Hi @YJiangcm, the performance of the model trained using the code in this repository differs from that claimed in the paper.

At your convenience, would you provide the trained models for evaluation please? Thank you so much.