I\'m confused about something in the code of this MLP tutorial using python tensorflow. the function is called in mlp in (preds = tf.nn.softmax(logits)) for validation and