0241          ; mapped                 ; 0242          # 4.1  LATIN CAPITAL LETTER GLOTTAL STOP
0242          ; valid                                  # 5.0  LATIN SMALL LETTER GLOTTAL STOP
0243          ; mapped                 ; 0180          # 5.0  LATIN CAPITAL LETTER B WITH STROKE
0244          ; mapped                 ; 0289          # 5.0  LATIN CAPITAL LETTER U BAR

        `(batch_size, 1)`. This enables `Model` subclasses to process scalar
        data in their `train_step()`/`test_step()`/`predict_step()` methods. \
        Note that this change may break certain subclassed models. You can
        revert back to the previous behavior by adding upranking yourself in the
        `train_step()`/`test_step()`/`predict_step()` methods, e.g. `if
        x.shape.rank == 1: x = tf.expand_dims(x, axis=-1)`. Functional models as