0.5 * (open_seen_acc + open_unseen_acc),
open_seen_acc,
open_unseen_acc,
objoracle_acc,
bias,
))
all_accuracies = [all_accuracies, area]
return all_accuracies
#----------------------------------------------------------------#
#----------------------------------------------------------------#
trainset = dset.CompositionDatasetActivations(root=args.data_dir,
phase='train',
split=args.splitname)
trainloader = torch.utils.data.DataLoader(trainset,
batch_size=args.batch_size,
shuffle=True,
num_workers=args.workers)
valset = dset.CompositionDatasetActivations(root=args.data_dir,
phase=args.test_set,
split=args.splitname,
subset=args.subset)
valloader = torch.utils.data.DataLoader(valset,
batch_size=args.test_batch_size,
shuffle=False,
num_workers=args.workers)
if args.model == 'modularpretrained':
open_seen_acc,
open_unseen_acc,
objoracle_acc,
bias,
))
all_accuracies = [all_accuracies, area]
return all_accuracies
#----------------------------------------------------------------#
#----------------------------------------------------------------#
testset = dset.CompositionDatasetActivations(root=args.data_dir, phase='test', split='compositional-split')
testloader = torch.utils.data.DataLoader(testset, batch_size=args.batch_size, shuffle=False, num_workers=2)
if args.model == 'visprodNN':
model = models.VisualProductNN(testset, args)
elif args.model == 'redwine':
model = models.RedWine(testset, args)
elif args.model =='labelembed+':
model = models.LabelEmbedPlus(testset, args)
elif args.model =='attributeop':
model = models.AttributeOperator(testset, args)
model.cuda()
evaluator = models.Evaluator(testset,model)
checkpoint = torch.load(args.load)
(open_seen_acc * open_unseen_acc)**0.5,
0.5 * (open_seen_acc + open_unseen_acc),
open_seen_acc,
open_unseen_acc,
objoracle_acc,
meanAP,
bias,
))
return all_accuracies
#----------------------------------------------------------------#
trainset = dset.CompositionDatasetActivations(
root=args.data_dir,
phase='train',
split=args.splitname,
num_negs=args.num_negs,
pair_dropout=args.pair_dropout,
)
trainloader = torch.utils.data.DataLoader(trainset,
batch_size=args.batch_size,
shuffle=True,
num_workers=args.workers)
valset = dset.CompositionDatasetActivations(
root=args.data_dir,
phase=args.test_set,
split=args.splitname,
subset=args.subset,
)
valloader = torch.utils.data.DataLoader(valset,
batch_size=args.test_batch_size,