Qubit-Discovery

Introduction | Usage | Installation | Contribution

Introduction

Qubit-Discovery is a set of software tools to perform gradient-based optimization of superconducting circuits using the SQcircuit Python package, with a special focus on qubit design. It provides a Python package, qubit-discovery, implementing loss functions and optimization algorithms, and a set of scripts using it in a reproducible workflow for parallelized optimization runs.

With these tools, superconducting circuits can be optimized over the values of their elements for metrics such as decoherence time, anharmonicity, qubit frequency, or other user-defined targets.

A description of the theory and example application (employing the provided scripts) is provided in the following paper:

Taha Rajabzadeh, Alex Boulton-McKeehan, Sam Bonkowsky, David I. Schuster, Amir H. Safavi-Naeini, A General Framework for Gradient-Based Optimization of Superconducting Quantum Circuits using Qubit Discovery as a Case Study, arXiv:2408.12704 (2024), https://arxiv.org/abs/2408.12704.

Usage

qubit-discovery

The qubit-discovery package provides the basic ingredients for running optimization with SQcircuit. The losses subpackage implements functions to evaluate qubit metrics (anharmonicity, decoherence time, …) and utilities to build loss functions out of them. The optim subpackage implements optimization algorithms (SGD, BFGS, and other PyTorch optimizers), utilities to set truncation numbers, and a qubit sampler.

See the notebooks in the tutorials directory for instructions on how to use these features.

• QD_transmon-optim.ipynb provides a quick demo on how to use qubit-discovery to optimize the $T_2$ time, qubit frequency, and anharmonicity of a transmon, in only a few lines of code.
• QD_overview.ipynb gives an introduction to the core functionality of the package.
• QD_advanced-features.ipynb shows the advanced features and customization possible.

Scripts

The scripts directory provides a set of programs which can be used to automate multiple optimization runs via the following workflow:

1. Execute optimize.py in parallel with different seeds, which correspond to distinct intialization points.
2. Collate the results of the different runs using plot_records.py.
3. Output the details of the best-performing circuits with circuit_summary.py and plot_analysis.py.

For a detailed description of this workflow, see parallel_optimization.md.

Installation

To install the qubit-discovery package, download from PyPI:

pip install qubit-discovery

To use the provided scripts, first install qubit-discovery with scripting support:

pip install qubit-discovery[scripts]

Then clone the Qubit-Discovery repository and run the scripts with Python >= 3.9.

Contribution

We welcome contributions to the develoment of Qubit-Discovery! Please feel free to fork the repository and send pull requests, or file bug reports on the issues page. All code contributions are acknowledged in the contributors' section below.