57544b1ff9f97d4da9f64d25c8ea5a3d8d247ffc,rllib/examples/centralized_critic_2.py,,,#,110

Before Change

        "opponent_obs": Discrete(6),
        "opponent_action": Discrete(2),
    })
    tune.run(
        "PPO",
        stop={
            "timesteps_total": args.stop,
            "episode_reward_mean": 7.99,
        },
        config={
            "env": TwoStepGame,
            "batch_mode": "complete_episodes",
            "callbacks": FillInActions,
            "num_workers": 0,
            "multiagent": {
                "policies": {
                    "pol1": (None, observer_space, action_space, {}),
                    "pol2": (None, observer_space, action_space, {}),
                },
                "policy_mapping_fn": lambda x: "pol1" if x == 0 else "pol2",
                "observation_fn": central_critic_observer,
            },
            "model": {
                "custom_model": "cc_model",
            },
        })

After Change

        "opponent_action": Discrete(2),
    })

    config = {
        "env": TwoStepGame,
        "batch_mode": "complete_episodes",
        "callbacks": FillInActions,
        "num_workers": 0,
        "multiagent": {
            "policies": {
                "pol1": (None, observer_space, action_space, {}),
                "pol2": (None, observer_space, action_space, {}),
            },
            "policy_mapping_fn": lambda x: "pol1" if x == 0 else "pol2",
            "observation_fn": central_critic_observer,
        },
        "model": {
            "custom_model": "cc_model",
        },
        "use_pytorch": args.torch,
    }

    stop = {
        "training_iteration": args.stop_iters,
        "timesteps_total": args.stop_timesteps,
        "episode_reward_mean": args.stop_reward,
    }

    results = tune.run("PPO", config=config, stop=stop)

    if args.as_test:
        check_learning_achieved(results, args.stop_reward)

Instances