>, <pad>, <eos>, <sos>
self.emb = crb.Embedding(vocab_size + 4, 100)
self.lstm = crb.LSTM(100, 100)
self.pool = crb.AvgPool1d(256)
self.fc1 = crb.Linear(100, 2)
def __init__(self, vocab_size, embedding_dim, pad_idx):
super().__init__()
# we will fix the dimensions here for the purpose of our experiments
hidden_dim = 256
output_dim = 1
n_layers = 1
self.embedding = nn.Embedding(vocab_size, embedding_dim, padding_idx = pad_idx)
self.rnn = nn.LSTM(embedding_dim,
hidden_dim,
num_layers=n_layers)
self.fc = nn.Linear(hidden_dim, output_dim)
def __init__(self,
input_size=(1, 32, 32),
n_layers=2,
n_channels=25,
factor=1.0,
kernel_size=5):
super().__init__()
self.input_size = input_size
# Layers grow at factor `factor`
conv_layers = []
for i in range(n_layers):
channels_in = int(n_channels * (factor ** (i - 1))) if i > 0 else \
input_size[0]
channels_out = int(n_channels * (factor**i))
conv_layers.append(
nn.Conv2d(channels_in, channels_out, kernel_size, 1,
padding=2))
self.conv_layers = nn.ModuleList(conv_layers)
self.conv_output = self.conv_block(torch.randn(1, *input_size)).size(1)
self.fc = nn.Linear(self.conv_output, 10)
def __init__(self):
super(Net, self).__init__()
self.conv1 = nn.Conv2d(1, 32, 3, 1)
self.conv2 = nn.Conv2d(32, 64, 3, 1)
self.fc1 = nn.Linear(9216, 128)
self.fc2 = nn.Linear(128, 10)
def __init__(self, **_):
super().__init__()
self.fc1 = crb.Linear(104, 1)
def __init__(self, **_):
super().__init__()
self.fc1 = crb.Linear(104, 50)
self.fc2 = crb.Linear(50, 2)
def __init__(self, **_):
super().__init__()
self.conv1 = crb.Conv2d(1, 16, 8, 2, padding=3)
self.conv2 = crb.Conv2d(16, 32, 4, 2)
self.fc1 = crb.Linear(32 * 4 * 4, 32)
self.fc2 = crb.Linear(32, 10)
