ML-Agents-SoccerTwos reinforcement-learning

trained 4499999 steps in colab env 3