QAOARuntime with CVAR optimization

[ghellstern]

Environment

Qiskit Software	Version
qiskit-terra	0.24.0
qiskit-aer	0.12.0
qiskit-ignis	0.6.0
qiskit-ibmq-provider	0.20.2
qiskit	0.43.0
qiskit-nature	0.3.1
qiskit-finance	0.2.0
qiskit-optimization	0.5.0
qiskit-machine-learning	0.2.1

Python version | 3.7.9
Python compiler | GCC 7.3.0
Python build | default, Aug 31 2020 12:42:55
OS | Linux
CPUs | 4
Memory (Gb) | 15.650867462158203

What is happening?

In the tutorial https://qiskit.org/ecosystem/optimization/tutorials/12_qaoa_runtime.html QAOAClient produces an error when alpha=0.75 ist used. When removing alpha everythin is fine. See attached notebook.

QAOA_client.zip

How can we reproduce the issue?

See Notebook

What should happen?

QAOAClient should work with alpha<1.0

Any suggestions?

No response

[Cryoris]

I can reproduce the above and looking at the error codes it seems the CVaRMeasurement is given an unsupported input:

[...]
2023-05-22T12:47:42.644038453Z     raise ValueError("Unsupported input to CVaRMeasurement.eval:", type(front))
2023-05-22T12:47:42.644047641Z ValueError: ('Unsupported input to CVaRMeasurement.eval:', )

Edit: Locally using the CVaRExpectation with the old QAOA code works as expected, this seems to be an issue of the runtime code only.

[ghellstern]

Maybe I should add the discussion on the Slack channel with Steve Wood:

https://qiskit.slack.com/archives/CB6C24TPB/p1684506534062659

[t-imamichi]

QAOARuntime and QAOAClient have been removed #542 because they are replaced with primitives and session.