Trinity S³AI
Adam: the optimizer most networks are trained with
Watch Adam carry a weight from 0 to the minimum at 3, then set three gradients of any size: the bias correction makes the first step lr times their sign, whatever their size.
Trinity S³AI
Watch Adam carry a weight from 0 to the minimum at 3, then set three gradients of any size: the bias correction makes the first step lr times their sign, whatever their size.