1 paper
Bryan A. Plummer, Shan Lu, Keval Doshi +4
Weight Factorization (WF) methods aim to reduce a model's latency, memory, or both. These typically decompose a pretrained model into two or more factors that approximate the origi…