The current interface of the methods in qudit_state_transform.py ask for a qudit_dimension and num_qudits, the implicit assumption being that all qudits in the system are of the same dimension. We should add support for systems having a mix of different dimensional qudits.
This would require changes to the tensor reshaping and remapping logic to generalize the transforms.
The interface can then accept a qid_shape tuple attribute similar to many Cirq methods, which will also be a nice consistency benefit.