/tequila

Rapid development of novel quantum algorithms

Primary LanguagePythonMIT LicenseMIT

Image

Tequila

Tequila is an Extensible Quantum Information and Learning Architecture where the main goal is to simplify and accelerate implementation of new ideas for quantum algorithms. It operates on abstract data structures allowing the formulation, combination, automatic differentiation and optimization of generalized objectives. Tequila can execute the underlying quantum expectation values on state of the art simulators as well as on real quantum devices.

You can get an overview from this presentation or from it's video recording

Get started with our Tutorials

or checkout our overview article

Quantum Backends

Currently supported

Tequila detects backends automatically if they are installed on your systems. All of them are available over standard pip installation like for example pip install qulacs. For best performance it is recommended to have qulacs installed.

QuantumChemistry:

Currently supported

  • Psi4. In a conda environment this can be installed with
conda install psi4 -c psi4

Here is a small tutorial that illustrates the usage.

  • Madness Currently you need to compile from a separate fork. See the github page of this fork for installation instruction. Here is a small tutorial that illustrates the usage.

  • PySCF
    Works similar as Psi4. Classical methods are also integrated in the madness interface allowing to use them in a basis-set-free representation.

Install from source

Do not install like this: (Minecraft lovers excluded) pip install tequila

We recommend installing in editable mode with

git clone https://github.com/aspuru-guzik-group/tequila.git
cd tequila
pip install -e .

You can install tequila directly with pip over:

pip install git+https://github.com/aspuru-guzik-group/tequila.git

Install from devel branch (most recent updates):

pip install git+https://github.com/aspuru-guzik-group/tequila.git@devel

Recommended Python version is 3.7. Python 3.8 works, but not all (optional) dependencies support it yet (e.g. Psi4). Python 3.6 works, but some (optional) dependencies might have issues with numpy >= 1.20.

Install from PyPi

Do not install like this: pip install tequila

this will install a Minecraft server manager (might be useful for other things, but probably not what you are looking for).

You can install tequila from PyPi as:

pip install tequila-basic

this will install tequila with all essential dependencies. We recommend to install some fast quantum backends, like qulacs or qibo, as well. Those can be installed before or after you install tequila.

# install basic tequila
pip install tequila-basic
# install qulacs and/or other backends and use it within tequila
pip install qulacs

Install with Windows

pip install git+https://github.com/aspuru-guzik-group/tequila.git@windows

See also the troubleshooting below if you want to tweak things manually.
The command above will not install the qulacs simulator.
You can install it on windows OS, but you need to have cmake and c++ compilers ready (can be installed for example over visual studio).
Of course you can also use one of the other backends (see above).

Getting Started

Check out the tutorial notebooks provided in tutorials.

Tequila Hello World

# optimize a one qubit example

# define a variable
a = tq.Variable("a")
# define a simple circuit
U = tq.gates.Ry(angle=a*pi, target=0)
# define an Hamiltonian
H = tq.paulis.X(0)
# define an expectation value
E = tq.ExpectationValue(H=H, U=U)
# optimize the expectation value
result = tq.minimize(method="bfgs", objective=E**2)
# check out the optimized wavefunction
wfn = tq.simulate(U, variables=result.angles)
print("optimized wavefunction = ", wfn)
# plot information about the optimization
result.history.plot("energies")
result.history.plot("angles")
result.history.plot("gradients")

Chemistry Hello World

# define a molecule within an active space
active = {"a1": [1], "b1":[0]}
molecule = tq.quantumchemistry.Molecule(geometry="lih.xyz", basis_set='6-31g', active_orbitals=active, transformation="bravyi-kitaev")

# get the qubit hamiltonian
H = molecule.make_hamiltonian()

# create an k-UpCCGSD circuit of order k
U = molecule.make_upccgsd_ansatz(order=1, include_singles=True)

# define the expectationvalue
E = tq.ExpectationValue(H=H, U=U)

# compute reference energies
fci = molecule.compute_energy("fci")
cisd = molecule.compute_energy("detci", options={"detci__ex_level": 2})

# optimize
result = tq.minimize(objective=E, method="BFGS", initial_values=0.0)

print("VQE : {:+2.8}f".format(result.energy))
print("CISD: {:+2.8}f".format(cisd))
print("FCI : {:+2.8}f".format(fci))

Do you want to create your own methods? Check out the tutorials!

Research projects using Tequila

J.S. Kottmann, A. Anand, A. Aspuru-Guzik.
A Feasible Approach for Automatically Differentiable Unitary Coupled-Cluster on Quantum Computers.
Chemical Science, 2021, doi.org/10.1039/D0SC06627C.
arxiv.org/abs/2011.05938
General techniques are implemented in the chemistry modules of tequila.
See the tutorials for examples.

