def compute(content: Tensor, target_hm: Tensor):
content_xy, _ = heatmap_to_measure(content)
target_xy, _ = heatmap_to_measure(target_hm)
lossyash = Loss(nn.BCELoss()(content, target_hm) * weight +
nn.MSELoss()(content_xy, target_xy) * weight * 0.001)
if name:
writer.add_scalar(name, lossyash.item(), counter.get_iter(name))
return lossyash
def hm_svoego_roda_loss(pred, target):
pred_coord = heatmap_to_measure(pred)[0]
target_coord = heatmap_to_measure(target)[0]
pred = pred.relu() + 1e-15
target[target < 1e-7] = 0
target[target > 1 - 1e-7] = 1
if torch.isnan(pred).any() or torch.isnan(target).any():
return Loss.ZERO()
bce = nn.BCELoss()(pred, target)
if torch.isnan(bce).any():
return Loss.ZERO()
return Loss(bce + nn.MSELoss()(pred_coord, target_coord) * 0.0005)
def hm_svoego_roda_loss(pred, target):
pred_xy, _ = heatmap_to_measure(pred)
t_xy, _ = heatmap_to_measure(target)
return Loss(nn.BCELoss()(pred, target) +
nn.MSELoss()(pred_xy, t_xy) * 0.0005 +
(pred - target).abs().mean() * 0.3)
def hm_svoego_roda_loss(pred, target):
pred_xy, _ = heatmap_to_measure(pred)
with torch.no_grad():
t_xy, _ = heatmap_to_measure(target)
return Loss(
nn.BCELoss()(pred, target) +
nn.MSELoss()(pred_xy, t_xy) * 0.001
# (pred - target).abs().mean() * 0.1
)
def stariy_hm_loss(pred, target, coef=1.0):
pred_mes = UniformMeasure2DFactory.from_heatmap(pred)
target_mes = UniformMeasure2DFactory.from_heatmap(target)
return Loss(
nn.BCELoss()(pred, target) * coef +
nn.MSELoss()(pred_mes.coord, target_mes.coord) * (0.001 * coef) +
nn.L1Loss()(pred_mes.coord, target_mes.coord) * (0.001 * coef)
# (pred - target).abs().mean() * 0.1 * coef
)
def hm_svoego_roda_loss(pred, target, coef=1.0, l1_coef=0.0):
pred_mes = UniformMeasure2DFactory.from_heatmap(pred)
target_mes = UniformMeasure2DFactory.from_heatmap(target)
# pred = pred.relu() + 1e-15
# target[target < 1e-7] = 0
# target[target > 1 - 1e-7] = 1
if torch.isnan(pred).any() or torch.isnan(target).any():
print("nan in hm")
return Loss.ZERO()
bce = nn.BCELoss()(pred, target)
if torch.isnan(bce).any():
print("nan in bce")
return Loss.ZERO()
return Loss(bce * coef + nn.MSELoss()(pred_mes.coord, target_mes.coord) *
(0.0005 * coef) +
nn.L1Loss()(pred_mes.coord, target_mes.coord) * l1_coef)
def _generator_loss(self, dgz: Tensor, real: List[Tensor],
fake: List[Tensor]) -> Loss:
batch_size = dgz.size(0)
nc = dgz.size(1)
real_labels = torch.full((
batch_size,
nc,
), 1, device=dgz.device)
errG = self.__criterion(
dgz.view(batch_size, nc).sigmoid(), real_labels)
return Loss(errG)
def _discriminator_loss(self, dx: Tensor, dy: Tensor) -> Loss:
batch_size = dx.size(0)
nc = dx.size(1)
real_labels = torch.full((
batch_size,
nc,
), 1, device=dx.device)
err_real = self.__criterion(
dx.view(batch_size, nc).sigmoid(), real_labels)
fake_labels = torch.full((
batch_size,
nc,
), 0, device=dx.device)
err_fake = self.__criterion(
dy.view(batch_size, nc).sigmoid(), fake_labels)
return Loss(-(err_fake + err_real))
def discriminator_loss(self, x: Tensor, y: Tensor):
L1 = nn.L1Loss()
x = x.detach()
y = y.detach()
tyx: Tensor = self.Tyx(y).detach() # detach ?
