Possibility to render multiple cameras at the same time in dm_control environments about shimmy HOT 6 CLOSED

def _internal_render(
      self,
      mode='rgb_array',
      height: int = HEIGHT,
      width: int = WIDTH,
      camera_id: typing.Optional[typing.Union[str, int]] = None,
      scene_callback = None
  ):
      assert mode in [
          'human', 'rgb_array'
      ], 'only support rgb_array and human mode, given %s' % mode
      if mode == "human":
          self._get_viewer().render()
          return
      else:
          physics = self._env.physics
          if camera_id is not None:
              return physics.render(
                  camera_id=camera_id,
                  height=height,
                  width=width,
              )
          # If no camera_id is specified, render all cameras in a grid.
          num_cameras = physics.model.ncam
          num_columns = int(math.ceil(math.sqrt(num_cameras)))
          num_rows = int(math.ceil(float(num_cameras) / num_columns))
          frame = np.zeros((num_rows * height, num_columns * width, 3), dtype=np.uint8)
          for col in range(num_columns):
              for row in range(num_rows):
                  camera_id = row * num_columns + col
                  if camera_id >= num_cameras:
                      break
                  subframe = physics.render(
                      camera_id=camera_id, height=height, width=width, scene_callback=scene_callback, scene_option=_SCENE_OPTION
                  )
                  frame[
                      row * height : (row + 1) * height, col * width : (col + 1) * width
                  ] = subframe
          return frame