1 paper · 1 filter
Cong Xu, Wenbin Liang, Mo Yu +7
The rapid scaling of models has led to prohibitively high training and fine-tuning costs. A major factor accounting for memory consumption is the widespread use of stateful optimiz…