Parameter-Efficient Mixture-of-Experts Architecture for Pre-trained Language Models

4934

Knowledge Triples