def init_loss(self, opt):
Base_Model.init_loss(self, opt)
# #####################
# define loss functions
# #####################
# GAN loss
if opt.ganloss == 'gan':
self.criterionGAN = GANLoss(use_lsgan=False).to(self.device)
elif opt.ganloss == 'lsgan':
self.criterionGAN = GANLoss(use_lsgan=True).to(self.device)
else:
raise ValueError()
# identity loss
self.criterionIdt = RestructionLoss(opt.restruction_loss).to(self.device)
# feature metric loss
self.criterionFea = torch.nn.L1Loss()
# map loss
self.criterionMap = RestructionLoss(opt.map_m_type).to(self.device)
# #####################
# initialize optimizers
# #####################
self.optimizer_G = torch.optim.Adam(self.netG.parameters(),
lr=opt.lr, betas=(opt.beta1, opt.beta2))
self.optimizer_D = torch.optim.Adam(self.netD.parameters(),
lr=opt.lr, betas=(opt.beta1, opt.beta2))
self.optimizers = []
self.optimizers.append(self.optimizer_G)
self.optimizers.append(self.optimizer_D)
def init_loss(self, opt):
Base_Model.init_loss(self, opt)
# #####################
# define loss functions
# #####################
# identity loss
self.criterionIdt = RestructionLoss(opt.idt_loss,
opt.idt_reduction).to(self.device)
# map loss
self.criterionMap = RestructionLoss(opt.map_projection_loss).to(
self.device)
# #####################
# initialize optimizers
# #####################
self.optimizers = []
if self.opt.weight_decay_if:
self.optimizer_G = torch.optim.Adam(self.netG.parameters(),
lr=opt.lr,
betas=(opt.beta1, opt.beta2),
weight_decay=1e-4)
else:
self.optimizer_G = torch.optim.Adam(self.netG.parameters(),
lr=opt.lr,
betas=(opt.beta1, opt.beta2))
self.optimizers.append(self.optimizer_G)
def init_loss(self, opt):
Base_Model.init_loss(self, opt)
# #####################
# define loss functions
# #####################
# GAN loss
self.criterionGAN = GANLoss(use_lsgan=True).to(self.device)
# identity loss
self.criterionIdt = RestructionLoss(opt.idt_loss, opt.idt_reduction).to(self.device)
# feature metric loss
self.criterionFea = torch.nn.L1Loss().to(self.device)
# map loss
self.criterionMap = RestructionLoss(opt.map_projection_loss).to(self.device)
# #####################
# initialize optimizers
# #####################
self.optimizer_G = torch.optim.Adam(self.netG.parameters(),
lr=opt.lr, betas=(opt.beta1, opt.beta2))
self.optimizer_D = torch.optim.Adam(self.netD.parameters(),
lr=opt.lr, betas=(opt.beta1, opt.beta2))
self.optimizers = []
self.optimizers.append(self.optimizer_G)
self.optimizers.append(self.optimizer_D)
def init_loss(self, opt):
Base_Model.init_loss(self, opt)
# feature metric loss
self.criterion = torch.nn.MSELoss()
self.optimizer_G = torch.optim.Adam(self.netG.parameters(),
lr=opt.lr,
betas=(opt.beta1, opt.beta2),
weight_decay=1e-4)
self.optimizers = []
self.optimizers.append(self.optimizer_G)
def init_loss(self, opt):
Base_Model.init_loss(self, opt)
# #####################
# define loss functions
# #####################
# GAN loss
if opt.ganloss == 'gan':
self.criterionGAN = GANLoss(use_lsgan=False).to(self.device)
elif opt.ganloss == 'lsgan':
self.criterionGAN = GANLoss(use_lsgan=True).to(self.device)
elif opt.ganloss == 'wgan':
self.criterionGAN = WGANLoss(grad_penalty=False).to(self.device)
elif opt.ganloss == 'wgan_gp':
self.criterionGAN = WGANLoss(grad_penalty=True).to(self.device)
else:
raise ValueError()
# identity loss
if opt.restruction_loss == 'mse':
print('Restruction loss: MSE')
self.criterionIdt = torch.nn.MSELoss()
elif opt.restruction_loss == 'l1':
print('Restruction loss: l1')
self.criterionIdt = torch.nn.L1Loss()
else:
raise ValueError()
# feature metric loss
self.criterionFea = torch.nn.L1Loss()
# map loss
self.criterionMap = Map_loss(direct_mean=opt.map_m_type,
predict_transition=self.opt.CT_MIN_MAX,
gt_transition=self.opt.XRAY1_MIN_MAX).to(
self.device)
# #####################
# initialize optimizers
# #####################
self.optimizer_G = torch.optim.Adam(self.netG.parameters(),
lr=opt.lr,
betas=(opt.beta1, opt.beta2))
self.optimizer_D = torch.optim.Adam(self.netD.parameters(),
lr=opt.lr,
betas=(opt.beta1, opt.beta2))
self.optimizers = []
self.optimizers.append(self.optimizer_G)
self.optimizers.append(self.optimizer_D)