Sample with Cirq

examples/cirq_sampler.py

The problem

Run existing Cirq code through cirq.work.Sampler. QubitraSampler is a cirq.work.Sampler, so code written against Cirq’s sampler interface runs on the platform by swapping the sampler object and nothing else. The example samples a Bell pair, then sweeps a rotation over five points as one job.

Needs the extra: pip install 'qubitra-sdk[cirq]'.

The walkthrough

The sampler

import cirq
import sympy

from qubitra import QubitraClient
from qubitra.cirq import QubitraSampler

    with QubitraClient() as client:
        sampler = QubitraSampler(client, backend_id=backend_id)
        q0, q1 = cirq.LineQubit.range(2)

QubitraSampler wraps a client and a backend id. Everything after this line is ordinary Cirq.

A Bell pair through run

        # -- a Bell pair, through cirq's single-run entry point -------------------
        bell = cirq.Circuit([cirq.H(q0), cirq.CNOT(q0, q1), cirq.measure(q0, q1, key="m")])
        print(f"running a Bell pair on {backend_id} ({SHOTS} shots)")
        result = sampler.run(bell, repetitions=SHOTS)

        histogram = Counter("".join(str(bit) for bit in row) for row in result.measurements["m"])
        print()
        for bits, count in sorted(histogram.items()):
            bar = "█" * round(40 * count / SHOTS)
            print(f"  {bits}  {count:5d}  {bar}")

sampler.run returns a cirq.Result whose measurements["m"] is one row per shot. The example folds the rows into a histogram; an even split across 00 and 11 says the entangling really happened.

A sweep, resolved before export

        # -- a sweep: every point in ONE job --------------------------------------
        theta = sympy.Symbol("theta")
        rotation = cirq.Circuit([cirq.ry(theta)(q0), cirq.measure(q0, key="m")])
        sweep = cirq.Linspace("theta", start=0.0, stop=math.pi, length=SWEEP_POINTS)

        print(f"\nsweeping theta over {SWEEP_POINTS} points, as one multi-PUB job")
        results = sampler.run_sweep(rotation, params=sweep, repetitions=SHOTS)

The parameter is a sympy.Symbol and the sweep a cirq.Linspace, Cirq’s own sweep vocabulary. Cirq exports OpenQASM 2, which carries no parameters, so the sampler resolves each point locally and submits the whole sweep as one multi-PUB job — five points is one round trip.

Read each point

        print()
        for point in results:
            angle = float(point.params.value_of("theta"))
            p_one = float(point.measurements["m"].mean())
            expected = math.sin(angle / 2) ** 2
            bar = "█" * round(20 * p_one)
            print(f"  theta={angle:5.3f}  P(1)={p_one:.3f}  (sin²={expected:.3f}) {bar}")
        return 0

run_sweep returns one result per point, each carrying its parameter binding. Rotating by theta about Y walks P(1) along sin²(θ/2), a curve you can check by eye.

Running it

QUBITRA_API_KEY=qpk_... python examples/cirq_sampler.py

Output from a live run:

running a Bell pair on sim-statevector-26q (1024 shots)

  00    510  ████████████████████
  11    514  ████████████████████

sweeping theta over 5 points, as one multi-PUB job

  theta=0.000  P(1)=0.000  (sin²=0.000)
  theta=0.785  P(1)=0.171  (sin²=0.146) ███
  theta=1.571  P(1)=0.490  (sin²=0.500) ██████████
  theta=2.356  P(1)=0.850  (sin²=0.854) █████████████████
  theta=3.142  P(1)=1.000  (sin²=1.000) ████████████████████

The Bell counts split evenly, and the five sampled probabilities track sin²(θ/2) to within shot noise.

Where to go next

• Cirq — the full adapter: export, bit order, and sweep batching.
• Primitives and PUBs — the multi-PUB job the sweep becomes.