def sample(epoch):
print('sampling')
fgen.eval()
n = 256
taus = [0.7, 0.8, 0.9, 1.0]
for t in taus:
z = torch.randn(n, 3, imageSize, imageSize).to(device)
z = z * t
img, _ = fgen.decode(z)
img = postprocess(img, n_bits)
image_file = 'sample{}.t{:.1f}.png'.format(epoch, t)
save_image(img, os.path.join(result_path, image_file), nrow=16)
def reconstruct(epoch):
print('reconstruct')
fgen.eval()
n = 16
np.random.shuffle(test_index)
img, _ = get_batch(test_data, test_index[:n])
img = preprocess(img.to(device), n_bits)
z, _ = fgen.encode(img)
img_recon, _ = fgen.decode(z)
abs_err = img_recon.add(img * -1).abs()
print('Err: {:.4f}, {:.4f}'.format(abs_err.max().item(), abs_err.mean().item()))
img = postprocess(img, n_bits)
img_recon = postprocess(img_recon, n_bits)
comparison = torch.cat([img, img_recon], dim=0).cpu()
reorder_index = torch.from_numpy(np.array([[i + j * n for j in range(2)] for i in range(n)])).view(-1)
comparison = comparison[reorder_index]
image_file = 'reconstruct{}.png'.format(epoch)
save_image(comparison, os.path.join(result_path, image_file), nrow=16)