Presumably if you have a GPU backed cloud DBN. Hyper parameter selection is faster than one param per day. Also how to you choose the parameters to the hyper pa

Presumably if you have a GPU backed cloud DBN. Hyper parameter selection is faster than one param per day. Also how to you choose the parameters to the hyper parameter tuner? I am never convinced these things work given no free lunch theorem.