Method _before_train

examples/DeepQNetwork/expreplay.py:384–390

(self)

        self.predictor = self.trainer.get_predictor(*self.predictor_io_names)

    def _before_train(self):
        env_runners = [
            EnvRunner(self.get_player(), self.predictor, self.mem, self.history_len)
            for k in range(self.num_parallel_players)
        ]
        self.runner = EnvRunnerManager(env_runners, self.update_frequency * 2)
        self._init_memory()

    def _trigger(self):
        mean, max = self.runner.reset_stats()