Solving a class of non-convex min-max games using adaptive momentum methods