AutoQASM

AutoQASM is not an officially supported AWS product.

This experimental module offers a new quantum-imperative programming experience embedded in Python for developing quantum programs. AutoQASM is experimental software. We may change, remove, or deprecate parts of the AutoQASM API without notice. The name AutoQASM is a working title and is also subject to change.

For a fully supported quantum development experience, please use the Amazon Braket Python SDK by following these instructions. If you are interested in our active development efforts on AutoQASM, and you are not afraid of a few bugs, please keep on reading!

Why AutoQASM?

AutoQASM provides a Pythonic developer experience for writing quantum programs. The working title "AutoQASM" is derived from the name of the AutoGraph module of TensorFlow. AutoQASM uses AutoGraph to construct quantum assembly (QASM) programs rather than TensorFlow graphs.

AutoQASM provides a natural interface for expressing quantum programs with mid-circuit measurements and classical control flow using native Python language features. It allows the construction of modular programs consisting of common programming constructs such as loops and subroutines. This enables a more imperative programming style than constructing programs via a series of function calls on a circuit object.

AutoQASM programs can be serialized to OpenQASM. This textual representation for quantum programs is widely supported and enables interoperability among various frameworks. A crucial part of our serialization process is that modular structures within the program, such as loops and subroutines, are preserved when serializing to OpenQASM.

Although it is still a work in progress, the intent is that AutoQASM will support any quantum programming paradigm which falls into the OpenQASM 3.0 language scope. AutoQASM supports serializing quantum programs to OpenQASM, which allows the programs to interoperate with any library or service that supports OpenQASM programs, such as Amazon Braket.

See the Quick Start section below, as well as the AutoQASM example notebooks, for examples of AutoQASM usage.

Installation

To install the latest experimental release of AutoQASM, use pip:

pip install autoqasm

Alternatively, to use AutoQASM in development mode, install from a local clone of this GitHub repository:

git clone https://github.com/amazon-braket/autoqasm.git
cd autoqasm
pip install -e .

Quick start

In this section, we will show how to get started with AutoQASM. AutoQASM allows you to build quantum programs with a simplified syntax and run the programs on the service. It uses the circuit model programming paradigm that is also used in the Amazon Braket SDK.

First, import the following modules and functions:

import autoqasm as aq
from autoqasm.instructions import h, cnot, measure

To create a quantum program using the AutoQASM experience, you decorate a function with @aq.main. This allows AutoQASM to hook into the program definition and generate an output format that is accepted by quantum devices.

For instance, we can create a Bell state like so:

# A program that generates a maximally entangled state
@aq.main
def bell_state():
    h(0)
    cnot(0, 1)

You can view the output format, which is OpenQASM, by running bell_state.build().display().

AutoQASM enables users to use more complicated program constructs with a compact and readable structure. We can demonstrate this with a program that conditionally prepares multiple Bell states on qubit pairs (1, 2) and (3, 4).

@aq.main(num_qubits=5)
def conditional_multi_bell_states() -> None:
    h(0)
    if measure(0):
        for i in aq.range(2):
            qubit = 2 * i + 1
            h(qubit)
            cnot(qubit, qubit+1)

    measure([0,1,2,3,4])

AutoQASM can support subroutines and complex control flow. You can use the Python runtime and quantum runtime side-by-side. There are rough edges at the moment, but we're actively smoothing them out!

AutoQASM includes a simulator which can be accessed using the Amazon Braket local simulator interface. Let's simulate the conditional_multi_bell_states program:

from braket.devices import LocalSimulator

device = LocalSimulator("autoqasm")
task = device.run(conditional_multi_bell_states, shots=100)
result = task.result()

Read more about AutoQASM decorators like @aq.main here.

Read more about using AutoQASM with Amazon Braket Hybrid Jobs here.

For more example usage of AutoQASM, visit the example notebooks.

Architecture

AutoQASM is built on top of the autograph component of TensorFlow. A quantum program is written as a Python function which is decorated with @aq.main. When calling this decorated function, the user’s Python function is converted into a transformed Python function by autograph. This transformed function is then executed to produce an AutoQASM Program object which can be simulated and/or serialized to OpenQASM.

The conversion process allows AutoQASM to provide custom handling for native Python control flow keywords such as if, for, and while and to preserve this control flow in the resulting quantum program in order to realize functionality such as classical feedback on mid-circuit measurement, efficient representation of loops, and modularity of subroutine definitions.

Plans

The AutoQASM project is undergoing rapid development. The current status and future plans are tracked in the AutoQASM GitHub project.

Contributing and sharing feedback

We welcome feature requests, bug reports, or general feedback, which you can share with us by opening up an issue. We also welcome pull requests, examples, and documentation -- please open an issue describing your work when you get started, or comment on an existing issue with your intentions. For more details on contributing to AutoQASM, please read the contributing guidelines.

For questions, you can get help via the Quantum Technologies section of AWS RePost. Please tag your question with "Amazon Braket" and mention AutoQASM in the question title.

Tests

To run AutoQASM unit tests, run:

tox -e unit-tests

Frequently asked questions

1. Will AutoQASM be extended to contain a library of quantum algorithms or quantum applications?

No, we are focused on AutoQASM as an interface for low-level expression of quantum programs: circuits, gates and pulses. Higher-level algorithm libraries could be implemented using AutoQASM and benefit from modular AutoQASM functionality such as subroutines.

2. What is the relationship between AutoQASM and OpenQASM?

AutoQASM is a quantum programming interface built in Python. OpenQASM is a quantum assembly language, often used as a serialization format for quantum programming frameworks and quantum hardware providers. We can represent a quantum program equivalently in either format, but using AutoQASM allows one to also make use of Python, including the Amazon Braket SDK.

AutoQASM can be seen as implementing a builder pattern for OpenQASM. It allows you serialize your program to OpenQASM by calling to_ir() on the built program. The interface is not strongly tied to OpenQASM, so we could serialize to other formats in the future.

3. What is the relationship between AutoQASM and the Amazon Braket SDK?

AutoQASM lives alongside the Amazon Braket SDK as an experimental package. It supplements the program building experience and integrates with Amazon Braket SDK features. For instance, one can build a program through AutoQASM, and then use the SDK to run the program on a local simulator or on an Amazon Braket device.

4. Does AutoQASM support other providers beyond Amazon Braket?

Yes. AutoQASM serializes to OpenQASM, and so it is applicable to any library or QPU that supports OpenQASM. AutoQASM does have some features that use the Amazon Braket SDK, such as device-specific validation. Because AutoQASM is open-source, anyone could build similar integrations for another service. Reach out if you're interested in doing this and would like support.

5. Does AutoQASM offer special support for device-specific programming?

Yes, AutoQASM has device-specific validation to support native programming. We plan to expand this functionality in the future. Learn more with our native programming example notebook.

6. Do the devices available through Amazon Braket support all of AutoQASM's features?

No, for example, the reset instruction is not supported by all devices. In general, different QPUs and QHPs support different sets of features, so AutoQASM will often support features that a particular device doesn't support. We intend that AutoQASM will eventually be able to generate any program representable by OpenQASM 3.0, with additional Python-side features such as validation and visualization.