ut3_to_t3
=========

.. py:function:: t3toolbox.backend.ut3_conversions.ut3_to_t3(x)

   .. code-block:: python

      def ut3_to_t3(
              x: typ.Tuple[
                  NDArray,                          # tucker_supercore
                  NDArray,                          # tt_supercore
                  typ.Tuple[int, ...],              # shape, static int tuple
                  typ.Tuple[NDArray, NDArray],      # (tucker_edge_mask, tt_edge_mask)
              ],
      ) -> typ.Union[
          typ.Tuple[typ.Tuple[NDArray, ...], typ.Tuple[NDArray, ...]],  # (tucker_cores, tt_cores), if unstacked
          typ.Tuple,                                                     # else a nested tree (shape stack_shape) of those
      ]:


   Convert uniform supercores + masks back to ragged TuckerTensorTrain core pairs.

   Unstacked: returns one ``(tucker_cores, tt_cores)``. Stacked: returns a nested tuple (shaped like
   ``stack_shape``) of such pairs -- a *tree*, since a varying-rank stack has no single stacked
   ``TuckerTensorTrain`` (``docs/uniform_ranks_and_varieties.md``). The real sub-blocks are selected *through
   the rank masks* (boolean indexing, ascending order) rather than by slicing a prefix, since an edge mask
   may be gappy after ``+``/``x`` (``docs/uniform_masks_vs_ranks.md``).