Tuple<matrix<T, M, N>, > transpose<T, N, M>.apply_bwd(matrix<T, N, M> ) where T : __BuiltinFloatingPointType;