transforms.Scale(256),
transforms.RandomCrop(224),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
]))
word_embedding = None
train_loader = data.DataLoader(train_data,
batch_size=args.batch_size,
shuffle=True,
num_workers=args.num_threads)
model = VisualSemanticEmbedding(args.embed_ndim)
if not args.no_cuda:
model.cuda()
optimizer = torch.optim.Adam(filter(lambda x: x.requires_grad, model.parameters()),
lr=args.learning_rate)
for epoch in range(args.num_epochs):
avg_loss = 0
for i, (img, desc, len_desc) in enumerate(train_loader):
img = Variable(img.cuda() if not args.no_cuda else img)
desc = Variable(desc.cuda() if not args.no_cuda else desc)
len_desc, indices = torch.sort(len_desc, 0, True)
indices = indices.numpy()
img = img[indices, ...]
desc = desc[indices, ...].transpose(0, 1)
word_embedding, args.max_nwords,
transforms.Compose([
transforms.Resize((224, 224)),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
]))
word_embedding = None
train_loader = data.DataLoader(train_data,
batch_size=args.batch_size,
shuffle=True,
num_workers=args.num_threads)
model = VisualSemanticEmbedding(args.embed_ndim)
if not args.no_cuda:
model.cuda()
optimizer = torch.optim.Adam(
[x for x in model.parameters() if x.requires_grad],
lr=args.learning_rate)
for epoch in range(args.num_epochs):
avg_loss = 0
for i, (img, seg, desc, len_desc) in enumerate(train_loader):
img = Variable(img.cuda() if not args.no_cuda else img)
desc = Variable(desc.cuda() if not args.no_cuda else desc)
len_desc, indices = torch.sort(len_desc, 0, True)
required=True,
help='root directory of output')
parser.add_argument('--no_cuda', action='store_true', help='do not use cuda')
args = parser.parse_args()
if not args.no_cuda and not torch.cuda.is_available():
print('Warning: cuda is not available on this machine.')
args.no_cuda = True
if __name__ == '__main__':
print('Loading a pretrained fastText model...')
word_embedding = fasttext.load_model(args.fasttext_model)
print('Loading a pretrained model...')
txt_encoder = VisualSemanticEmbedding(args.embed_ndim)
txt_encoder.load_state_dict(torch.load(args.text_embedding_model))
txt_encoder = txt_encoder.txt_encoder
G = Generator()
G.load_state_dict(torch.load(args.generator_model))
G.eval()
if not args.no_cuda:
txt_encoder.cuda()
G.cuda()
transform = transforms.Compose([
transforms.Scale(74),
transforms.CenterCrop(64),
transforms.ToTensor()
transforms.Scale(256),
transforms.RandomCrop(224),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
]))
word_embedding = None
train_loader = data.DataLoader(train_data,
batch_size=args.batch_size,
shuffle=True,
num_workers=args.num_threads)
model = VisualSemanticEmbedding(args.embed_ndim)
if not args.no_cuda:
model.cuda()
optimizer = torch.optim.Adam(filter(lambda x: x.requires_grad,
model.parameters()),
lr=args.learning_rate)
for epoch in range(args.num_epochs):
avg_loss = 0
for i, (img, desc, len_desc) in enumerate(train_loader):
img = Variable(img.cuda() if not args.no_cuda else img)
desc = Variable(desc.cuda() if not args.no_cuda else desc)
len_desc, indices = torch.sort(len_desc, 0, True)
indices = indices.numpy()
img = img[indices, ...]