replay_memory.py 文件源码

def get_idxs(self, idxs):
        """
        Returns the states a the provided indexes.

        Args:
            idxs (list): the indexes of the states to return.

        Returns:
            The states at the provided indexes.

        """
        if not self._full and np.any(idxs < self._history_length):
            idxs[np.argwhere(
                idxs < self._history_length).ravel()] += self._history_length

        s = self._get_state(idxs - 1)
        ss = self._get_state(idxs)

        return s, self._actions[idxs - 1, ...], self._rewards[idxs - 1, ...],\
            ss, self._absorbing[idxs - 1, ...], self._last[idxs - 1, ...]