Text this: Accelerating phase-field-based microstructure evolution predictions via surrogate models trained by machine learning methods