AerSimulator seed should accept Numpy random Generator

[kevinsung]

What is the expected behavior?

This should work:

import numpy as np
from qiskit.circuit import QuantumCircuit
from qiskit_aer import AerSimulator

rng = np.random.default_rng()
sim = AerSimulator(seed_simulator=rng)
sim.run(QuantumCircuit()).result()

TypeError: invalid option type: name=seed_simulator, type=Generator, expected=int

[hhorii]

@kevinsung Could you tell the situation that application specifies a random number generator of numpy as a seed instead of int value? I know Qiskit takes a random number generator of numpy as a seed. Do you need to reuse the same instance?

[kevinsung]

Yes, exactly, if you already have a random Generator instance, it's best to use it everywhere to ensure that the results are as "random" as possible.

[chriseclectic]

@kevinsung The issue with this request is the seed option is only used in the C++ code which has no idea about numpy objects and can only handle int values to initialize the C++ rng generator. You would need to convert the Python object into an int for the C++ code to be able to use.

[kevinsung]

To work around this issue I've been passing seed_simulator=rng.integers(2**32). I'm not sure if this is totally valid, but if something like this is valid, then the AerSimulator can do it itself.