Does Self-Attention Need Separate Weights in Transformers?