def forward(self, x: Tensor) -> Tensor:
x = self.conv(x)
x = x.view(-1, 512 * 2 * 2)
x = self.linear(x)
return x
def forward(self, x: Tensor) -> Tensor:
x = self.bn1(self.fc1(x)).relu()
x = self.bn2(self.fc2(x)).relu()
x = x.view(-1, 16, self.img_size_h, self.img_size_w)
x = self.up1(self.cbn1(self.conv1(x)).relu())
if self.dataset == 'CelebA':
x = self.up2(self.cbn2(self.conv2(x)).relu())
x = self.cbn3(self.conv3(x)).relu()
x = self.conv4(x).sigmoid()
return x
def forward(self, x: Tensor) -> Tuple[Tensor]:
x = self.down1(self.cbn1(self.conv1(x)).relu())
if self.dataset == 'CelebA':
x = self.down2(self.cbn2(self.conv2(x)).relu())
x = self.cbn3(self.conv3(x)).relu()
x = x.view(-1, 16*self.img_size_h*self.img_size_w)
x = self.bn1(self.fc1(x)).relu()
d = self.out_d(x).sigmoid()
q_cat = self.out_q_cat(x)
q_non_cat_mean, q_non_cat_logvar = self.out_q_non_cat_mean(x), self.out_q_non_cat_logvar(x)
return d, q_cat, q_non_cat_mean, q_non_cat_logvar
def forward(self, img: Tensor, labels: int):
"""Forward pass of the Discriminator.
Args:
img: the image that should be classified in fake or real.
labels: the label of the image.
Returns:
"""
d_in = torch.cat(
(img.view(img.size(0), -1), self.label_embedding(labels)), -1)
score = self.model(d_in)
return score
def forward(self, input: Tensor, mask: Tensor = None, hx: Tuple[Tensor, Tensor] = None) -> Tuple[Tensor, Tensor]:
batch_size = input.size(0) if self.batch_first else input.size(1)
if hx is None:
num_directions = 2 if self.bidirectional else 1
hx = input.new_zeros((self.num_layers * num_directions, batch_size, self.hidden_size))
hx = (hx, hx)
func = rnn_f.autograd_var_masked_rnn(num_layers=self.num_layers,
batch_first=self.batch_first,
bidirectional=self.bidirectional,
lstm=True)
self.reset_noise(batch_size)
output, hidden = func(input, self.all_cells, hx, None if mask is None else mask.view(mask.size() + (1,)))
return output, hidden
def forward(self, x: Tensor) -> Tensor:
y: Tensor = self.conv1(x)
x = self.conv128(y)
x += y
y = self.conv128to256(x)
x = self.conv256(y)
x += y
y = self.conv256to512(x)
x = self.conv512(y)
x += y
x = x.view(-1, 512 * 4 * 4)
return self.linear(x)
def forward(self, x: Tensor) -> Tensor:
x = self.conv(x)
x = x.view(-1, 256 * 3 * 3)
x = self.linear(x)
return x
def forward(self, x: Tensor) -> Tensor:
x = self.conv(x)
x = x.view(-1, 512 * 4 * 4) # -1 means infer this dimension
x = self.linear(x)
return x
