[BUG] qml.QNSPSAOptimizer includes op.adjoint instead of qml.adjoint(op) in source

[ikurecic]

Expected behavior

Circuit that includes qml.RandomLayers and uses qml.QNSPSAOptimizer executes normally.

Actual behavior

Execution raises AdjointUndefinedError. Other optimizers execute fine.

Additional information

qml.QNSPSAOptimizer includes op.adjoint in the source instead of qml.adjoint(op), as spotted by @albi3ro

Bug discovered by Lucas Tecot on the PennyLane Discussion Forum.

Source code

import pennylane as qml
from pennylane import numpy as plnp

num_qubits = 2
dev = qml.device("default.qubit", wires=num_qubits)
@qml.qnode(dev)
def cost(params):
    qml.RandomLayers(weights=params, wires=range(num_qubits), seed=42)
    return qml.expval(qml.PauliZ(0) @ qml.PauliZ(1))

params = plnp.random.normal(0, plnp.pi, (2, 4))
opt = qml.QNSPSAOptimizer(stepsize=5e-2)
for i in range(51):
    params, loss = opt.step_and_cost(cost, params)
    if i % 10 == 0:
        print(f"Step {i}: cost = {loss:.4f}")

Tracebacks

No response

System information

(Not relevant here.)

Existing GitHub issues

• I have searched existing GitHub issues to make sure the issue does not already exist.