Port BackendSampler and BackendEstimator from Qiskit 1.1

[t-imamichi]

Summary

(Update May 22)

If #1690 is merged, we don't need this PR anymore. We can remove all files related to backend primitives.

---

Port Qiskit/qiskit#12291
I copied SamplerPubResult so that Qiskit 1.0 works.
If we will update the requirement with qiskit>=1.1 after Qiskit 1.1 release, we don't have to keep them.
We just need backend_sampler_v2.py and backend_estimator_v2.py.

Fixes #1631

If #1690

Details and comments

microbenchmarking

from timeit import timeit

from qiskit_aer import AerSimulator
from qiskit_ibm_runtime.fake_provider import FakeSherbrooke

from qiskit import QuantumCircuit
from qiskit.transpiler.preset_passmanagers import generate_preset_pass_manager

from qiskit_ibm_runtime.qiskit.primitives import BackendSamplerV2, BackendEstimatorV2

backend = AerSimulator.from_backend(FakeSherbrooke())
shots = 10000
precision = 0.01
num_copies = 10


def gen_circuit(num_qubits: int, reps: int):
    qc = QuantumCircuit(num_qubits)
    for _ in range(reps):
        qc.h(range(num_qubits))
        for i in range(0, num_qubits - 1, 2):
            qc.cx(i, i + 1)
    qc.measure_all()
    return qc


def bench_sampler_v2(qc: QuantumCircuit):
    print("\nBackendSamplerV2")
    sampler = BackendSamplerV2(backend=backend)
    print(f"{timeit(lambda: sampler.run([qc] * num_copies, shots=shots).result(), number=1)} sec")


def bench_estimator_v2(qc: QuantumCircuit):
    print("\nBackendEstimatorV2")
    estimator = BackendEstimatorV2(backend=backend)
    op = "Z" + "I" * (qc.num_qubits - 1)
    print(
        f"{timeit(lambda: estimator.run([(qc, op)] * num_copies, precision=precision).result(), number=1)} sec"
    )


qc = gen_circuit(5, 5)
pm = generate_preset_pass_manager(optimization_level=2, backend=backend)
qc2 = pm.run(qc)
bench_sampler_v2(qc2)
bench_estimator_v2(qc2)

main branch

BackendSamplerV2
3.3306877909999457 sec

BackendEstimatorV2
3.359602416996495 sec

this PR

BackendSamplerV2
0.45033658400643617 sec

BackendEstimatorV2
0.4832707500027027 sec

[coveralls]

Pull Request Test Coverage Report for Build 9089237108

Details

• 109 of 130 (83.85%) changed or added relevant lines in 3 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage decreased (-0.06%) to 82.707%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
qiskit_ibm_runtime/qiskit/primitives/sampler_pub_result.py	9	30	30.0%

Totals
Change from base Build 9084268669:	-0.06%
Covered Lines:	6313
Relevant Lines:	7633

---

💛 - Coveralls

[ihincks]

Thanks @t-imamichi the improvement to performance will be nice.

We just need backend_sampler_v2.py and backend_estimator_v2.py.

I'm a bit confused. If we wait for qiskit 1.1, can't we use BackendSamplerV2 and BackendEstimatorV2 from there, and not have any such implementations in this repo?

[t-imamichi]

I think BackendSamplerV2 and BackendEstimatorV2 were copied from qiskit to qiskit-ibm-runtime because they are not in qiskit 1.0 release yet (but they will be part of qiskit 1.1 release).
They are used here.

qiskit-ibm-runtime/qiskit_ibm_runtime/fake_provider/local_service.py

Line 30 in 58722ee

from ..qiskit.primitives import BackendEstimatorV2, BackendSamplerV2 # type: ignore[attr-defined]

So, if we can update the requirement qiskit>=1.1.0, we then can remove them from qiskit-ibm-runtime.
If we need to keep qiskit>=1.0.0, we need this PR to fix the performance issue of the local mode.

qiskit-ibm-runtime/requirements.txt

Line 10 in 58722ee

qiskit>=1.0.0

[ihincks]

Thanks @t-imamichi. Let's then merge this PR and wait for #1690.

[ihincks]

I'm not confident that including this is worth it: it may cause isinstance confusion when people have one in qiskit and here.

However, this is not a blocker for me.

[t-imamichi]

It depends on how we control the compatibility with Qiskit 1.0.
If we still allow users to use Qiskit 1.0.*, we need SamplerPubResult here. Yes, it would be confusing to Qiskit 1.1 users.
But if #1690 is merged, we can simply remove all backend primitives as well as SamplerPubResult.
I'm curious about the status of #1690

[ihincks]

I hope that #1690 can be done this sprint, we'll see what happens in sprint planning, and that we can cut a new qiskit_ibm_runtime release with the change. What do you think, @kt474 ?

[kt474]

Yeah I can help @t-imamichi with that issue this sprint (based on what needs to be split out of #1700). Then we can cut a new release.

[t-imamichi]

Superseded by #1700