cartpole-impala-tf:
    env: CartPole-v0
    run: IMPALA
    stop:
        episode_reward_mean: 150
        timesteps_total: 500000
    config:
        use_pytorch: false
        num_gpus: 0