Flattening Sharpness for Dynamic Gradient Projection Memory Benefits Continual Learning

Open in new window