assert len(source_set) == 152397
assert len(target_set) == 28978
source_loader = torch.utils.data.DataLoader(source_set,
batch_size=args.batch_size,
shuffle=args.shuffle,
num_workers=args.num_workers)
target_loader = torch.utils.data.DataLoader(target_set,
batch_size=args.batch_size,
shuffle=args.shuffle,
num_workers=args.num_workers)
if args.model == 'resnet101':
netG = ResBase101().cuda()
elif args.model == 'resnet50':
netG = ResBase50().cuda()
else:
raise ValueError('Unexpected value of args.model')
netF = ResClassifier(class_num=args.class_num, extract=args.extract).cuda()
netF.apply(weights_init)
def get_L2norm_loss_self_driven(x):
l = (x.norm(p=2, dim=1).mean() - args.radius)**2
return args.weight_ring * l
def get_cls_loss(pred, gt):
cls_loss = F.nll_loss(F.log_softmax(pred), gt)
return cls_loss
source_set = VisDAImage(source_root, source_label, train_transform)
target_set = VisDAImage(target_root, target_label, train_transform)
assert len(source_set) == 152397
assert len(target_set) == 28978
source_loader = torch.utils.data.DataLoader(source_set,
batch_size=args.batch_size,
shuffle=args.shuffle,
num_workers=args.num_workers)
target_loader = torch.utils.data.DataLoader(target_set,
batch_size=args.batch_size,
shuffle=args.shuffle,
num_workers=args.num_workers)
netG = ResBase50()
netF = ResClassifier(class_num=args.class_num, extract=args.extract)
netF.apply(weights_init)
netG.cuda()
netF.cuda()
def get_cls_loss(pred, gt):
cls_loss = F.nll_loss(F.log_softmax(pred), gt)
return cls_loss
def get_L2norm_loss_self_driven(x):
radius = x.norm(p=2, dim=1).detach()
assert radius.requires_grad == False
radius = radius + 0.3