datasets.MNIST(DATA_PATH, train=False, download=True,
transform=transforms.ToTensor()),
batch_size=args.batch_size, shuffle=True)
print('>>> data loaded')
print('train: ', len(train_data.dataset))
print('test: ', len(test_data.dataset))
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, zShared_dim=args.n_shared)
decA = DecoderA(args.wseed, zShared_dim=args.n_shared)
encB = EncoderB(args.wseed, zShared_dim=args.n_shared)
decB = DecoderB(args.wseed, zShared_dim=args.n_shared)
if CUDA:
encA.cuda()
decA.cuda()
encB.cuda()
decB.cuda()
cuda_tensors(encA)
cuda_tensors(decA)
cuda_tensors(encB)
cuda_tensors(decB)
optimizer = torch.optim.Adam(
list(encB.parameters()) + list(decB.parameters()) + list(encA.parameters()) + list(decA.parameters()),
print('test: ', len(test_data.dataset))
BIAS_TRAIN = (len(train_data.dataset) - 1) / (args.batch_size - 1)
BIAS_VAL = (len(val_data.dataset) - 1) / (args.batch_size - 1)
BIAS_TEST = (len(test_data.dataset) - 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())
encA = EncoderA(args.wseed,
zPrivate_dim=args.n_private,
zShared_dim=args.n_shared)
decA = DecoderA(args.wseed,
zPrivate_dim=args.n_private,
zShared_dim=args.n_shared)
encB = []
decB = []
for _ in range(N_ATTR):
encB.append(EncoderB(args.wseed, num_attr=1, zShared_dim=args.n_shared))
decB.append(DecoderB(args.wseed, num_attr=1, zShared_dim=args.n_shared))
if CUDA:
encA.cuda()
decA.cuda()
cuda_tensors(encA)
test_data = torch.utils.data.DataLoader(DIGIT('./data', train=False),
batch_size=args.batch_size,
shuffle=False)
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())
encA = EncoderA(args.wseed, zPrivate_dim=args.n_private, num_pixels=NUM_PIXELS)
decA = DecoderA(args.wseed, zPrivate_dim=args.n_private, num_pixels=NUM_PIXELS)
encB = EncoderB(args.wseed, zPrivate_dim=args.n_private, num_pixels=NUM_PIXELS)
decB = DecoderB(args.wseed, zPrivate_dim=args.n_private, num_pixels=NUM_PIXELS)
if CUDA:
encA.cuda()
decA.cuda()
encB.cuda()
decB.cuda()
cuda_tensors(encA)
cuda_tensors(decA)
cuda_tensors(encB)
cuda_tensors(decB)
optimizer = torch.optim.Adam(list(encB.parameters()) +
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())
encA = EncoderA(args.wseed,
zPrivate_dim=args.n_privateA,
zShared_dim=args.n_shared,
num_hidden=args.num_hidden)
decA = DecoderA(args.wseed,
zPrivate_dim=args.n_privateA,
zShared_dim=args.n_shared,
num_hidden=args.num_hidden)
encB = EncoderB(args.wseed,
zShared_dim=args.n_shared,
num_hidden=args.num_hidden)
decB = DecoderB(args.wseed,
zShared_dim=args.n_shared,
num_hidden=args.num_hidden)
if CUDA:
encA.cuda()
decA.cuda()
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())
encA = EncoderA(args.wseed, zPrivate_dim=args.n_privateA, zSharedAttr_dim=ATTR_DIM)
decA = DecoderA(args.wseed, zPrivate_dim=args.n_privateA, zSharedAttr_dim=ATTR_DIM)
encB = EncoderB(args.wseed, zSharedAttr_dim=ATTR_DIM)
decB = DecoderB(args.wseed, zSharedAttr_dim=ATTR_DIM)
if CUDA:
encA.cuda()
decA.cuda()
encB.cuda()
decB.cuda()
cuda_tensors(encA)
cuda_tensors(decA)
cuda_tensors(encB)
cuda_tensors(decB)
if len(args.gpu) > 2:
print('multi: ' + args.gpu)
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())
encA = EncoderA(args.wseed)
decA = DecoderA(args.wseed)
if CUDA:
encA.cuda()
decA.cuda()
cuda_tensors(encA)
cuda_tensors(decA)
if len(args.gpu) > 2:
print('multi: ' + args.gpu)
