How to use the neuroglancer.PrefetchState function in neuroglancer

def prefetch(self):
    prefetch_states = []
    for i in range(self.num_to_prefetch):
      idx = self.index + (i + 1) * self.batch
      if idx >= len(self.todo):
        break
      prefetch_state = copy.deepcopy(self.viewer.state)
      prefetch_state.layers['seg'].segments = self.list_segments(idx)
      prefetch_state.layout = '3d'
      prefetch_states.append(prefetch_state)

    with self.viewer.config_state.txn() as s:
      s.prefetch = [
          neuroglancer.PrefetchState(state=prefetch_state, priority=-i)
          for i, prefetch_state in enumerate(prefetch_states)
      ]

def set_state(self, viewer, frame_i, prefetch_frames):
        states = self.get_frames(frame_i, frame_i + prefetch_frames)
        viewer.set_state(states[0])
        with viewer.config_state.txn() as s:
            del s.prefetch[:]
            for i, state in enumerate(states[1:]):
                s.prefetch.append(
                    neuroglancer.PrefetchState(state=state, priority=prefetch_frames - i))

with self.viewer.txn() as s:
            modify_state_for_body(s, body)

        prefetch_states = []
        for i in range(self.num_to_prefetch):
            prefetch_index = self.index + i + 1
            if prefetch_index >= len(self.bodies):
                break
            prefetch_state = copy.deepcopy(self.viewer.state)
            prefetch_state.layout = '3d'
            modify_state_for_body(prefetch_state, self.bodies[prefetch_index])
            prefetch_states.append(prefetch_state)

        with self.viewer.config_state.txn() as s:
            s.prefetch = [
                neuroglancer.PrefetchState(state=prefetch_state, priority=-i)
                for i, prefetch_state in enumerate(prefetch_states)
            ]

        label = self.state.body_labels.get(body.segment_id, '')
        with self.viewer.config_state.txn() as s:
            s.status_messages['status'] = (
                '[Segment %d/%d  : %d/%d voxels labeled = %.3f fraction] label=%s' %
                (index, len(self.bodies), self.cumulative_voxels[index], self.total_voxels,
                 self.cumulative_voxels[index] / self.total_voxels, label))

else:
                    cur_state = neuroglancer.ViewerState.interpolate(a, b, t)
                    states_to_capture.append((frame_number, i + t, cur_state))
                    frame_number += 1
        for frame_number, t, cur_state in states_to_capture:
            prefetch_states = [
                x[2]
                for x in states_to_capture[frame_number + 1:frame_number + 1 + num_prefetch_frames]
            ]
            viewer.set_state(cur_state)
            if num_prefetch_frames > 0:
                with viewer.config_state.txn() as s:
                    del s.prefetch[:]
                    for i, state in enumerate(prefetch_states[1:]):
                        s.prefetch.append(
                            neuroglancer.PrefetchState(state=state, priority=num_prefetch_frames - i))
            frame_number, path = saver.capture(frame_number)
            with lock:
                num_frames_written[0] += 1
                cur_num_frames_written = num_frames_written[0]
                print('[%07d/%07d] keypoint %.3f/%5d: %s' %
                      (cur_num_frames_written, total_frames, t, len(keypoints), path))