rewriting tests

    labels = 2

    logits = 0.0001

    input_tensor = np.random.uniform(labels, logits, [1, 1])

    logits_tensor = np.random.uniform(logits)

    labesl_tensor = np.random.rand(labels)

    with self.test_session() as sess:

      input_tensor = tf.convert_to_tensor(input_tensor, dtype=tf.float32)

      out_tensor = Hingeloss()(input_tensor)

      logits_tensor = tf.convert_to_tensor(logits_tensor, dtype=tf.float32)

      labels_tesnor = tf.convert_to_tensor(labels_tensor, dtype=tf.float32)

      out_tensor = Hingeloss()(labels_tensor, logits_tensor)

      out_tensor = out_tensor.eval()

      assert type(out_tensor) is FloatType

      assert out_tensor.shape == (labels,)

def test_hingeloss_pickle():

  lables = 2

  logits = 0.001

  tg = TensorGraph()

  feature = Feature(shape=(tg.batch_size, 1))

  layer = Hingeloss(in_layers=[feature, feature])

  layer = Hingeloss(labels=labels, logits=logits, in_layers=[feature, feature])

  tg.add_output(layer)

  tg.set_loss(layer)

  tg.build()