Release Notes¶

Upcoming release (`stable/0.2`)¶

This adds a workaround for an issue sometimes encountered when calling TwoQubitWeylDecomposition inside generate_ansatz_from_circuit(). Sometimes an error is thrown despite the input matrix being unitary to numerical precision, and returning is_unitary()`() as True. This is an ongoing issue in qiskit, see https://github.com/Qiskit/qiskit/issues/4159 for more details.

With this workaround, any time the error is encountered, it will retry by transpiling the two-qubit circuit (from which the matrix is derived) prior to passing the matrix to TwoQubitWeylDecomposition. It should be noted that this is does not fix the root issue, and is not guaranteed to always work, but has been observed to work for some instances.

This release renames the primary objective to MaximizeStateFidelity and adds support for Qiskit 2.0.

Adds the parametrize_circuit() function for generating a parametrized version of a given circuit. In contrast to generate_ansatz_from_circuit(), parametrize_circuit does not change the types of gates in the circuit. It simply replaces numerical parameters with free parameters.

The OneMinusFidelity objective function has been renamed and is now known as MaximizeStateFidelity. As a related change, one should now call the loss_function() method to obtain the value and gradient of the loss function, instead of calling the instance directly (through its __call__ method).

Qiskit SDK version 1.3 or higher is now required. Qiskit SDK version 2.0 is now supported.

This release fixes the import of quimb_gate() to work on versions of qiskit-quimb greater than 0.0.5. The aforementioned function is now imported from the toplevel qiskit_quimb instead of qiskit_quimb.circuit.

This release adds quimb-autograd as an optional dependency, so that one can now run pip install 'qiskit-addon-aqc-tensor[quimb-autograd]' to install the addon along with quimb and autograd.

Initial release.