Goal-Conditioned Reinforcement Learning (Manipulation) on puzzle-3x3-play state-based v0 (test)