def test_save_image(self):
arr = jt.array(np.random.randn(16, 3, 10, 10))
jt.save_image(arr, "/tmp/a.jpg")
parser.add_argument('--n_col',
type=int,
default=5,
help='number of columns of sample matrix')
parser.add_argument('path', type=str, help='path to checkpoint file')
args = parser.parse_args()
generator = StyledGenerator(512)
ckpt = jt.load(args.path)
generator.load_state_dict(ckpt)
generator.eval()
mean_style = get_mean_style(generator)
step = int(math.log(args.size, 2)) - 2
img = sample(generator, step, mean_style, args.n_row * args.n_col)
jt.save_image(img,
'style_mixing/sample.png',
nrow=args.n_col,
normalize=True,
range=(-1, 1))
for j in range(20):
img = style_mixing(generator, step, mean_style, args.n_col, args.n_row)
jt.save_image(img,
f'style_mixing/sample_mixing_{j}.png',
nrow=args.n_col + 1,
normalize=True,
range=(-1, 1))
def test_save_image(self):
arr = jt.array(np.random.randn(16, 3, 10, 10))
jt.save_image(arr, jt.flags.cache_path + "/tmp/a.jpg")
used_sample += real_image.shape[0]
if (i + 1) % 100 == 0:
images = []
gen_i, gen_j = (10, 5)
with jt.no_grad():
for _ in range(gen_i):
images.append(
g_running(jt.randn(gen_j, code_size),
step=step,
alpha=alpha).data)
jt.save_image(
jt.concat(images, 0),
f'FFHQ/sample/{str(i + 1).zfill(6)}.png',
nrow=gen_i,
normalize=True,
range=(-1, 1),
)
if (i + 1) % 10000 == 0:
jt.save(g_running.state_dict(),
f'FFHQ/checkpoint/{str(i + 1).zfill(6)}.model')
state_msg = (
f'Size: {4 * 2 ** step}; G: {gen_loss_val:.3f}; D: {disc_loss_val:.3f};'
f' Grad: {grad_loss_val:.3f}; Alpha: {alpha:.5f}')
pbar.set_description(state_msg)