FiQCI EMS

FiQCI Error Mitigation Service (EMS) is a Python library for quantum error mitigation as part of the Finnish Quantum Computing Infrastructure (FiQCI). It wraps IQM quantum backends and applies error mitigation transparently, allowing users to run circuits with improved accuracy by specifying a mitigation level.

This python package can be pre-installed on a HPC system or installed by the user. The main goal of the project is to allow users using FiQCI quantum computers to easily add flags to run error mitigated quantum jobs.

Mitigation Levels

Sampler

Level	Mitigation Applied	Technique
0	None	Raw results
1	Readout Error Mitigation	M3 (matrix-free measurement mitigation)
2	Level 1 + Dynamical Decoupling	Dynamical Decoupling
3	Level 2 + Pauli Twirling	Pauli Twirling

Estimator

Level	Mitigation Applied	Technique
0	None	Raw results
1	Readout Error Mitigation	M3 (matrix-free measurement mitigation)
2	Level 1 + Dynamical Decoupling	Dynamical Decoupling
3	Level 2 + Zero Noise Extrapolation	Exponential Extrapolation, Local Folding

Note

FiQCIBackend mitigation levels correspond to the Sampler levels.

The default is level 1, which applies M3 readout error mitigation.

Installation

UV is recommended for installation

uv pip install fiqci-ems
#or
uv add fiqci-ems

Requires Python 3.11 or 3.12.

The example notebooks also need matplotlib, which Qiskit ships only as an optional extra:

uv pip install matplotlib
#or
uv add matplotlib

Usage

Start by initialising your IQM backend and a quantum circuit.

from iqm.qiskit_iqm import IQMProvider
from qiskit import QuantumCircuit, transpile

# Initialise backend
provider = IQMProvider()
backend = provider.get_backend()

# Define a quantum circuit
qc = QuantumCircuit(2)
qc.h(0)
qc.cx(0, 1)
qc.measure_all()

# Transpile the circuit
qc_transpiled = transpile(qc, backend=backend, initial_layout=qubit_indices)

EMS provides three interfaces depending on your use case.

FiQCISampler - sampling interface

For users who need measurement counts with built-in mitigation:

from fiqci.ems import FiQCISampler

# Using mitigation_level
sampler = FiQCISampler(backend, mitigation_level=1)

# Execute the job
job = sampler.run(qc_transpiled, shots=2048)

# Get results
result = job.result()

# Or manually set mitigation options
sampler.rem(enabled=True, calibration_shots=2000, calibration_file="cals.json")

# See applied and available options
sampler.mitigation_options

FiQCIEstimator - expectation values

Computes expectation values of Pauli observables directly from circuits:

from fiqci.ems import FiQCIEstimator
from iqm.qiskit_iqm import transpile_to_IQM
from qiskit.quantum_info import SparsePauliOp

# Using mitigation_level
estimator = FiQCIEstimator(backend, mitigation_level=1)

# Transpile for the backend. remove_final_rzs=False is required: the estimator measures in the X and
# Y bases, and dropping the final RZ gates changes those expectation values.
tr_qc = transpile_to_IQM(qc, backend, remove_final_rzs=False, optimization_level=3)

# Define observables
observables = SparsePauliOp.from_list([("ZZ", 1), ("IX", 1)])

# Map observables to transpiled layout
device_observables = observables.apply_layout(tr_qc.layout)

# Execute the job
job_collection = estimator.run(tr_qc, observables=device_observables, shots=2048)

# Get expectation values
evs = job_collection.expectation_values()

# Access all jobs executed by estimator
jobs = job_collection.jobs()

# Or manually set mitigation options
estimator.rem(enabled=True, calibration_shots=2000, calibration_file="cals.json")

# See applied and available options
estimator.mitigation_options

FiQCIBackend - drop-in backend replacement

FiQCIBackend is used under the hood by both sampler and estimator. Wraps any IQM backend and applies error mitigation to run() calls:

from fiqci.ems import FiQCIBackend

# Using mitigation_level
backend = FiQCIBackend(backend, mitigation_level=1)

# Execute the job
job = backend.run(circuit, shots=1024)

# Get the results
result = job.result()

# Or manually set mitigation options
backend.rem(enabled=True, calibration_shots=2000, calibration_file="cals.json")

# See applied and available options
backend.mitigation_options

Access raw (pre-mitigation) counts via backend.raw_counts (populated only after the run's result() has been retrieved, since post-processing is computed lazily).

run() returns a lazy job handle immediately without waiting for results. The per-batch job_id()s and an aggregated status()/done() are available right away; error mitigation and result combination are computed on the first result() call. The handle also exposes job_ids() and partial_results() (per-batch results for batches that have already completed), and result() raises BatchFailedError identifying the failing batch if any batch fails.

If the backend rejects a circuit during submission, run() does not throw: it logs a warning, stops submitting, and returns a handle covering every batch. Submitted batches keep their job ids and status, the rejected batch reports ERROR, and skipped batches report CANCELLED (with a None job id). Inspect the outcome via statuses()/status()/partial_results().

Advanced Usage

It is also possible to manually configure and directly use the M3 mitigator without the wrapper classes above. Consult the docs for how this is done.

Configuration

All three interfaces accept the same core options:

Parameter	Default	Description
mitigation_level	1	Mitigation level (0-3)
calibration_shots	1000	Shots used for M3 calibration circuits
calibration_file	None	Path to save/load calibration data (JSON)

Mitigation can also be configured directly. See the docs for FiQCISampler, FiQCIEstimator, and FiQCIBackend to see all available options.

Documentation

Full documentation including API reference, guides, and Jupyter notebook examples is available at docs

Development

# Install with dev dependencies
uv sync

# Run tests
uv run pytest

# Run tests with coverage
uv run pytest --cov

# Lint and format
uv run ruff check --fix
uv run ruff format

# Type check
uv run pyrefly check

Building docs

#Install docs dependencies
uv sync --group docs

#Build docs
uv run sphinx-build docs/ docs/_build

License

Apache 2.0, see LICENSE for details.

Having trouble?

Contact servicedesk@csc.fi or raise an issue here.