Abstract: To transfer the representation capacity of large pre-trained models to lightweight models, knowledge distillation has been widely explored. However, conventional single-stage distillation ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results