EstimatorV2 extended_stabilizer fails" invalid instructions {"instructions": {save_expval}}

[zlatko-minev]

Informations

• Qiskit Aer version: 0.14.1
• Python version: 3.11
• Operating system: Mac

What is the current behavior?

Can't run extended stabilizer with Estimator, see code:

Steps to reproduce the problem

from qiskit_aer.primitives import EstimatorV2 
qc = QuantumCircuit(2)
qc.sx(0)
qc.cx(0,1)
obs = ['ZI']

options = {
    "backend_options": {  
        "method": "extended_stabilizer",
    }
}
estimator = EstimatorV2(options=options)
print(estimator._backend)
job = estimator.run([(qc, obs)])

Return

AerSimulator('aer_simulator_extended_stabilizer')
Simulation failed and returned the following error message:
ERROR:  [Experiment 0] Circuit circuit-394 contains invalid instructions {"instructions": {save_expval}} for "extended_stabilizer" method.Circuit circuit-394 contains invalid parameters  for "extended_stabilizer" method.

What is the expected behavior?

Should run extended stabilizer with Estimator.

Suggested solutions

NA

[kevinsung]

So strange. When I run the code in @zlatko-minev 's repro, I get a different error. I'm using the same Aer version and Python 0.14.1, but my OS is Fedora Linux 40.

Traceback (most recent call last):
  File "/home/kevinsung/projects/documentation/test.py", line 13, in <module>
    estimator = EstimatorV2(options=options)
                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/kevinsung/.pyenv/versions/documentation/lib/python3.11/site-packages/qiskit_aer/primitives/estimator_v2.py", line 74, in __init__
    self._backend = AerSimulator(method=method, **self.options.backend_options)
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
TypeError: qiskit_aer.backends.aer_simulator.AerSimulator() got multiple values for keyword argument 'method'

[doichanj]

I found save_expval is disabled because there are some issue in implementation of save_expval of extended stabilizer.

qiskit-aer/src/simulators/extended_stabilizer/extended_stabilizer_state.hpp

Lines 401 to 405 in 83d679e

	// Disabled until can fix bug in expval
	// case Operations::OpType::save_expval:
	// case Operations::OpType::save_expval_var:
	// apply_save_expval(op, result, rng);
	// break;

[doichanj]

So strange. When I run the code in @zlatko-minev 's repro, I get a different error. I'm using the same Aer version and Python 0.14.1, but my OS is Fedora Linux 40.

Traceback (most recent call last):
  File "/home/kevinsung/projects/documentation/test.py", line 13, in <module>
    estimator = EstimatorV2(options=options)
                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/kevinsung/.pyenv/versions/documentation/lib/python3.11/site-packages/qiskit_aer/primitives/estimator_v2.py", line 74, in __init__
    self._backend = AerSimulator(method=method, **self.options.backend_options)
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
TypeError: qiskit_aer.backends.aer_simulator.AerSimulator() got multiple values for keyword argument 'method'

This is fixed by PR #2120

[zlatko-minev]

Thank you Sent from my iPhoneOn May 27, 2024, at 5:25 AM, Jun Doi ***@***.***> wrote: So strange. When I run the code in @zlatko-minev 's repro, I get a different error. I'm using the same Aer version and Python 0.14.1, but my OS is Fedora Linux 40. Traceback (most recent call last): File "/home/kevinsung/projects/documentation/test.py", line 13, in <module> estimator = EstimatorV2(options=options) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/home/kevinsung/.pyenv/versions/documentation/lib/python3.11/site-packages/qiskit_aer/primitives/estimator_v2.py", line 74, in __init__ self._backend = AerSimulator(method=method, **self.options.backend_options) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ TypeError: qiskit_aer.backends.aer_simulator.AerSimulator() got multiple values for keyword argument 'method' This is fixed by PR #2120 —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you were mentioned.Message ID: ***@***.***>

[doichanj]

I'm feeling difficulty to solve expval bug in extended stabilizer, so I think I will not include fix in Aer 014.2

[zlatko-minev]

Thank you for trying. Can you add a little more on what the issue is? That's tricky to resolve.

I hope somebody else here might be able to contribute as well.

The thing is that we're increasingly moving to using the estimator class and so this is now the default preferred pathway.

[doichanj]

I opened PR #2161 for this issue.
I'm fixing bugs in save_expval and now it works for clifford circuits
I think it is not easy because of accuracy for non-clifford circuits

[zlatko-minev]

Thanks! What do you mean by accuracy for non Clifford’s? Just the number of t gates and cost associated in memory with that Sent from my iPhoneOn Jun 3, 2024, at 5:57 AM, Jun Doi ***@***.***> wrote: I opened PR #2161 for this issue. I'm fixing bugs in save_expval and now it works for clifford circuits I think it is not easy because of accuracy for non-clifford circuits —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you were mentioned.Message ID: ***@***.***>

[doichanj]

It seems the estimation is unstable for non clifford circuits that causes wrong expectation values

[zlatko-minev]

You mean as a function of the truncation?And if a user increases the truncation it converges.  On Jun 3, 2024, at 10:31 PM, Jun Doi ***@***.***> wrote: It seems the estimation is unstable for non clifford circuits that causes wrong expectation values —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you were mentioned.Message ID: ***@***.***>