CTMP mitigator crashes Jupyter notebook kernel

[nonhermitian]

Informations

• Qiskit Ignis version: latest
• Python version: 3.7
• Operating system: Linux

What is the current behavior?

The CTMP mitigator crashes a Jupyter notebook kernel on the following circuit:

qubits=[14, 13, 11, 12, 16, 8, 15, 5, 18, 19, 3, 2, 22, 21, 25, 1]

qc = QuantumCircuit(27, len(qubits))
qc.h(14)
qc.cx(14, 13)
qc.cx(14, 11)
qc.cx(13, 12)
qc.cx(14, 16)
qc.cx(11, 8)
qc.cx(12, 15)
qc.cx(8, 5)
qc.cx(15, 18)
qc.cx(16, 19)
qc.cx(5, 3)
qc.cx(19, 22)
qc.cx(18, 21)
qc.cx(3, 2)
qc.cx(22, 25)
qc.cx(2, 1)
qc.measure(qubits, range(len(qubits)))

when targeting a 27Q system. Performing the mitigation using:

circuits2, metadata2 = expval_meas_mitigator_circuits(len(qubits), method='ctmp')
job_cal2 = execute(circuits2, backend, shots=8192, initial_layout=qubits)
result_cal2 = job_cal2.result()
ctmp_mitigator = ExpvalMeasMitigatorFitter(result_cal2, metadata2).fit()
ctmp_val = expectation_value(counts_noise, meas_mitigator=ctmp_mitigator)

crashes on the last line. Note that method='tensored' works correctly.

Steps to reproduce the problem

Try the above.

What is the expected behavior?

Does not crash.

Suggested solutions

There is no error given save for the kernel crashing modal window.

[nonhermitian]

A bit more work uncovered that the Killed return value is related to SIGKILL and that it is probably due to excessive memory consumption. Indeed, looking at the memory while running, it uses all 32Gb of memory and all my swap space before terminating.

[chriseclectic]

@nonhermitian Can you check if this is fixed by #563?

[nonhermitian]

Yep, that does it! thanks.