Transposed shape in lookup parameters?

`dy.Model().add_lookup_parameters((2, 3)).shape()` returns `(3, 2)`.

The doc says that:
- The first argument of `add_lookup_parameters` is the "shape of the parameter. The first dimension is the vocab size" [[1]](https://dynet.readthedocs.io/en/latest/python_ref.html#dynet.ParameterCollection.add_lookup_parameters)
- When called on a lookup parameter, `shape()` returns the "shape of the lookup parameter. The first dimension is the lookup dimension" [[2]](https://dynet.readthedocs.io/en/latest/python_ref.html#dynet.LookupParameters.shape)

Unless I'm mistaken, "vocab size" and "size of the lookup dimension" refer to the same thing, so shouldn't `shape()` return the declared shape?