【论文阅读】【ViT系列】DeiT:数据高效的图像transformers的训练&通过注意力的蒸馏
论文:Trainingdata-efficientimagetransformers&distillationthroughattention代码:https://github.com/facebookresearch/deit目录1主要贡献2原理2.1VisionTransformer2.2Distillationthroughattention2.2.1软蒸馏2.2.2硬蒸馏2.2.3Dist