J.S. Kottmann, P. Schleich, T. Tamayo-Mendoza, A. Aspuru-Guzik.
Reducing Qubit Requirements while Maintaining Numerical Precision for the Variational Quantum Eigensolver: A Basis-Set-Free Approach.
J.Phys.Chem.Lett., 2021, doi.org/10.1021/acs.jpclett.0c03410.
arxiv.org/abs/2008.02819
example code
tutorial on the madness interface

A. Cervera-Lierta, J.S. Kottmann, A. Aspuru-Guzik.
The Meta-Variational Quantum Eigensolver.
arxiv.org/abs/2009.13545
example code

J.S. Kottmann, M. Krenn, T.H. Kyaw, S. Alperin-Lea, A. Aspuru-Guzik.
Quantum Computer-Aided design of Quantum Optics Hardware.
arxiv.org/abs/2006.03075
example code
slides

A. Anand, M. Degroote, A. Aspuru-Guzik.
Natural Evolutionary Strategies for Variational Quantum Computation.
arxiv.org/abs/2012.00101

J. S. Kottmann, A. Aspuru-Guzik,
Optimized Low-Depth Quantum Circuits for Molecular Electronic Structure using a Separable Pair Approximation,
arxiv.org/abs/2105.03836
example code

K. Choudhary,
Quantum Computation for Predicting Electron and Phonon Properties of Solids
arxiv.org/abs/2102.11452

Let us know, if you want your research project and/or tutorial to be included in this list!

Dependencies

Support for additional optimizers or quantum backends can be activated by intalling them in your environment. Tequila will then detect them automatically. Currently those are: Phoenics and GPyOpt. Quantum backends are treated in the same way.

Documentation

You can build the documentation by navigating to docs and entering make html. Open the documentation with a browser over like firefox docs/build/html/index.html Note that you will need some additional python packages like sphinx and mr2 that are not explicitly listed in the requirements.txt

You can also visit our prebuild online documentation that will correspond to the github master branch

How to contribute

If you find any bugs or inconveniences in tequila please don't be shy and let us know. You can do so either by raising an issue here on github or contact us directly.

If you already found a solution you can contribute to tequila over a pull-request. Here is how that works:

  1. Make a fork of tequila to your own github account.
  2. Checkout the devel branch and make sure it is up to date with the main github repository.
  3. Create and checkout a new branch from devel via git branch pr-my-branch-name followed by git checkout pr-my-branch-name. By typing git branch afterwards you can check which branch is currently checked out on your computer.
  4. Introduce changes to the code and commit them with git.
  5. Push the changes to your github account
  6. Log into github and create a pull request to the main github repository. The pull-request should be directed to the devel branch (but we can also change that afterwards).

If you plan to introduce major changes to the base library it can be beneficial to contact us first. This way we might be able to avoid conflicts before they arise.

If you used tequila for your research, feel free to include your algorithms here, either by integrating it into the core libraries or by demonstrating it with a notebook in the tutorials section. If you let us know about it, we will also add your research article in the list of research projects that use tequila (see above).

Troubleshooting

If you experience trouble of any kind or if you either want to implement a new feature or want us to implement a new feature that you need: Don't hesitate to contact us directly or raise an issue here on github.

PySCF

If pyscf crashes on import with

Using default_file_mode other than 'r' is no longer supported. Pass the mode to h5py.File() instead

then you need to downgrade the h5py version

pip install --upgrade 'h5py <= 3.1'

The issue will probably be fixed soon in pyscf.

Qiskit backend

Qiskit version 0.25 is not yet supported. pip install --upgrade qiskit<0.25 fixes potential issues. If not: Please let us know.

Circuit drawing

Standard graphical circuit representation within a Jupyter environment is often done using tq.draw. Without further keywords tequial will try to create and compile a qpic file. For proper display you will need the following dependencies: qpic, pdflatex and convert/ImageMagick (pre-installed on most GNU/Linux distributions, not pre-installed on macs). On GNU/Linux distributions sometimes the permissions of convert to convert pdf to png are not granted, resulting in an error when trying to use tq.draw. Click here for a possible solution.

In general, there is no reason to worry if tq.draw does not function properly. It is just one way to display circuits, but not neccessary to have. Alternatives are:

  • Use tq.draw(circuit, backend="qiskit") (or backend=cirq )
  • translate to qiskit/cirq and use their functionality ( qiskit_circuit = tq.compile(circuit, backend='qiskit').circuit )
  • directly create pdfs: tq.circuit.export_to(circuit, filename="my_name.pdf") (will also create my_name.qpic that can be used with qpic)
  • use print(circuit) (does not look pretty, but carries the same information).
  • become a contributor and implement your own graphical circuit representation and create a pull-request.

Qulacs simulator

You will need cmake to install the qulacs simulator pip install cmake

You don't need qulacs for tequila to run (although is is recommended) To install without qulacs just remove the qulacs line from requirements.txt It can be replaced by one (or many) of the other supported simulators. Note that simulators can also be installed on a later point, they don't need to be installed with tequila. As long as they are installed within the same python environment tequila can detect them.

Windows

You can in principle use tequila with windows as OS and have almost full functionality. You will need to replace Jax with autograd for it to work. In order to do so: Remove jax and jaxlib from setup.py and requirements.txt and add autograd instead.

In order to install qulacs you will need latest GNU compilers (at least gcc-7). They can be installed for example over visual studio.

Mac OS

Tequila runs on Mac OSX. You might get in trouble with installing qulacs since it currently does not work with Apple's clang compiler. You need to install latest GNU compile (at least gcc-7 and g++7) and set them as default before installing qulacs over pip.

Qibo and GPyOpt

Currently you can't use Qibo and GPyOpt within the same environment.