Memory Efficient Continual Learning with Transformers