Text this: Exploring in and out-of-equilibrium learning regimes of restricted Boltzmann machines