if opt.dataset == 'mnist':
os.makedirs('data/mnist', exist_ok=True)
dataloader_x = torch.utils.data.DataLoader(
dset.MNIST('data/mnist', train=True, download=True,
transform=transforms.Compose([
transforms.Resize(opt.imageSize),
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
])),
batch_size=opt.batchSize, shuffle=True)
os.makedirs('data/mnistm', exist_ok=True)
dataloader_y = torch.utils.data.DataLoader(
mnistm.MNISTM('data/mnistm', train=True, download=True,
transform=transforms.Compose([
transforms.Resize(opt.imageSize),
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
])),
batch_size=opt.batchSize, shuffle=True)
#
elif opt.dataset == 'photo-monet':
dataset1 = dset.ImageFolder(root=opt.dataroot+'/train_monet',
transform=transforms.Compose([
transforms.Scale(opt.imageSize),
transforms.CenterCrop(opt.imageSize),
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5)),
]))
dataset2 = dset.ImageFolder(root=opt.dataroot+'/train_photo',
transform=transforms.Compose([
transforms.Scale(opt.imageSize),
train=True,
download=True,
transform=transforms.Compose([
transforms.Resize(opt.img_size),
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
])),
batch_size=opt.batch_size,
shuffle=True)
os.makedirs('../../data/mnistm', exist_ok=True)
dataloader2 = torch.utils.data.DataLoader(mnistm.MNISTM(
'../../data/mnistm',
train=True,
download=True,
transform=transforms.Compose([
transforms.Resize(opt.img_size),
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
])),
batch_size=opt.batch_size,
shuffle=True)
# Optimizers
optimizer_G = torch.optim.Adam(coupled_generators.parameters(),
lr=opt.lr,
betas=(opt.b1, opt.b2))
optimizer_D = torch.optim.Adam(coupled_discriminators.parameters(),
lr=opt.lr,
betas=(opt.b1, opt.b2))
transforms.ToTensor(),
transforms.Normalize([0.5], [0.5])
# transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
])
# Loading dataset
train_data = datasets.MNIST("~/Desktop/Datasets/", transform=transform)
train_loader = torch.utils.data.DataLoader(train_data, **kwargs)
train_loader2 = torch.utils.data.DataLoader(
mnistm.MNISTM(
"~/Desktop/Datasets/",
train=True,
download=True,
transform=transforms.Compose([
transforms.Resize(ndf),
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5)),
]),
),
batch_size=args.batch_size,
shuffle=True,
)
# Initialize weights
def weight_init(m):
classname = m.__class__.__name__
if classname.find('Conv') != -1:
torch.nn.init.normal_(m.weight, 0.0, 0.02)
elif classname.find('BatchNorm') != -1:
adversarial_loss = nn.MSELoss()
# Initialize models
coupled_generators = CoupledGenerators()
coupled_discriminators = CoupledDiscriminators()
print(coupled_generators)
print(coupled_discriminators)
transform = transform.Compose([
transform.Resize(opt.img_size),
transform.ImageNormalize(mean=[0.5], std=[0.5]),
])
dataloader1 = MNIST(train=True, transform=transform).set_attrs(batch_size=opt.batch_size, shuffle=True)
dataloader2 = mnistm.MNISTM(mnist_root = "../../data/mnistm", train=True, transform = transform).set_attrs(batch_size=opt.batch_size, shuffle=True)
# Optimizers
optimizer_G = nn.Adam(coupled_generators.parameters(), lr=opt.lr, betas=(opt.b1, opt.b2))
optimizer_D = nn.Adam(coupled_discriminators.parameters(), lr=opt.lr, betas=(opt.b1, opt.b2))
# ----------
# Training
# ----------
for epoch in range(opt.n_epochs):
for i, ((imgs1, _), (imgs2, _)) in enumerate(zip(dataloader1, dataloader2)):
jt.sync_all(True)
batch_size = imgs1.shape[0]
# Adversarial ground truths
valid = jt.ones([batch_size, 1]).float32().stop_grad()
transforms.Resize(args.img_size),
transforms.ToTensor(),
transforms.Normalize([0.5], [0.5])
]),
),
batch_size=args.batch_size,
shuffle=True,
)
os.makedirs("../../data/mnistm", exist_ok=True)
dataloader2 = torch.utils.data.DataLoader(
mnistm.MNISTM(
"../../data/mnistm",
train=True,
download=True,
transform=transforms.Compose([
transforms.Resize(args.img_size),
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5)),
]),
),
batch_size=args.batch_size,
shuffle=True,
)
lr = args.lr
b1, b2 = args.b1, args.b2
# Optimizers
optimizer_G = torch.optim.Adam(coupled_generators.parameters(),
lr=lr,
betas=(b1, b2))