1 paper
Mingxing Rao, Bohan Jiang, Daniel Moyer
In the present work we present Training Noise Token (TNT) Pruning for vision transformers. Our method relaxes the discrete token dropping condition to continuous additive noise, pr…