factor out shared code in qaoa proxies (#1091)

most of the code in QAOAFermionicSwapProxy is identical to that of
QAOAVanillaProxy, so instead of repeating it we can inherit

checks-superstaq/checks_superstaq/checks-pyproject.toml

@@ -137,7 +137,7 @@ enable = [
 accept-no-param-doc=false # Do not accept totally missing parameter documentation.
 accept-no-raise-doc=false # Do not accept totally missing raises documentation.
 accept-no-return-doc=false # Do not accept totally missing return documentation.
-accept-no-yield-doc=false # Do not accept missing yields annotations.
+accept-no-yields-doc=false # Do not accept missing yields annotations.
 default-docstring-type = "google"
 
 # Ignore long lines containing urls or pylint directives.

supermarq-benchmarks/supermarq/benchmarks/qaoa_fermionic_swap_proxy.py

@@ -2,18 +2,13 @@
 
 from __future__ import annotations
 
-from collections.abc import Mapping
-
 import cirq
 import numpy as np
-import numpy.typing as npt
-import scipy
 
-import supermarq
-from supermarq.benchmark import Benchmark
+from supermarq.benchmarks.qaoa_vanilla_proxy import QAOAVanillaProxy
 
 
-class QAOAFermionicSwapProxy(Benchmark):
+class QAOAFermionicSwapProxy(QAOAVanillaProxy):
     """Proxy of a full Quantum Approximate Optimization Algorithm (QAOA) benchmark.
 
     This benchmark targets MaxCut on a Sherrington-Kirkpatrick (SK) model. Device
@@ -34,28 +29,7 @@ class QAOAFermionicSwapProxy(Benchmark):
     #. Finding approximately optimal angles (rather than random values)
     """
 
-    def __init__(self, num_qubits: int) -> None:
-        """Generate a new benchmark instance.
-
-        Args:
-            num_qubits: The number of nodes (qubits) within the SK graph.
-        """
-        self.num_qubits = num_qubits
-        self.hamiltonian = self._gen_sk_hamiltonian()
-        self.params = self._gen_angles()
-
-    def _gen_sk_hamiltonian(self) -> list[tuple[int, int, float]]:
-        """Randomly pick +1 or -1 for each edge weight."""
-        hamiltonian = []
-        for i in range(self.num_qubits):
-            for j in range(i + 1, self.num_qubits):
-                hamiltonian.append((i, j, np.random.choice([-1, 1])))
-
-        np.random.shuffle(hamiltonian)
-
-        return hamiltonian
-
-    def _gen_swap_network(self, gamma: float, beta: float) -> cirq.Circuit:
+    def _gen_ansatz(self, gamma: float, beta: float) -> cirq.Circuit:
         qubits = cirq.LineQubit.range(self.num_qubits)
         circuit = cirq.Circuit()
 
@@ -105,90 +79,5 @@ def _gen_swap_network(self, gamma: float, beta: float) -> cirq.Circuit:
         return circuit
 
