import math import torch from torch import autograd as autograd from torch import nn as nn from torch.nn import functional as F from basicsr.utils.registry import LOSS_REGISTRY @LOSS_REGISTRY.register() class GANLoss(nn.Module): """Define GAN loss. Args: gan_type (str): Support 'vanilla', 'lsgan', 'wgan', 'hinge'. real_label_val (float): The value for real label. Default: 1.0. fake_label_val (float): The value for fake label. Default: 0.0. loss_weight (float): Loss weight. Default: 1.0. Note that loss_weight is only for generators; and it is always 1.0 for discriminators. """ def __init__(self, gan_type, real_label_val=1.0, fake_label_val=0.0, loss_weight=1.0): super(GANLoss, self).__init__() self.gan_type = gan_type self.loss_weight = loss_weight self.real_label_val = real_label_val self.fake_label_val = fake_label_val if self.gan_type == 'vanilla': self.loss = nn.BCEWithLogitsLoss() elif self.gan_type == 'lsgan': self.loss = nn.MSELoss() elif self.gan_type == 'wgan': self.loss = self._wgan_loss elif self.gan_type == 'wgan_softplus': self.loss = self._wgan_softplus_loss elif self.gan_type == 'hinge': self.loss = nn.ReLU() else: raise NotImplementedError(f'GAN type {self.gan_type} is not implemented.') def _wgan_loss(self, input, target): """wgan loss. Args: input (Tensor): Input tensor. target (bool): Target label. Returns: Tensor: wgan loss. """ return -input.mean() if target else input.mean() def _wgan_softplus_loss(self, input, target): """wgan loss with soft plus. softplus is a smooth approximation to the ReLU function. In StyleGAN2, it is called: Logistic loss for discriminator; Non-saturating loss for generator. Args: input (Tensor): Input tensor. target (bool): Target label. Returns: Tensor: wgan loss. """ return F.softplus(-input).mean() if target else F.softplus(input).mean() def get_target_label(self, input, target_is_real): """Get target label. Args: input (Tensor): Input tensor. target_is_real (bool): Whether the target is real or fake. Returns: (bool | Tensor): Target tensor. Return bool for wgan, otherwise, return Tensor. """ if self.gan_type in ['wgan', 'wgan_softplus']: return target_is_real target_val = (self.real_label_val if target_is_real else self.fake_label_val) return input.new_ones(input.size()) * target_val def forward(self, input, target_is_real, is_disc=False): """ Args: input (Tensor): The input for the loss module, i.e., the network prediction. target_is_real (bool): Whether the targe is real or fake. is_disc (bool): Whether the loss for discriminators or not. Default: False. Returns: Tensor: GAN loss value. """ target_label = self.get_target_label(input, target_is_real) if self.gan_type == 'hinge': if is_disc: # for discriminators in hinge-gan input = -input if target_is_real else input loss = self.loss(1 + input).mean() else: # for generators in hinge-gan loss = -input.mean() else: # other gan types loss = self.loss(input, target_label) # loss_weight is always 1.0 for discriminators return loss if is_disc else loss * self.loss_weight @LOSS_REGISTRY.register() class MultiScaleGANLoss(GANLoss): """ MultiScaleGANLoss accepts a list of predictions """ def __init__(self, gan_type, real_label_val=1.0, fake_label_val=0.0, loss_weight=1.0): super(MultiScaleGANLoss, self).__init__(gan_type, real_label_val, fake_label_val, loss_weight) def forward(self, input, target_is_real, is_disc=False): """ The input is a list of tensors, or a list of (a list of tensors) """ if isinstance(input, list): loss = 0 for pred_i in input: if isinstance(pred_i, list): # Only compute GAN loss for the last layer # in case of multiscale feature matching pred_i = pred_i[-1] # Safe operation: 0-dim tensor calling self.mean() does nothing loss_tensor = super().forward(pred_i, target_is_real, is_disc).mean() loss += loss_tensor return loss / len(input) else: return super().forward(input, target_is_real, is_disc) def r1_penalty(real_pred, real_img): """R1 regularization for discriminator. The core idea is to penalize the gradient on real data alone: when the generator distribution produces the true data distribution and the discriminator is equal to 0 on the data manifold, the gradient penalty ensures that the discriminator cannot create a non-zero gradient orthogonal to the data manifold without suffering a loss in the GAN game. Reference: Eq. 9 in Which training methods for GANs do actually converge. """ grad_real = autograd.grad(outputs=real_pred.sum(), inputs=real_img, create_graph=True)[0] grad_penalty = grad_real.pow(2).view(grad_real.shape[0], -1).sum(1).mean() return grad_penalty def g_path_regularize(fake_img, latents, mean_path_length, decay=0.01): noise = torch.randn_like(fake_img) / math.sqrt(fake_img.shape[2] * fake_img.shape[3]) grad = autograd.grad(outputs=(fake_img * noise).sum(), inputs=latents, create_graph=True)[0] path_lengths = torch.sqrt(grad.pow(2).sum(2).mean(1)) path_mean = mean_path_length + decay * (path_lengths.mean() - mean_path_length) path_penalty = (path_lengths - path_mean).pow(2).mean() return path_penalty, path_lengths.detach().mean(), path_mean.detach() def gradient_penalty_loss(discriminator, real_data, fake_data, weight=None): """Calculate gradient penalty for wgan-gp. Args: discriminator (nn.Module): Network for the discriminator. real_data (Tensor): Real input data. fake_data (Tensor): Fake input data. weight (Tensor): Weight tensor. Default: None. Returns: Tensor: A tensor for gradient penalty. """ batch_size = real_data.size(0) alpha = real_data.new_tensor(torch.rand(batch_size, 1, 1, 1)) # interpolate between real_data and fake_data interpolates = alpha * real_data + (1. - alpha) * fake_data interpolates = autograd.Variable(interpolates, requires_grad=True) disc_interpolates = discriminator(interpolates) gradients = autograd.grad( outputs=disc_interpolates, inputs=interpolates, grad_outputs=torch.ones_like(disc_interpolates), create_graph=True, retain_graph=True, only_inputs=True)[0] if weight is not None: gradients = gradients * weight gradients_penalty = ((gradients.norm(2, dim=1) - 1)**2).mean() if weight is not None: gradients_penalty /= torch.mean(weight) return gradients_penalty