def forward(ctx, x):
x = x.clamp(-1 + 1e-15, 1 - 1e-15)
ctx.save_for_backward(x)
dtype = x.dtype
x = x.double()
res = (torch.log_(1 + x).sub_(torch.log_(1 - x))).mul_(0.5)
return res.to(dtype)
def forward(ctx, x):
x = x.clamp(-1 + 1e-15, 1 - 1e-15)
ctx.save_for_backward(x)
z = x.double()
return (torch.log_(
(1 + z).clamp_min(1e-15)).sub_(torch.log_(
(1 - z).clamp_min(1e-15)))).mul_(0.5).to(x.dtype)
def forward(ctx, x):
x_dtype = x.dtype
x = x.double()
x = x.clamp(-1 + 1e-5, 1 - 1e-5)
ctx.save_for_backward(x)
z = x.double()
temp = torch.log_(1 + z).sub_(torch.log_(1 - z))
res = (temp).mul_(0.5).to(x_dtype)
return res
def forward(self, x):
out = self.conv1(x)
out = F.leaky_relu(out)
out = torch.log_(1 + torch.abs_(out))
out = self.conv2(out)
out = F.leaky_relu(self.bn2(out))
out = F.avg_pool1d(out, kernel_size=7, padding=0, stride=3)
out1 = self.layer1(out)
out2 = self.layer2(out1)
out3 = self.layer3(out2)
out4 = self.layer4(out3)
out5 = self.layer5(out4)
# skip1 = F.avg_pool1d(out1, 15, stride=15)
# skip2 = F.avg_pool1d(out2, 7, stride=7)
skip3 = F.avg_pool1d(out3, 4, stride=4, padding=1)
skip4 = F.avg_pool1d(out4, 3, stride=2, padding=1)
out = torch.cat([out5, skip4, skip3], 1)
out = out[:, :,
1:-1] # обрезаем концы, чтобы padding не вызывал глюков
out = F.avg_pool1d(out, 26)
out = out.view(out.size(0), -1)
out = self.linear(F.dropout(out, training=self.training, p=0.25))
out = self.logsoftmax(out)
return out
def _make_ordinal(logits):
""" Ordinal Distribution Network """
sigmoid = torch.nn.Sigmoid()
logits = torch.squeeze(logits, 0)
s_i = sigmoid(logits)
one_minus_s = tensor(1) - s_i
_part_1 = torch.log_(s_i)
_part_2 = torch.log_(one_minus_s)
ordinal_logits = [
torch.sum(_part_1[:i + 1]).item() +
torch.sum(_part_2[i + 1:]).item() for i in range(len(s_i))
]
# ordinal_logits = [torch.sum(torch.log_(s_i[:i+1])).item() + torch.sum(torch.log_(one_minus_s[i+1:])).item() for i in range(len(s_i))]
ordinal_logits = tensor(ordinal_logits)
original_logits = self.softmax(ordinal_logits)
ordinal_logits = torch.unsqueeze(ordinal_logits, dim=0)
dist = torch.distributions.Categorical(logits=ordinal_logits)
return dist
def mish(input, inplace = False):
'''
Applies the mish function element-wise:
.. math::
mish(x) = x * tanh(softplus(x)) = x * tanh(ln(1 + e^{x}))
See additional documentation for :mod:`echoAIAI.Activation.Torch.mish`.
'''
if inplace:
inp = input.clone()
torch.exp_(input)
input += 1
torch.tanh_(torch.log_(input))
input *= inp
return input
else:
return input * torch.tanh(F.softplus(input))
optimizer_D.zero_grad()
if setting.distance == "wd_gp":
real_decision = D(real_data)
fake_decision = D(fake_data)
gradient_penalty = compute_gradient_penalty(
D, real_data.data, fake_data.data)
d_loss = -torch.mean(real_decision) + torch.mean(
fake_decision) + setting.lambda_gp * gradient_penalty
elif setting.distance == "jsd":
real_decision = D(real_data)
fake_decision = D(fake_data)
d_loss = -np.log(2) - 0.5 * torch.mean(
torch.log_(real_decision + 1 - 1)) - 0.5 * torch.mean(
torch.log_(1 - fake_decision + 1 - 1))
d_loss.backward()
optimizer_D.step()
print(
"[Epoch %d/%d] [Batch %d/%d] [D loss: %f] [Teta = %f]" %
(epoch, setting.n_epochs, i, sample_num, d_loss.item(), theta))
wsd.append(d_loss.item())
D_LOSSES.append(-d_loss)
plt.plot(np.arange(-1., 1.1, 0.1), D_LOSSES, 'ro')
plt.plot(np.arange(-1., 1.1, 0.1), D_LOSSES, 'c--')
plt.title("Distributions Distances")
plt.xlabel(r'$\Phi$')
if setting.distance == "wd_gp":
plt.ylabel("Wasserstein Distance (WD)")
def forward(self, x):
torch.log_(x)
return x
def forward(ctx: Any, x: torch.Tensor) -> torch.Tensor:
x = clamp(x, min=-1. + 4 * eps, max=1. - 4 * eps)
ctx.save_for_backward(x)
res = (torch.log_(1 + x).sub_(torch.log_(1 - x))).mul_(0.5)
return res
def forward(ctx, x):
x = x.clamp(-1 + 1e-5, 1 - 1e-5)
ctx.save_for_backward(x)
res = (torch.log_(1 + x).sub_(torch.log_(1 - x))).mul_(0.5)
return res
print(a.matmul(b)) print(a.matmul(b).shape) # 指数运算 a = torch.tensor([1, 2]) print(torch.pow(a, 3)) print(a.pow(3)) print(a**3) print(a.pow_(3)) # exp a = torch.tensor([1, 2], dtype=torch.float32) print(a.type()) print(torch.exp(a)) print(torch.exp_(a)) print(a.exp()) print(a.exp_()) # 对数 a = torch.tensor([10, 2], dtype=torch.float32) print(torch.log(a)) print(torch.log_(a)) print(a.log()) print(a.log_()) # sqart a = torch.tensor([10, 2], dtype=torch.float32) print(torch.sqrt(a)) print(torch.sqrt_(a)) print(a.sqrt()) print(a.sqrt_())
def forward(self, rgb, spad):
img_features = self.feature_extractor(rgb)
spad_features = self.hints_extractor(rgb, spad)
spad_features = spad_features.expand(-1, -1, img_features.size(2), img_features.size(3))
img_features.add_(self.spad_weight * torch.log_(spad_features + 1e-5))
return img_features
def forward(ctx, z):
z = z.clamp(-1 + 1e-5, 1 - 1e-5)
ctx.save_for_backward(z)
return (torch.log_(1 + z).sub_(torch.log_(1 - z))).mul_(0.5)
