exit(0)
if len(args.gpus) > 0:
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpus
torch_devs = torch.device("cuda" if torch.cuda.is_available() else "cpu")
itr_out_dir = args.expName + '-itrOut'
if os.path.isdir(itr_out_dir):
shutil.rmtree(itr_out_dir)
os.mkdir(itr_out_dir) # to save temp output
# redirect print to a file
if args.print == 0:
sys.stdout = open(os.path.join(itr_out_dir, 'iter-prints.log'), 'w')
mb_data_iter = bkgdGen(data_generator=gen_train_batch_bg(mbsize=args.mbsize, \
psz=args.psz, nvar=args.nvar), \
max_prefetch=16)
def main(args):
model = unet()
# model = DnCNN(1, num_of_layers = 8)
_ = model.apply(model_init) # init model weights and bias
masker = Masker(width=4, mode='zero')
if torch.cuda.is_available():
if torch.cuda.device_count() > 1:
model = torch.nn.DataParallel(model)
model = model.to(torch_devs)
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpus
os.environ[
'TF_CPP_MIN_LOG_LEVEL'] = '3' # disable printing INFO, WARNING, and ERROR
itr_out_dir = args.expName + '-itrOut'
if os.path.isdir(itr_out_dir):
shutil.rmtree(itr_out_dir)
os.mkdir(itr_out_dir) # to save temp output
# redirect print to a file
if args.print == 0:
sys.stdout = open('%s/%s' % (itr_out_dir, 'iter-prints.log'), 'w')
# build minibatch data generator with prefetch
mb_data_iter = bkgdGen(data_generator=gen_train_batch_bg(
dsfn=args.dsfn, mb_size=args.mbsz, \
in_depth=args.depth, img_size=args.psz), \
max_prefetch=args.mbsz*4)
generator = make_generator_model(input_shape=(None, None, args.depth),
nlayers=args.lunet)
discriminator = make_discriminator_model(input_shape=(args.psz, args.psz, 1))
feature_extractor_vgg = tf.keras.applications.VGG19(\
weights='vgg19_weights_notop.h5', \
include_top=False)
# This method returns a helper function to compute cross entropy loss
cross_entropy = tf.keras.losses.BinaryCrossentropy(from_logits=True)