Args: player (gym.Env) predictor (callable): the model forward function which takes a state and returns the prediction. memory (ReplayMemory): the replay memory to store experience to. history_len (int):
(self, player, predictor, memory, history_len)
| 116 | and fill the results to experience replay buffer. |
| 117 | """ |
| 118 | def __init__(self, player, predictor, memory, history_len): |
| 119 | """ |
| 120 | Args: |
| 121 | player (gym.Env) |
| 122 | predictor (callable): the model forward function which takes a |
| 123 | state and returns the prediction. |
| 124 | memory (ReplayMemory): the replay memory to store experience to. |
| 125 | history_len (int): |
| 126 | """ |
| 127 | self.player = player |
| 128 | self.num_actions = player.action_space.n |
| 129 | self.predictor = predictor |
| 130 | self.memory = memory |
| 131 | self.state_shape = memory.state_shape |
| 132 | self.dtype = memory.dtype |
| 133 | self.history_len = history_len |
| 134 | |
| 135 | self._current_episode = [] |
| 136 | self._current_ob = player.reset() |
| 137 | self._current_game_score = StatCounter() # store per-step reward |
| 138 | self.total_scores = [] # store per-game total score |
| 139 | |
| 140 | self.rng = get_rng(self) |
| 141 | |
| 142 | def step(self, exploration): |
| 143 | """ |
nothing calls this directly
no test coverage detected