def __init__()

in rlmeta/agents/ppo/ppo_agent.py [0:0]

• 36 lines of code
• 2 McCabe index (conditional complexity)

    def __init__(self,
                 model: ModelLike,
                 deterministic_policy: bool = False,
                 replay_buffer: Optional[ReplayBufferLike] = None,
                 controller: Optional[ControllerLike] = None,
                 optimizer: Optional[torch.optim.Optimizer] = None,
                 batch_size: int = 128,
                 grad_clip: float = 50.0,
                 gamma: float = 0.99,
                 gae_lambda: float = 0.95,
                 eps_clip: float = 0.2,
                 entropy_ratio: float = 0.01,
                 advantage_normalization: bool = True,
                 reward_rescaling: bool = True,
                 value_clip: bool = True,
                 push_every_n_steps: int = 1) -> None:
        super(PPOAgent, self).__init__()

        self.model = model
        self.deterministic_policy = deterministic_policy

        self.replay_buffer = replay_buffer
        self.controller = controller

        self.optimizer = optimizer
        self.batch_size = batch_size
        self.grad_clip = grad_clip

        self.gamma = gamma
        self.gae_lambda = gae_lambda
        self.eps_clip = eps_clip
        self.entropy_ratio = entropy_ratio
        self.advantage_normalization = advantage_normalization
        self.reward_rescaling = reward_rescaling
        if self.reward_rescaling:
            self.reward_rescaler = NormRescaler(size=1)
        self.value_clip = value_clip

        self.push_every_n_steps = push_every_n_steps
        self.done = False
        self.trajectory = []