Aer sampler does not truncate to active number of qubits

[nonhermitian]

Informations

• Qiskit Aer version: latest
• Python version:
• Operating system:

What is the current behavior?

For many years now, the Aer simulator has truncated a circuit to only the active qubits, allowing for simulating transpiled circuits against device noise models; even when the full device number of qubits is larger than Aer supports (but the circuits active width is small enough for Aer to handle). However, the Aer Sampler does not perform the same truncation. E.g. using 5 qubits, and transpiling to an Eagle, the old Aer simulator will still truncate and run the circuit.

sim = AerSimulator()
job1 = sim.run(trans_qc, shots=int(1e4), dynamic=True)

works fine. But doing the same thing via the Aer sampler raises:

from qiskit_aer.primitives import Sampler as AerSampler
sampler = AerSampler()
job1 = sampler.run(trans_qc, shots=int(1e4), dynamic=True)

CircuitTooWideForTarget: 'Number of qubits (127) in circuit-8195 is greater than maximum (30) in the coupling_map'

Steps to reproduce the problem

Run a transpiled circuit that is smaller than 30Q in active qubits, but transpiled against a 127Q system

What is the expected behavior?

The sampler should truncate to active qubits only, just like the simulator does.

Suggested solutions

[zlatko-minev]

+1

[doichanj]

Could you provide full script to reproduce the problem?

[nonhermitian]

from qiskit import *
from qiskit_aer.primitives import Sampler
sampler = Sampler()

qc = QuantumCircuit(100, 2)
qc.h(98)
qc.cx(98, 99)
qc.measure([98, 99], [0, 1])
sampler.run(qc).result()

[doichanj]

This issue is fixed by release 0.14.1

[nonhermitian]

It is not fixed in Aer 0.14.1:

[doichanj]

I'm sorry but I had implemented wrongly this one:

qiskit-aer/qiskit_aer/backends/aerbackend.py

Line 355 in 7b93719

if self._target is not None:

I fixed this

[haimeng-zhang]

The issue does not seem to be fixed when one of the following simulation methods is specified:
statevector, density_matrix, extended_stabilizer, matrix_product_state, unitary, superop.
For example, when using the method matrix_product_state, I got the following error

The script to reproduce this error is the following:

from qiskit import *
from qiskit_aer.primitives import Sampler
shots = 5000
options = {'shots':shots, 'method':'matrix_product_state'}
sampler = Sampler(backend_options=options)

qc = QuantumCircuit(100, 2)
qc.h(98)
qc.cx(98, 99)
qc.measure([98, 99], [0, 1])
sampler.run(qc).result()

I am using Aer version '0.14.2'.

[hhorii]

I confirmed that 0.15 can run

from qiskit import *
from qiskit_aer.primitives import Sampler
sampler = Sampler()

qc = QuantumCircuit(100, 2)
qc.h(98)
qc.cx(98, 99)
qc.measure([98, 99], [0, 1])
sampler.run(qc).result()

and

from qiskit import *
from qiskit_aer.primitives import Sampler
shots = 5000
options = {'shots':shots, 'method':'matrix_product_state'}
sampler = Sampler(backend_options=options)

qc = QuantumCircuit(100, 2)
qc.h(98)
qc.cx(98, 99)
qc.measure([98, 99], [0, 1])
sampler.run(qc).result()

, which were reported in this issue.