Running Jobs

A job is one call to MENOSBackend.run. This page covers the choices you make per call.

Choosing the output

Each call asks for some combination of outputs. At least one is required.

You want	Ask with
Measurement counts	a circuit with measure and shots > 0
The exact statevector	want_statevector=True (usually with shots=0)
Both	a measured circuit, shots > 0 and want_statevector=True
An expectation value	use MENOSEstimator instead

from qiskit import QuantumCircuit

from menos.qiskit import MENOSBackend

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

# Counts only.
counts_only = backend.run(qc, shots=100, seed_simulator=1).result()
assert "counts" in counts_only.data(0)

# Nothing requested: refused.
try:
    backend.run(qc, shots=0)
except Exception as error:
    print(type(error).__name__)
    assert type(error).__name__ == "InvalidCircuitError"

Asking for a statevector rules out some engines: only the dense engine can produce one, so the circuit must be small enough for a dense state.

Shots and seeds

shots is the number of samples. seed_simulator fixes the random stream:

from qiskit import QuantumCircuit

from menos.qiskit import MENOSBackend

backend = MENOSBackend()
qc = QuantumCircuit(1, 1)
qc.h(0)
qc.measure(0, 0)

a = backend.run(qc, shots=200, seed_simulator=9).result().get_counts()
b = backend.run(qc, shots=200, seed_simulator=9).result().get_counts()
assert a == b

The same circuit, shots, options and seed always give the same counts. See Determinism. If you do not pass a seed, a fresh one is drawn and recorded in the result, so you can reproduce a run afterwards.

The job object

run returns a MENOSJob. Because execution is synchronous it is already finished:

from qiskit import QuantumCircuit

from menos.qiskit import MENOSBackend

qc = QuantumCircuit(1, 1)
qc.x(0)
qc.measure(0, 0)
job = MENOSBackend().run(qc, shots=5, seed_simulator=2)
print(job.status(), job.job_id())
assert str(job.status()) == "JobStatus.DONE"
assert job.result().get_counts() == {"1": 5}

job.result(timeout=...) accepts a timeout for interface compatibility and ignores it.

Failures

A job that cannot run raises when you call run, not when you ask for the result, because the work happens inside run. The common cases:

Situation	Exception
Nothing requested	InvalidCircuitError
shots > 0 but no measure	InvalidCircuitError
Gate or feature MENOS cannot represent	UnsupportedFeatureError
No plan fits the memory budget	ResourceExceededError

See Common Errors.

Timing

result.time_taken is the time spent executing plus sampling, in seconds. It excludes planning and conversion.