augment_func=Duplicate)
if args.adv:
model, teacher, disc, kwargs = experiment_setup(args)
experiment = solver.AdversarialDropoutSolver(model, loader_plain,
**kwargs)
experiment.optimize()
if args.vada:
model, teacher, disc, kwargs = experiment_setup(args)
experiment = solver.VADASolver(model, disc, loader_plain, **kwargs)
experiment.optimize()
if args.dann:
model, teacher, disc, kwargs = experiment_setup(args)
experiment = solver.DANNSolver(model, disc, loader_plain, **kwargs)
experiment.optimize()
if args.assoc:
model, teacher, disc, kwargs = experiment_setup(args)
experiment = solver.AssociativeSolver(model, loader_plain, **kwargs)
experiment.optimize()
if args.coral:
print(args.null)
model, teacher, disc, kwargs = experiment_setup(args)
experiment = solver.DeepCoralSolver(model,
loader_plain,
use_nullspace=args.null,
**kwargs)
experiment.optimize()
if osp.exists(args.checkpoint):
print("Resume from checkpoint file at {}".format(args.checkpoint))
model = torch.load(args.checkpoint)
else:
model = models.SVHNmodel()
# Dataset
data = datasets.load_dataset(path="data", train=True, img_size=32)
train_loader = torch.utils.data.DataLoader(data[args.source],
batch_size=args.sourcebatch,
shuffle=True,
num_workers=4)
val_loader = torch.utils.data.DataLoader(data[args.target],
batch_size=args.targetbatch,
shuffle=True,
num_workers=4)
dataset = datasets.JointLoader(train_loader, val_loader)
# Initialize the solver for this experiment
experiment = solver.DANNSolver(model,
dataset,
n_epochs=args.epochs,
savedir=args.log,
dryrun=args.dryrun,
learningrate=args.learningrate,
gpu=args.gpu if not args.cpu else None)
experiment.optimize()