pytket-cutensornet

Pytket is a python module for interfacing with tket, a quantum computing toolkit and optimising compiler developed by Quantinuum.

cuTensorNet is a high-performance library for tensor network computations, developed by NVIDIA. It is part of the cuQuantum SDK - a high-performance library aimed at quantum circuit simulations on the NVIDIA GPUs.

pytket-cutensornet is an extension to pytket that allows pytket circuits and expectation values to be simulated using cuTensorNet via an interface to cuQuantum Python.

Some useful links:

• API Documentation

Getting started

pytket-cutensornet is available for Python 3.10, 3.11 and 3.12 on Linux. In order to use it, you need access to a Linux machine (or WSL) with an NVIDIA GPU of Compute Capability +7.0 (check it here). You will need to install cuda-toolkit and cuquantum-python before pytket-cutensornet; for instance, in Ubuntu 24.04:

sudo apt install cuda-toolkit
pip install cuquantum-python
pip install pytket-cutensornet

Alternatively, you may install cuQuantum Python following their instructions using conda-forge. This will include the necessary dependencies from CUDA toolkit. Then, you may install pytket-cutensornet using pip.

Bugs, support and feature requests

Please file bugs and feature requests on the Github issue tracker.

Development

To install an extension in editable mode, from its root folder run:

pip install -e .

Contributing

Pull requests are welcome. To make a PR, first fork the repo, make your proposed changes on the main branch, and open a PR from your fork. If it passes tests and is accepted after review, it will be merged in.

Code style

Docstrings

We use the Google style docstrings, please see this page for reference.

Formatting

All code should be formatted using black, with default options. This is checked on the CI. The CI is currently using version 22.12.0. You can install it (as well as pylint as described below) by running from the root package folder:

pip install -r lint-requirements.txt

Type annotation

On the CI, mypy is used as a static type checker and all submissions must pass its checks. You should therefore run mypy locally on any changed files before submitting a PR. Because of the way extension modules embed themselves into the pytket namespace this is a little complicated, but it should be sufficient to run the script mypy-check and passing as a single argument the root directory of the module to test. The directory path should end with a /. For example, to run mypy on all Python files in this repository, when in the root folder, run:

./mypy-check ./

The script requires mypy 0.800 or above.

Linting

We use pylint on the CI to check compliance with a set of style requirements (listed in .pylintrc). You should run pylint over any changed files before submitting a PR, to catch any issues.

Tests

To run the tests for a module:

pip install -r tests/test-requirements.txt
pytest tests/

When adding a new feature, please add a test for it. When fixing a bug, please add a test that demonstrates the fix.