qitesse 0.2.0

High-throughput CPU backend for repeated evaluation of parameterized quantum circuits

qitesse

qitesse is a high-throughput CPU backend for repeated evaluation of parameterized quantum circuits.

qitesse is built upon qitesse-sim, the high-performance CPU-based state-vector execution engine for quantum circuits, fully built in Rust.

This PyPI module provides a Python interface designed for hybrid quantum algorithms, repeated circuit evaluation, backend integration, and low-overhead CPU execution from Python.

Features

• Rust-based CPU statevector simulator with a Python API
• One-off circuit execution through Gate and Circuit
• Compiled parameterized circuits through CircuitSpec, Parameter, and CompiledCircuit
• Reusable zero-copy parameter buffers through ParamBuffer and ParamBatchBuffer
• Expectation-value workflows for Pauli observables and Hamiltonians
• Batch expectation execution for parameter sweeps and optimizer loops
• Gradient APIs for compiled circuits via parameter-shift evaluation
• Reusable ExecutionContext buffers for repeated scalar execution
• Full statevector output for both one-off and compiled execution paths
• Mid-circuit measure, reset, and barrier operations
• Custom unitary operations with Gate.unitary(...)
• Controlled custom unitaries with Gate.controlled_unitary(...)
• Common single-, two-, and multi-qubit gates
• Read the Docs documentation with autogenerated API reference pages

Installation

qitesse requires Python 3.8+. Install it via pip:

pip install qitesse

Or install from source:

git clone https://github.com/OsamaMIT/qitesse.git

pip install maturin

maturin develop --release

To run examples:

python examples/h_example.py

python examples/qft._example.py

python examples/custom_unitary.py

Documentation

The documentation is now set up for Read the Docs with automatic API generation.

• Entry point: docs/index.rst
• Current features: docs/current_features.rst
• Compiled execution guide: docs/guides/compiled_execution.rst
• Generated API reference root: docs/api/index.rst
• Read the Docs config: .readthedocs.yaml

To add a new public class to the API docs, add it once in docs/api/index.rst. Sphinx will generate the class page and include its methods and attributes automatically.

Current Capabilities

qitesse currently has two main usage modes:

1. General-purpose simulation with Gate and Circuit for one-off execution.
2. Compiled execution with CircuitSpec and CompiledCircuit for repeated parameterized workloads.

The compiled path currently supports:

• reusable zero-copy parameter buffers
• scalar expectation evaluation
• batched expectation evaluation
• scalar gradients
• batched gradients
• reusable execution contexts
• statevector inspection when needed

The simulator path currently supports:

• standard single-qubit gates
• controlled and multi-qubit gates
• custom unitaries
• measurement, reset, and barrier operations

Compiled Circuits

import numpy as np
import qitesse

spec = qitesse.CircuitSpec(2)
theta = spec.param("theta")

spec.ry(0, theta)
spec.cx(0, 1)

compiled = spec.compile()
observable = qitesse.Observable.pauli_z(1)

value = compiled.expectation(np.array([0.4], dtype=np.float32), observable)
gradient = compiled.gradient(np.array([0.4], dtype=np.float32), observable)
value_again, grad_again = compiled.value_and_gradient(
    np.array([0.4], dtype=np.float32),
    observable,
)
state = compiled.statevector(np.array([0.4], dtype=np.float32))

params_batch = np.array([[0.1], [0.2], [0.3]], dtype=np.float32)
values = compiled.batch_expectation(params_batch, observable)
grads = compiled.batch_gradient(params_batch, observable)

context = compiled.execution_context()
value_again = context.expectation(compiled, np.array([0.5], dtype=np.float32), observable)
grad_again = context.gradient(compiled, np.array([0.5], dtype=np.float32), observable)

This execution path is intended for repeated evaluation of the same circuit structure with different parameter values.

Backend Integration

The compiled execution path is the primary integration surface for higher-level libraries.

The intended backend pattern is:

1. Translate a framework circuit into CircuitSpec
2. Compile once per circuit structure
3. Keep parameters in contiguous numpy.float32 arrays
4. Call expectation, batch_expectation, gradient, or value_and_gradient

For sequential scalar calls, reuse compiled.execution_context() to keep internal buffers alive.

For sweeps, minibatches, or optimizer batches, prefer batch_expectation(...) and batch_gradient(...) instead of looping in Python.

Supported Gates

Single-qubit gates:

• i
• x
• y
• z
• h
• s
• sdg
• t
• tdg
• rx
• ry
• rz
• p / phase
• u

Two-qubit gates:

• cnot / cx
• cy
• cz
• ch
• swap
• iswap
• crx
• cry
• crz
• cp / cphase
• cu

Three-qubit and larger:

• ccx / toffoli
• cswap / fredkin
• mcx
• mcz
• mcp / mcphase
• controlled_unitary

Circuit operations:

• measure
• reset
• barrier

Custom unitaries:

import numpy as np
import qitesse

hadamard = np.array([[1, 1], [1, -1]], dtype=np.complex64) / np.sqrt(2)

circuit = qitesse.Circuit([
    qitesse.Gate.unitary([0], hadamard),
    qitesse.Gate.controlled_unitary([0], [1], hadamard),
])

state = circuit.run(2)

Use run_with_measurements(num_qubits) if the circuit contains measurement gates and you want the observed bit values back.

Planned Features

• Additional simulation backends

Contributing

Contributions are welcome! To contribute:

1. Fork the repository
2. Create a new branch (feature-branch)
3. Commit your changes and open a pull request

License

This project is licensed under the MIT License. See the LICENSE file for details.