encA = nn.DataParallel(encA)
decA = nn.DataParallel(decA)
optimizer = torch.optim.Adam(list(encA.parameters()) + list(decA.parameters()),
lr=args.lr)
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())
encA = EncoderA(args.wseed,
zPrivate_dim=args.n_privateA,
zSharedAttr_dim=ATTR_DIM,
zSharedLabel_dim=N_CLASSES)
decA = DecoderA2(args.wseed,
zPrivate_dim=args.n_privateA,
zSharedAttr_dim=ATTR_DIM,
zSharedLabel_dim=N_CLASSES)
if CUDA:
encA.cuda()
decA.cuda()
cuda_tensors(encA)
cuda_tensors(decA)
if len(args.gpu) > 2:
print('multi: ' + args.gpu)
encA = nn.DataParallel(encA)
decA = nn.DataParallel(decA)
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())
encA = EncoderA(args.wseed,
zPrivate_dim=args.n_privateA,
zSharedAttr_dim=ATTR_DIM)
if CUDA:
encA.cuda()
cuda_tensors(encA)
if len(args.gpu) > 2:
print('multi: ' + args.gpu)
encA = nn.DataParallel(encA)
optimizer = torch.optim.Adam(list(encA.parameters()), lr=args.lr)
def train(data, encA, optimizer):
epoch_elbo = epoch_correct = 0.0
epoch_pred = np.zeros(sum(ATTR_DIM))
)
viz_port = args.viz_port # port number, eg, 8097
VIZ = visdom.Visdom(port=args.viz_port)
viz_init()
viz_ll_iter = args.viz_ll_iter
viz_la_iter = args.viz_la_iter
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(num_pixels=4096, num_hidden=512, zPrivate_dim=args.n_private, zShared_dim=args.n_shared)
decA = DecoderA(num_pixels=4096, num_hidden=512, zPrivate_dim=args.n_private, zShared_dim=args.n_shared)
encB = EncoderB(num_pixels=4096, num_hidden=512, zPrivate_dim=args.n_private, zShared_dim=args.n_shared)
decB = DecoderB(num_pixels=4096, num_hidden=512, zPrivate_dim=args.n_private, zShared_dim=args.n_shared)
if CUDA:
encA.cuda()
decA.cuda()
encB.cuda()
decB.cuda()
cuda_tensors(encA)
cuda_tensors(decA)
cuda_tensors(encB)
cuda_tensors(decB)
optimizer = torch.optim.Adam(list(encB.parameters())+list(decB.parameters())+list(encA.parameters())+list(decA.parameters()),
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:
encA.cuda()
cuda_tensors(encA)
optimizer = torch.optim.Adam(
list(encA.parameters()),
lr=args.lr)
def train(data, encA, optimizer):
encA.train()
N = 0
total_loss = 0
for b, (images, attributes) in enumerate(data):
N += 1
batch_size=args.batch_size,
shuffle=True)
print('>>> data loaded')
print('train: ', len(train_data.dataset))
print('test: ', len(test_data.dataset))
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, zShared_dim=args.n_shared)
if CUDA:
encA.cuda()
cuda_tensors(encA)
optimizer = torch.optim.Adam(list(encA.parameters()), lr=args.lr)
def train(data, encA, epoch, optimizer, gt_std):
encA.train()
for param in encA.parameters():
param.requires_grad = False
encA.fc32.weight.requires_grad = True
N = 0
encA = EncoderImgF(args.wseed,
zPrivate_dim=args.n_privateA,
zShared_dim=args.n_shared,
num_hidden=args.num_hidden)
decA = DecoderImgF(args.wseed,
zPrivate_dim=args.n_privateA,
zShared_dim=args.n_shared,
num_hidden=args.num_hidden)
encB = EncoderAttr(args.wseed,
zShared_dim=args.n_shared,
num_hidden=args.num_hidden)
decB = DecoderAttr(args.wseed,
zShared_dim=args.n_shared,
num_hidden=args.num_hidden)
ae_encA = EncoderA(0)
ae_decA = DecoderA2(0)
if CUDA:
encA.cuda()
decA.cuda()
encB.cuda()
decB.cuda()
ae_encA.cuda()
ae_decA.cuda()
cuda_tensors(encA)
cuda_tensors(decA)
cuda_tensors(encB)
cuda_tensors(decB)
cuda_tensors(ae_encA)
cuda_tensors(ae_decA)
