f33e7b5201494eb52360efc14dbfc7e6eeb02096,examples/mujoco_all_sac_real_nvp_hierarchy.py,,run_experiment,#Any#,191

Before Change

        ant_env = normalize(AntEnv())
        env = HierarchyProxyEnv(wrapped_env=ant_env,
                                low_level_policy=low_level_policy)
    elif variant["env_name"] == "random-goal-swimmer":
        random_goal_swimmer_env = normalize(RandomGoalSwimmerEnv(
            reward_type=variant["env_reward_type"],
            goal_reward_weight=variant["env_goal_reward_weight"],
            goal_radius=variant["env_goal_radius"],
            terminate_at_goal=variant["env_terminate_at_goal"],
        ))
        env = HierarchyProxyEnv(wrapped_env=random_goal_swimmer_env,
                                low_level_policy=low_level_policy)

    elif variant["env_name"] == "random-goal-ant":
        random_goal_ant_env = normalize(RandomGoalAntEnv(
            reward_type=variant["env_reward_type"],
            goal_reward_weight=variant["env_goal_reward_weight"],
            goal_radius=variant["env_goal_radius"],
            terminate_at_goal=variant["env_terminate_at_goal"],
        ))
        env = HierarchyProxyEnv(wrapped_env=random_goal_ant_env,
                                low_level_policy=low_level_policy)

    elif variant["env_name"] == "humanoid-rllab":
        humanoid_env = normalize(HumanoidEnv())
        env = HierarchyProxyEnv(wrapped_env=humanoid_env,
                                low_level_policy=low_level_policy)

After Change

    env_name = variant["env_name"]
    env_type = env_name.split("-")[-1]

    if "random-goal" in env_name:
        EnvClass = RANDOM_GOAL_ENVS[env_type]
        env_args = {
            name.replace("env_", "", 1): value
            for name, value in variant.items()
            if name.startswith("env_") and name != "env_name"
        }
        env = normalize(EnvClass(**env_args))
    elif "rllab" in variant["env_name"]:
        EnvClass = RLLAB_ENVS[variant["env_name"]]
        base_env = normalize(EnvClass())
        env = HierarchyProxyEnv(wrapped_env=base_env,
                                low_level_policy=low_level_policy)
    else:
        raise NotImplementedError

    pool = SimpleReplayBuffer(
        env_spec=env.spec,
        max_replay_buffer_size=variant["max_pool_size"],
    )

Instances