On CY-LG correspondence for (0, 2) toric models