def get_media(self,media_id,file_path): """获取media""" try: url = self.create_get_media_url(media_id) data = HttpClient().get(url) utils.mkdir(os.path.dirname(file_path)) f = open(file_path, 'wb') im = Image.open(BytesIO(data)) im_file = BytesIO() im.save(im_file, format='png') im_data = im_file.getvalue() f.write(im_data) f.close() return 1 except: return 0
os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID" flags = tf.app.flags FLAGS = flags.FLAGS flags.DEFINE_integer('batch_size', 64, '') flags.DEFINE_integer('max_iter', 200000, '') flags.DEFINE_integer('snapshot_interval', 1000, 'interval of snapshot') flags.DEFINE_integer('evaluation_interval', 5000, 'interval of evalution') flags.DEFINE_integer('display_interval', 100, 'interval of displaying log to console') flags.DEFINE_float('adam_alpha', 0.0001, 'learning rate') flags.DEFINE_float('adam_beta1', 0.5, 'beta1 in Adam') flags.DEFINE_float('adam_beta2', 0.999, 'beta2 in Adam') flags.DEFINE_integer('n_dis', 1, 'n discrminator train') mkdir('tmp') os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID" INCEPTION_FILENAME = 'inception_score.pkl' config = FLAGS.__flags config = {k: FLAGS[k]._value for k in FLAGS} generator = DCGANGenerator(**config) discriminator = SNDCGAN_Discrminator(**config) data_set = Cifar10(batch_size=FLAGS.batch_size) global_step = tf.Variable(0, name="global_step", trainable=False) increase_global_step = global_step.assign(global_step + 1) is_training = tf.placeholder(tf.bool, shape=()) z = tf.placeholder(tf.float32, shape=[None, generator.generate_noise().shape[1]])
if FLAGS.isResNet is False: tmp_dir = 'tmp' snapshot_dir = 'snapshots' generator = DCGANGenerator(**config) discriminator = SNDCGAN_Discrminator(**config) else: tmp_dir = 'resnet_tmp' snapshot_dir = 'resnet_snapshots' generator = ResNetGenerator(**config) discriminator = ResNetDiscrminator(**config) INCEPTION_FILENAME = tmp_dir + '/' + INCEPTION_FILENAME FID_FILENAME = tmp_dir + '/' + FID_FILENAME mkdir(tmp_dir) global_step = tf.Variable(0, name="global_step", trainable=False) increase_global_step = global_step.assign(global_step + 1) is_training = tf.placeholder(tf.bool, shape=()) z = tf.placeholder(tf.float32, shape=[None, generator.generate_noise().shape[1]]) x_hat = generator(z, is_training=is_training) x = tf.placeholder(tf.float32, shape=x_hat.shape) d_fake = discriminator(x_hat, update_collection=None) # Don't need to collect on the second call, put NO_OPS d_real = discriminator(x, update_collection="NO_OPS") # Softplus at the end as in the official code of author at chainer-gan-lib github repository # d_loss = tf.reduce_mean(tf.nn.softplus(d_fake) + tf.nn.softplus(-d_real)) + 1e-3 * tf.reduce_mean(tf.get_collection('partialL2')) d_loss = tf.reduce_mean(tf.nn.relu(1.0 - d_real)) + tf.reduce_mean(
from hmc import HMCSampler flags = tf.app.flags FLAGS = flags.FLAGS flags.DEFINE_integer('batch_size', 64, '') flags.DEFINE_integer('max_iter', 200000, '') flags.DEFINE_integer('snapshot_interval', 1000, 'interval of snapshot') flags.DEFINE_integer('evaluation_interval', 5000, 'interval of evalution') flags.DEFINE_integer('display_interval', 100, 'interval of displaying log to console') flags.DEFINE_float('adam_alpha', 1e-4, 'learning rate') flags.DEFINE_float('adam_beta1', 0.0, 'beta1 in Adam') flags.DEFINE_float('adam_beta2', 0.999, 'beta2 in Adam') flags.DEFINE_integer('n_dis', 1, 'n discrminator train') mkdir('results') mkdir('results/tmp') os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID" INCEPTION_FILENAME = 'inception_score.pkl' config = FLAGS.__flags config = {k: FLAGS[k]._value for k in FLAGS} generator = Generator(**config) discriminator = Discrminator(**config) data_set = Cifar10(batch_size=FLAGS.batch_size) global_step = tf.Variable(0, name="global_step", trainable=False) increase_global_step = global_step.assign(global_step + 1) is_training = tf.placeholder(tf.bool, shape=()) z = tf.placeholder(tf.float32,