loss: Tensor = self.Dx(x).mean() + self.product(
tyx, y) - self.Dx(tyx).mean()
x0 = self.gradient_point([tyx], [x])[0]
tx0y = self.d_grad(x0)
x0_pred = self.Tyx(tx0y)
y0 = self.gradient_point([y], [self.d_grad(x, False)])[0]
ty0x = self.Tyx(y0)
y0_pred = self.d_grad(ty0x)
pen = L1(y0_pred, y0) + L1(x0_pred, x0)
return Loss(loss + self.pen_weight * pen)
def _compute(self, delta: Tensor) -> Loss:
gradient_penalty_value = delta.relu().norm(2, dim=1).pow(2).mean()
return Loss(self.weight * gradient_penalty_value)
def add_generator_loss(self, loss: nn.Module, weight=1.0):
return self.__add__(
GANLossObject(
lambda dx, dy: Loss.ZERO(), lambda dgz, real, fake: Loss(
loss(fake[0], real[0].detach()) * weight),
self.discriminator))
def _compute(self, gradients: List[Tensor]) -> Loss:
gradients_cat = torch.cat([g.view(g.size(0), -1) for g in gradients],
dim=1)
# gradients: Tensor = gradients.view((gradients.size(0), -1))
gradient_penalty_value = ((gradients_cat.norm(2, dim=1) - 1)**2).mean()
return Loss(self.weight * gradient_penalty_value)
def otdelnaya_function(content: Tensor, measure: ProbabilityMeasure):
content_cropped = content
lossyash = Loss((content_cropped - measure.coord).abs().mean())
return lossyash
def generator_loss(self, x: Tensor):
L1 = nn.L1Loss()
tx: Tensor = self.d_grad(x, False).detach()
return Loss(L1(x, tx))
def _discriminator_loss(self, d_real: Tensor, d_fake: Tensor) -> Loss:
discriminator_loss = (1 - d_real).relu().mean() + (1 + d_fake).relu().mean()
return Loss(-discriminator_loss)
def compute(t1: Tensor, t2: Tensor):
loss = l1_loss(t1, t2)
if name:
if counter.get_iter(name) % 10 == 0:
writer.add_scalar(name, loss, counter.get_iter(name))
return Loss(loss)
lr=5e-4,
betas=(0.5, 0.97))
g_transforms: albumentations.DualTransform = albumentations.Compose([
ToNumpy(),
NumpyBatch(
albumentations.ElasticTransform(p=0.8,
alpha=150,
alpha_affine=1,
sigma=10)),
NumpyBatch(albumentations.ShiftScaleRotate(p=0.5, rotate_limit=10)),
ToTensor(device)
])
R_s = UnoTransformRegularizer.__call__(
g_transforms, lambda trans_dict, img, ltnt: Loss(nn.L1Loss()(
ltnt, style_encoder(trans_dict['image']))))
W300DatasetLoader.batch_size = 24
W300DatasetLoader.test_batch_size = 64
Celeba.batch_size = 24
heatmaper = ToHeatMap(64)
# tuner = GoldTuner([1.0, 1.0], device=device, rule_eps=0.02, radius=0.5, active=True)
w300_test = next(iter(LazyLoader.w300().test_loader))
w300_test_image = w300_test['data'].to(device)[:8]
def hm_svoego_roda_loss(pred, target, coef=1.0, l1_coef=0.0):
pred_mes = UniformMeasure2DFactory.from_heatmap(pred)
def hm_loss_bes_xy(pred, target):
return Loss(
nn.BCELoss()(pred, target)
# (pred - target).abs().mean() * 0.1
)
def transport_loss(self, y: Tensor):
y = y.detach()
ty: Tensor = self.Tyx(y)
return Loss(self.product(ty, y) - self.Dx(ty).mean())
def entropy(hm: Tensor):
B, N, D, D = hm.shape
return Loss(-(hm * hm.log()).sum() / (B * D * D))
def _generator_loss(self, dgz: Tensor, real: List[Tensor], fake: List[Tensor]) -> Loss:
return Loss(-dgz.mean())
