Exemple #1
0
                                            'recon_poeA', 'recon_poeB',
                                            'recon_crA', 'recon_crB',
                                            'total_loss', 'test_total_loss',
                                            'test_acc', 'val_total_loss',
                                            'val_acc', 'test_f1', 'val_f1')
    VIZ = visdom.Visdom(port=args.viz_port)
    viz_init()

preprocess_data = transforms.Compose([
    transforms.CenterCrop((168, 178)),
    transforms.Resize((64, 64)),
    transforms.ToTensor(),
])

train_data = torch.utils.data.DataLoader(datasets(
    partition='train',
    data_dir='../../data/celeba2',
    image_transform=preprocess_data),
                                         batch_size=args.batch_size,
                                         shuffle=True)

test_data = torch.utils.data.DataLoader(datasets(
    partition='test',
    data_dir='../../data/celeba2',
    image_transform=preprocess_data),
                                        batch_size=args.batch_size,
                                        shuffle=False)
val_data = torch.utils.data.DataLoader(datasets(
    partition='val',
    data_dir='../../data/celeba2',
    image_transform=preprocess_data),
                                       batch_size=args.batch_size,
Exemple #2
0

if args.viz_on:
    WIN_ID = dict(
        llA='win_llA', llB='win_llB',
        total_losses='win_total_losses',
        llB_test='win_llB_test', llA_test='win_llA_test'
    )
    LINE_GATHER = probtorch.util.DataGather(
        'epoch', 'recon_A', 'recon_B',
        'total_loss', 'test_total_loss', 'recon_A_test', 'recon_B_test'
    )
    VIZ = visdom.Visdom(port=args.viz_port)
    viz_init()

train_data = torch.utils.data.DataLoader(datasets(path, ATTR_IDX, train=True, crop=1.2), batch_size=args.batch_size,
                                         shuffle=True,
                                         num_workers=len(GPU))
test_data = torch.utils.data.DataLoader(datasets(path, ATTR_IDX, train=False, crop=1.2), batch_size=args.batch_size,
                                        shuffle=True,
                                        num_workers=len(GPU))

BIAS_TRAIN = (train_data.dataset.__len__() - 1) / (args.batch_size - 1)
BIAS_TEST = (test_data.dataset.__len__() - 1) / (args.batch_size - 1)


def cuda_tensors(obj):
    for attr in dir(obj):
        value = getattr(obj, attr)
        if isinstance(value, torch.Tensor):
            setattr(obj, attr, value.cuda())
Exemple #3
0
                  llB_test='win_llB_test',
                  llA_test='win_llA_test',
                  llB_val='win_llB_val',
                  llA_val='win_llA_val',
                  acc='win_acc')
    LINE_GATHER = probtorch.util.DataGather('epoch', 'recon_A', 'recon_B',
                                            'total_loss', 'test_total_loss',
                                            'recon_A_test', 'recon_B_test',
                                            'val_total_loss', 'recon_A_val',
                                            'recon_B_val', 'tr_acc', 'te_acc',
                                            'val_acc')
    VIZ = visdom.Visdom(port=args.viz_port)
    viz_init()

train_data = torch.utils.data.DataLoader(datasets(path,
                                                  ATTR_IDX,
                                                  train=True,
                                                  crop=1.2),
                                         batch_size=args.batch_size,
                                         shuffle=True,
                                         num_workers=len(GPU))
test_data = torch.utils.data.DataLoader(datasets(path,
                                                 ATTR_IDX,
                                                 train=False,
                                                 crop=1.2),
                                        batch_size=args.batch_size,
                                        shuffle=True,
                                        num_workers=len(GPU))

val_data = torch.utils.data.DataLoader(datasets(path,
                                                ATTR_IDX,
                                                train=True,
Exemple #4
0
                                            'test_acc')
    VIZ = visdom.Visdom(port=args.viz_port)
    viz_init()

path = '../../data/awa/Animals_with_Attributes2/'
test_classes = np.genfromtxt(path + 'testclasses.txt',
                             delimiter='\n',
                             dtype=str)
class_meta = np.genfromtxt(path + 'classes.txt', delimiter='\n', dtype=str)
test_labels = []
for test_class in test_classes:
    for i in range(len(class_meta)):
        if test_class in class_meta[i]:
            test_labels.append(i)

train_data = torch.utils.data.DataLoader(datasets(train=True),
                                         batch_size=args.batch_size,
                                         shuffle=True)
test_data = torch.utils.data.DataLoader(datasets(train=False),
                                        batch_size=args.batch_size,
                                        shuffle=True)

BIAS_TRAIN = (train_data.dataset.__len__() - 1) / (args.batch_size - 1)
BIAS_TEST = (test_data.dataset.__len__() - 1) / (args.batch_size - 1)


def cuda_tensors(obj):
    for attr in dir(obj):
        value = getattr(obj, attr)
        if isinstance(value, torch.Tensor):
            setattr(obj, attr, value.cuda())
Exemple #5
0
             win=WIN_ID['total_losses'],
             update='append',
             opts=dict(xlabel='epoch',
                       ylabel='loss',
                       title='Total Loss',
                       legend=['train_loss', 'test_loss']))


if args.viz_on:
    WIN_ID = dict(total_losses='win_total_losses')
    LINE_GATHER = probtorch.util.DataGather('epoch', 'total_loss',
                                            'test_total_loss')
    VIZ = visdom.Visdom(port=args.viz_port)
    viz_init()

train_data = torch.utils.data.DataLoader(datasets(path, train=True, crop=1.2),
                                         batch_size=args.batch_size,
                                         shuffle=True,
                                         num_workers=len(GPU))
test_data = torch.utils.data.DataLoader(datasets(path, train=False, crop=1.2),
                                        batch_size=args.batch_size,
                                        shuffle=True,
                                        num_workers=len(GPU))


def cuda_tensors(obj):
    for attr in dir(obj):
        value = getattr(obj, attr)
        if isinstance(value, torch.Tensor):
            setattr(obj, attr, value.cuda())
Exemple #6
0
        acc='win_acc', total_losses='win_total_losses', f1='win_f1'
    )
    LINE_GATHER = probtorch.util.DataGather(
        'epoch',
        'total_loss', 'test_acc', 'test_f1'
    )
    VIZ = visdom.Visdom(port=args.viz_port)
    viz_init()

preprocess_data = transforms.Compose([
    transforms.CenterCrop((168, 178)),
    transforms.Resize((64, 64)),
    transforms.ToTensor(),
])

train_data = torch.utils.data.DataLoader(datasets(partition='train', data_dir='../../data/celeba2',
                                                  image_transform=preprocess_data), batch_size=args.batch_size,
                                         shuffle=True)

test_data = torch.utils.data.DataLoader(datasets(partition='test', data_dir='../../data/celeba2',
                                                 image_transform=preprocess_data), batch_size=args.batch_size,
                                        shuffle=False)

def cuda_tensors(obj):
    for attr in dir(obj):
        value = getattr(obj, attr)
        if isinstance(value, torch.Tensor):
            setattr(obj, attr, value.cuda())


encA = EncoderA(args.wseed, n_attr=N_ATTR)
if CUDA: