print 'Prepare data loaders...'
transform = transforms.Compose([transforms.ToPILImage(), transforms.RandomHorizontalFlip(),\
transforms.ToTensor(), transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])
trainDataset = CELEBA(root=opts.root, train=True, transform=transform)
trainLoader = torch.utils.data.DataLoader(trainDataset,
batch_size=opts.batchSize,
shuffle=True)
transform = transforms.Compose([ transforms.ToPILImage(), \
transforms.ToTensor(), transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])
testDataset = CELEBA(root=opts.root, train=False, transform=transform)
testLoader = torch.utils.data.DataLoader(testDataset,
batch_size=opts.batchSize,
shuffle=False)
print 'Data loaders ready.'
###### Create model #####
IM_SIZE = 64
gen = GEN(imSize=IM_SIZE, nz=opts.nz, fSize=opts.fSize)
dis = DIS(imSize=IM_SIZE, fSize=opts.fSize, WGAN=opts.WGAN)
gen, dis = train_mode(gen,
dis,
trainLoader,
useNoise=opts.useNoise,
beta1=0.5,
c=opts.c,
k=opts.k,
WGAN=opts.WGAN)
])
testDataset = CELEBA(
root=opts.root,
train=False,
labels=opts.labels,
transform=transform,
Ntest=1000
) #most models trained with Ntest=1000, but using 100 to prevent memory errors
testLoader = torch.utils.data.DataLoader(testDataset,
batch_size=opts.batchSize,
shuffle=False)
print 'Data loader ready.'
#Load model
gen = GEN(imSize=IM_SIZE, nz=opts.nz, fSize=opts.fSize)
if gen.useCUDA:
torch.cuda.set_device(opts.gpuNo)
gen.cuda()
gen.load_params(opts.exDir, gpuNo=opts.gpuNo)
gen.eval()
print 'params loaded'
# Get the data:
data = iter(testLoader).next()
x, y = prep_data(data, useCUDA=gen.useCUDA)
# Get men with glasses
idx_men_w_glasses = np.argwhere(
torch.prod(y.data == torch.Tensor([1, 1]).cuda(), dim=1))[0]
img_men_w_glasses = x[idx_men_w_glasses[:opts.numSamples]]
print 'already exists'
print 'Outputs will be saved to:', exDir
save_input_args(exDir, opts)
####### Test data set #######
IM_SIZE = opts.imSize
print 'Prepare data loaders...'
transform = transforms.Compose([ transforms.ToPILImage(), transforms.Resize((IM_SIZE, IM_SIZE)), \
transforms.ToTensor(), transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])
if opts.data == 'CELEBA':
testDataset = CELEBA(
root=opts.root, train=False, transform=transform, Ntest=100
) #most models trained with Ntest=1000, but using 100 to prevent memory errors
gen = GEN(imSize=IM_SIZE, nz=opts.nz, fSize=opts.fSize)
elif opts.data == 'OMNI':
print 'using Omniglot eval dataset...'
testDataset = OMNI(root=opts.root, train=False, transform=transform)
gen = GEN1D(imSize=IM_SIZE, nz=opts.nz, fSize=opts.fSize)
else:
testDataset = SHOES(root=opts.root, train=False, transform=transform)
gen = GEN(imSize=IM_SIZE, nz=opts.nz, fSize=opts.fSize)
testLoader = torch.utils.data.DataLoader(testDataset,
batch_size=opts.batchSize,
shuffle=False)
print 'Data loaders ready.'
print "Test Data:", len(testDataset)
###### Create model and load parameters #####