Fetching the paper…

DeepZero: Scaling up Zeroth-Order Optimization for Deep Model Training · Around