why to pad features with zero vector

[anny0316]

In supervised_train.py, why to pad features with zero vector?

"if not features is None: # pad with dummy zero vector features = np.vstack([features, np.zeros((features.shape[1],))])"