marginal_counts fails on qiskit Result structure

[eendebakpt]

Environment

• Qiskit Terra version: 0.24.1
• Python version: 3.11
• Operating system: Windows

What is happening?

The marginalize_counts method fails on the output of the Aer simulator (a qiskit.result.result.Result structure). A minimal example is included below.

The error generated is:

Traceback (most recent call last):

  Cell In[7], line 12
    result = marginal_counts(result, [0])

  File C:\develop\env311\Lib\site-packages\qiskit\result\utils.py:67 in marginal_counts
    result = deepcopy(result)

...

  File C:\develop\Python311\Lib\copy.py:231 in _deepcopy_dict
    y[deepcopy(key, memo)] = deepcopy(value, memo)

  File C:\develop\Python311\Lib\copy.py:161 in deepcopy
    rv = reductor(4)

TypeError: cannot pickle 'qiskit_aer.backends.controller_wrappers.AerCircuit' object

How can we reproduce the issue?

from qiskit import QuantumCircuit
from qiskit.result.utils import marginal_counts
from qiskit_aer import Aer

qc = QuantumCircuit(2, 2)
qc.h(0)
qc.h(1)
qc.measure([0, 1], [0, 1])

backend = Aer.get_backend("aer_simulator")
result = backend.run(qc, shots=100, memory=True).result()
result = marginal_counts(result, [0])

What should happen?

This code worked in previous version of qiskit-terra. The marginalize_counts should work without error on results generated by qiskit.

Any suggestions?

No response

[Abhiraj-Shrotriya]

Calling marginalize_counts on result would not be correct since the result object holds a lot of things (for the code given by you) -

result = Result(backend_name='qasm_simulator', backend_version='0.12.1', qobj_id='', job_id='---------', success=True, results=[ExperimentResult(shots=1000, success=True, meas_level=2, data=ExperimentResultData(counts={'0x1': 255, '0x3': 239, '0x2': 267, '0x0': 239}), header=QobjExperimentHeader(creg_sizes=[['c', 2]], global_phase=0.0, memory_slots=2, metadata={}, n_qubits=2, name='circuit-239', qreg_sizes=[['q', 2]]), status=DONE, circuit=Circuit(qubit=2, num_memory=2, num_registers=0, ops={h[0],[],h[1],[],measure[0],[],measure[1],[]}, shots=100, seed=2825605101, global_phase_angle=0), seed_simulator=2825605101, metadata={'batched_shots_optimization': False, 'method': 'stabilizer', 'active_input_qubits': [0, 1], 'device': 'CPU', 'remapped_qubits': False, 'num_qubits': 2, 'num_clbits': 2, 'sample_measure_time': 0.004132, 'input_qubit_map': [[0, 0], [1, 1]], 'measure_sampling': True, 'noise': 'ideal', 'parallel_shots': 1, 'parallel_state_update': 16, 'fusion': {'enabled': False}}, time_taken=0.0060285)], date=----------, status=COMPLETED, header=None, metadata={'parallel_experiments': 1, 'omp_enabled': True, 'max_memory_mb': 16333, 'max_gpu_memory_mb': 0, 'num_processes_per_experiments': 1, 'mpi_rank': 0, 'num_mpi_processes': 1, 'time_taken_execute': 0.0061044}, time_taken=0.00600433349609375)

Your problem should be fixed if you call marginalize_counts on result.get_counts()

[eendebakpt]

@Abhiraj-Shrotriya Using marginal_counts(result.get_counts()) indeed works. But this results in a dictionary, instead of a Result structure.

The signature for marginalize_counts is:

marginal_counts(
    result: Union[dict, qiskit.result.result.Result],
    indices: Optional[List[int]] = None,
    ...

So it seems reasonable to expect the marginal_counts to work in Results input structures.

[jakelishman]

This was caused by a bug (of sorts) in Aer (see Qiskit/qiskit-aer#1845). If you're able to use Aer 0.11 temporarily until Aer 0.12.2 can be released, hopefully in the next week or two, that should work around the problem, I think.

[jakelishman]

Oh, Aer 0.12.0 might be fine too - I forget exactly which version the issue appeared in.

[eendebakpt]

The issue will be fixed in the next release of qiskit aer (confirmed by running from git with commit 6802355b7f3f721098d62255ea994ce6abda5ed3), closing this issue.