Target object info missing in the Gym wrapper?

[hnyu]

Currently the gym wrapper resets an episode and simply discards the task descriptions which encode the key information about the target object for a task.

def reset(self) -> Dict[str, np.ndarray]:
        descriptions, obs = self.task.reset()
        del descriptions  # Not used.
        return self._extract_obs(obs)

For example, in the pick_up_cup.py task, there is actually a variation about which cup to pick up and the cup's color might be different each episode.

This target information is not in image inputs. Also in the default get_low_dim_state(), I don't see it. The only place I can see this is in the returned descriptions. One exception is reach_target.py which defines its own get_low_dim_state() and returns the target object position. Am I missing something here?

(Edit: upon further inspection, I found that self._variation_number seems not set (always 0))