     def _get_energy_for_bitstring(self, bitstring: str) -> float:
-        energy_val = 0.0
-        for i, j, weight in self.hamiltonian:
-            if bitstring[i] == bitstring[j]:
-                energy_val -= weight  # if edge is UNCUT, weight counts against objective
-            else:
-                energy_val += weight  # if edge is CUT, weight counts towards objective
-        return energy_val
-
-    def _get_expectation_value_from_probs(self, probabilities: Mapping[str, float]) -> float:
-        expectation_value = 0.0
-        for bitstring, probability in probabilities.items():
-            expectation_value += probability * self._get_energy_for_bitstring(bitstring)
-        return expectation_value
-
-    def _get_opt_angles(self) -> tuple[npt.NDArray[np.float_], float]:
-        def f(params: npt.NDArray[np.float_]) -> float:
-            """The objective function to minimize.
-
-            Args:
-                params: The parameters at which to evaluate the objective.
-
-            Returns:
-                Evaluation of objective given parameters.
-            """
-            gamma, beta = params
-            circ = self._gen_swap_network(gamma, beta)
-            # Reverse bitstring ordering due to SWAP network
-            raw_probs = supermarq.simulation.get_ideal_counts(circ)
-            probs = {bitstring[::-1]: probability for bitstring, probability in raw_probs.items()}
-            h_expect = self._get_expectation_value_from_probs(probs)
-
-            return -h_expect  # because we are minimizing instead of maximizing
-
-        init_params = [np.random.uniform() * 2 * np.pi, np.random.uniform() * 2 * np.pi]
-        out = scipy.optimize.minimize(f, init_params, method="COBYLA")
-
-        return out["x"], out["fun"]
-
-    def _gen_angles(self) -> npt.NDArray[np.float_]:
-        # Classically simulate the variational optimization 5 times,
-        # return the parameters from the best performing simulation
-        best_params, best_cost = np.zeros(2), 0.0
-        for _ in range(5):
-            params, cost = self._get_opt_angles()
-            if cost < best_cost:
-                best_params = params
-                best_cost = cost
-        return best_params
-
-    def circuit(self) -> cirq.Circuit:
-        """Generate a QAOA circuit for the Sherrington-Kirkpatrick model.
-
-        This particular benchmark utilizes a quantum circuit structure called
-        the fermionic swap network. We restrict the depth of this proxy benchmark
-        to p=1 to keep the classical simulation scalable.
-
-        Returns:
-            The S-K model QAOA `cirq.Circuit`.
-        """
-        gamma, beta = self.params
-        return self._gen_swap_network(gamma, beta)
-
-    def score(self, counts: Mapping[str, float]) -> float:
-        """Compare the experimental output to the output of noiseless simulation.
-
-        The implementation here has exponential runtime and would not scale. However, it could in
-        principle be done efficiently via https://arxiv.org/abs/1706.02998, so we're good.
-
-        Args:
-            counts: A dictionary containing the measurement counts from circuit execution.
-
-        Returns:
-            The QAOA Fermionic SWAP proxy benchmark score.
-        """
         # Reverse bitstring ordering due to SWAP network
-        raw_probs = supermarq.simulation.get_ideal_counts(self.circuit())
-        ideal_counts = {
-            bitstring[::-1]: probability for bitstring, probability in raw_probs.items()
-        }
-        total_shots = sum(counts.values())
-        # Reverse the order of the bitstrings due to the fermionic swap ansatz
-        experimental_counts = {k[::-1]: v / total_shots for k, v in counts.items()}
-
-        ideal_value = self._get_expectation_value_from_probs(ideal_counts)
-        experimental_value = self._get_expectation_value_from_probs(experimental_counts)
-
-        return 1 - abs(ideal_value - experimental_value) / (2 * ideal_value)
+        return super()._get_energy_for_bitstring(bitstring[::-1])

supermarq-benchmarks/supermarq/benchmarks/qaoa_vanilla_proxy.py

@@ -101,7 +101,7 @@ def f(params: npt.NDArray[np.float_]) -> float:
             """The objective function to minimize.
 
             Args:
-                params: parameters for objective.
+                params: The parameters at which to evaluate the objective.
 
             Returns:
                 Evaluation of objective given parameters.
@@ -137,7 +137,7 @@ def circuit(self) -> cirq.Circuit:
         the classical simulation scalable.
 
         Returns:
-            The S-K model QAOA circuit.
+            The S-K model QAOA `cirq.Circuit`.
         """
         gamma, beta = self.params
         return self._gen_ansatz(gamma, beta)
@@ -149,10 +149,10 @@ def score(self, counts: Mapping[str, float]) -> float:
         principle be done efficiently via https://arxiv.org/abs/1706.02998, so we're good.
 
         Args:
-            counts: A dictionary containing measurement counts from circuit execution.
+            counts: A dictionary containing the measurement counts from circuit execution.
 
         Returns:
-            The QAOA Vanilla proxy benchmark score.
+            The QAOA proxy benchmark score.
         """
         ideal_counts = supermarq.simulation.get_ideal_counts(self.circuit())
         total_shots = sum(counts.values())

supermarq-benchmarks/supermarq/qcvv/conftest.py

@@ -8,7 +8,7 @@
 
 
 @pytest.fixture(scope="session", autouse=True)
-def patch_tqdm() -> Generator[None, None, None]:
+def _patch_tqdm() -> Generator[None, None, None]:
     """Disable progress bars during tests."""
     with mock.patch.dict(os.environ, {"TQDM_DISABLE": "1"}):
         yield