def __init__(self, args):
self.model = ACGAN()
self.G_optimizer = torch.optim.Adam(self.model.G.parameters(),
lr=1e-4,
betas=(0.5, 0.999))
self.D_optimizer = torch.optim.Adam(self.model.D.parameters(),
lr=1e-4,
betas=(0.5, 0.999))
self.criterion = torch.nn.CrossEntropyLoss()
self.dataloader = celeba_loader(args.batch_size)
self.epochs = args.epochs
self.modelpath = 'ckpt/%s/model.pth' % args.taskname
self.img_dpath = 'ckpt/' + args.taskname + '/img_%d-%s.png'
self.args = args
if args.resume is not None:
print('resuming model', args.resume)
self.model.load(args.resume)
self.device = 'cuda' if torch.cuda.is_available() else 'cpu'
self.model.to(self.device)
self.LAMBDA = 10
self.p = 1
self.q = 1
self.one = torch.FloatTensor([1]).to(self.device)
self.mone = self.one * -1
print('finish init')
def __init__(self, args):
self.model = DCGAN()
self.G_optimizer = torch.optim.Adam(self.model.G.parameters())
self.D_optimizer = torch.optim.Adam(self.model.D.parameters())
self.criterion = torch.nn.BCELoss()
self.dataloader = celeba_loader(args.batch_size)
self.device = 'cuda' if torch.cuda.is_available() else 'cpu'
self.model.to(self.device)
self.k = 1
self.epochs = args.epochs
self.modelpath = 'ckpt/%s/model.pth' % args.taskname
self.img_dpath = 'ckpt/' + args.taskname + '/img_%d.png'
def __init__(self, args):
self.model = ACGAN(args)
self.G_optimizer = torch.optim.Adam(self.model.G.parameters())
self.D_optimizer = torch.optim.Adam(self.model.D.parameters())
self.criterion = torch.nn.BCELoss()
self.label_criterion = torch.nn.CrossEntropyLoss()
self.dataloader = celeba_loader(args.batch_size)
self.epochs = args.epochs
self.modelpath = 'ckpt/%s/model.pth' % args.taskname
self.img_dpath = 'ckpt/' + args.taskname + '/img_%d-%s.png'
self.args = args
if args.resume is not None:
print('loading model')
self.model.load(self.modelpath)
self.device = 'cuda' if torch.cuda.is_available() else 'cpu'
self.model.to(self.device)
self.p = 1
self.q = 1