def build_model(self): if self.model == 'model_34': self.G = model34_DeepSupervion() elif self.model == 'model_50A': self.G = model50A_DeepSupervion() elif self.model == 'model_50A_slim': self.G = model50A_slim_DeepSupervion() elif self.model == 'model_101A': self.G = model101A_DeepSupervion() elif self.model == 'model_101B': self.G = model101B_DeepSupervion() elif self.model == 'model_152': self.G = model152_DeepSupervion() elif self.model == 'model_154': self.G = model154_DeepSupervion() self.g_optimizer = torch.optim.SGD(filter(lambda p: p.requires_grad, self.G.parameters()), self.g_lr, weight_decay=0.0002, momentum=0.9) self.print_network(self.G, 'G') if torch.cuda.is_available(): self.G = torch.nn.DataParallel(self.G) self.G.cuda()
def build_model(self): if self.model == 'model_34': self.G = model34_DeepSupervion() elif self.model == 'model_50A': self.G = model50A_DeepSupervion() elif self.model == 'model_101A': self.G = model101A_DeepSupervion() self.g_optimizer = torch.optim.SGD(filter(lambda p: p.requires_grad, self.G.parameters()), self.g_lr, weight_decay=0.0002, momentum=0.9) self.print_network(self.G, 'G') if torch.cuda.is_available(): self.G = torch.nn.DataParallel(self.G) self.G.cuda()