qiskit_noise_learning.aer_executor.AerExecutor

class qiskit_noise_learning.aer_executor.AerExecutor(qasm_simulator: AerSimulator, noise_dict: dict[str, PauliLindbladMap] | None = None, angle_decimals: int = 5, warn_absent: bool = True, root_seed: int | None = None)

Bases: object

Local Aer-based executor mimicking the IBM Runtime executor interface.

Runs a QuantumProgram eagerly on construction of the returned job — the result is available immediately when AerRuntimeJob.result() is called.

Noise injection

When noise_dict is provided, Pauli-Lindblad noise is injected into circuits at tagged barriers via InsertNoisePass. Samplomatic inserts three barriers around each boxed gate — left (L), middle (M), and right (R) — with labels of the form <pos><idx>@tag=<tag> (e.g. R0@tag=r0). By default, noise is injected at the R (right) barriers, i.e. after the gate. Use noise_after=False on InsertNoisePass to target M barriers instead (noise before the gate).

The noise_dict format is:

• Keys — layer name tags (strings, e.g. "r0", "my_tag"). Each key must match the ref of a Tag annotation used when building the QuantumProgram. A warning is emitted (if warn_absent=True) when a tagged barrier’s tag is absent from the dict; the barrier is left as-is (no noise inserted for that layer).
• Values — PauliLindbladMap instances describing the Pauli-Lindblad noise channel for that gate. The map’s num_qubits must equal the number of qubits on the corresponding barrier in the circuit.
• Qubit indexing — indices inside the map are local to the barrier’s qubit set, independent of global circuit qubit numbering. Local index i refers to the i-th qubit of the barrier in ascending physical-qubit order, so a device-wide map can be converted with PauliLindbladMap.keep_qubits(sorted(qubits)).

Parameters

• qasm_simulator – The Aer simulator to run programs on.
• noise_dict – A map from barrier label refs to Pauli-Lindblad noise maps. Pass None (default) to run without noise injection.
• angle_decimals – Gate angles are rounded to the nearest multiple of π/2 at this decimal precision before simulation. This prevents floating-point drift from preventing Clifford-method simulation when angles are nominally Clifford.
• warn_absent – If True (default), emit a warning when a tagged barrier’s tag is not found in noise_dict. Set to False when partial coverage of tags is intentional.
• root_seed – Root seed for random number generation, covering both the sampling of shots and the sampling of twirls. Rather than being used directly, it seeds a sequence that each call to run() draws the next seed from, so that runs are independently random. With the default of None a root seed is drawn nondeterministically; it is available as root_seed either way.

__init__

Methods

Attributes

root_seed

run