Selene

Selene is a quantum computer emulation platform written primarily in Rust with a python frontend.

Selene is built with flexibility in mind. This includes:

• A plugin system for the addition of additional components including simulators, error models, quantum runtimes to be provided within Selene or as third party plugins
• Support for custom input formats and device APIs through the selene-core build system.

What's included

Out of the box, Selene provides first-class support for the HUGR ecosystem, including execution of Guppy programs in an emulation environment, making use of our open-source compiler. You can find many examples of guppy usage in our unit tests.

Selene provides a range of simulators, including:

• Statevector simulation using QuEST and the quest-sys crate.
• Stabilizer simulation using Stim
• Coinflip simulation with customisable bias
• Classical Replay, for running pre-recorded measurements without direct simulation
• Quantum Replay, for running pre-recorded measurements with postselection-based simulation

Error models that are currently provided include:

• An 'ideal' error model which adds no noise to simulations
• A depolarizing error model which adds noise to qubit initialisation, measurement, and single- and two-qubit gates

And we offer two example quantum runtimes, including:

• Simple, which executes the program as-is, without any modifications
• SoftRZ, which elides Z rotations through RXY gates, providing the same observable behaviour with fewer quantum operations

Usage example

Although examples are provided in our tests folder, here is a quick walkthrough to get you started with Selene, HUGR and Guppy.

• First, we define the guppy program that we're interested in emulating:

from guppylang import guppy
from guppylang.std.quantum import *
from hugr.qsystem.result import QsysShot, QsysResult

@guppy
def main() -> None:
    # allocate 10 qubits
    qubits = array(qubit() for _ in range(10))

    # prepare the 10-qubit GHZ state (|0000000000> + |1111111111>)/sqrt(2)
    h(qubits[0])
    for i in range(9):
        cx(qubits[i], qubits[i+1])

    # measure all qubits
    ms = measure_array(qubits)

    # report measurements to the results stream
    result("measurements", ms)

compiled_hugr = main.compile()

• Then we compile the resulting HUGR Envelope to LLVM IR or bitcode using the HUGR-QIS compiler

from selene_sim import build
runner = build(compiled_hugr)

• Then we can utilise run or run_shots on the resulting selene instance, choosing a simulator (in this case Quest or Stim) and an error model (in this case DepolarizingErrorModel) to run the program.

from selene_sim import Quest, Stim
# run a single shot with Quest, the statevector simulator
shot = QsysShot(runner.run(simulator=Quest(), n_qubits=10))
print(shot)

# run a single shot with Stim, the stabilizer simulator
shot = QsysShot(runner.run(simulator=Stim(), n_qubits=10))
print(shot)

# run_shots runs efficient multi-shot simulations
# n_processes provides multi-processing across shots
# deterministic results can be achieved by providing a random seed
shots = QsysShot(runner.run(
    simulator=Stim(random_seed=5),
    n_qubits=10,
    n_shots=100,
    n_processes=8
))
print(shots)

• As well as simulators, we can customise the emulation by providing an error model, such as the depolarizing error model:

from selene_sim import DepolarizingErrorModel
error_model = DepolarizingErrorModel(
    random_seed=12478918,
    p_init=1e-3,
    p_meas=1e-2,
    p_1q=1e-5,
    p_2q=1e-6,
)

shots = QsysResult(runner.run_shots(
    simulator=Stim(
        random_seed=10
    ), 
    error_model=error_model,
    n_qubits=10,
    n_shots=20,
    n_processes=4,
))
print(shots)

• And/or a runtime, such as the SoftRZRuntime, which elides physical RZ gates through subsequent RXY gates:

from selene_sim import SoftRZRuntime

shots = QsysResult(runner.run_shots(
    simulator=Stim(),
    runtime=SoftRZRuntime(),
    error_model=error_model,
    n_qubits=10,
    n_shots=20
))
print(shots)