Reinforcement Learning - PacMan runs for only one episode

Question

I\'m trying to train an agent based on pacman but have the problem that it only runs for one episode. Since one episode comprises of three lives, I added the variable dead