Decision Transformer for Robot Imitation Learning