TY  - RPRT
TI  - DEFT: Exploiting Gradient Norm Difference between Model Layers for Scalable Gradient Sparsification
AU  - Daegun Yoon
AU  - Sangyoon Oh
PY  - 2023
DO  - 10.1145/3605573.3605609
UR  - https://arxiv.org/abs/2307.03500
ID  - 2307.03500
ER  -