def __init__(self, image_size, max_classes=None, class_temperature=2.):
super().__init__()
assert image_size in (
128, 256, 512), 'image size must be one of 128, 256, or 512'
self.biggan = BigGAN.from_pretrained(f'biggan-deep-{image_size}')
self.max_classes = max_classes
self.class_temperature = class_temperature
self.init_latents()
def __init__(self,
image_size,
max_classes=None,
class_temperature=2.0,
ema_decay=0.99):
super().__init__()
assert image_size in (
128,
256,
512,
), "image size must be one of 128, 256, or 512"
self.biggan = BigGAN.from_pretrained(f"biggan-deep-{image_size}")
self.max_classes = max_classes
self.class_temperature = class_temperature
self.ema_decay = ema_decay
self.init_latents()
# ladataan latenssitiedot
lat1 = torch.load(opt.lat1)
lat2 = torch.load(opt.lat2)
best1 = lat1.best
best2 = lat2.best
noise1 = lat1.normu.to(device)
noise2 = lat2.normu.to(device)
class1 = lat1.cls.to(device)
class2 = lat2.cls.to(device)
# ladataan biggan
# load biggan
model = BigGAN.from_pretrained(f'biggan-deep-{imgSize}')
model.eval()
truncation = opt.trunc
model.to(device)
#n_delta = (noise2 - noise1) / opt.steps
#c_delta = (class2 - class1) / opt.steps
#noise_vector = noise1
#class_vector = class1
alphas = np.linspace(0., 1., opt.steps)
with torch.no_grad():
for i in range(0, opt.steps):
def __init__(self, image_size):
super().__init__()
assert image_size in (
128, 256, 512), 'image size must be one of 128, 256, or 512'
self.biggan = BigGAN.from_pretrained(f'biggan-deep-{image_size}')
self.init_latents()