cartpole-pg:
    env: CartPole-v1
    run: PG
    stop:
        sampler_results/episode_reward_mean: 150
        timesteps_total: 100000
    config:
        # Works for both torch and tf.
        framework: torch
        num_workers: 0