Amazon Braket Python Schemas

Amazon Braket Python Schemas is an open source library that contains the schemas for Braket, including:

• intermediate representations (IR) for Amazon Braket quantum tasks and offers serialization and deserialization of those IR payloads. Think of the IR as the contract between the Amazon Braket SDK and Amazon Braket API for quantum programs.
• schemas for the S3 results of each quantum task
• schemas for the device capabilities of each device

Installation

Prerequisites

• Python 3.7+

Steps

The preferred way to get Amazon Braket Python Schemas is by installing the Amazon Braket Python SDK, which will pull in the schemas. Follow the instructions in the README for setup.

However, if you only want to use the schemas, it can be installed on its own as follows:

pip install amazon-braket-schemas

You can install from source by cloning this repository and running a pip install command in the root directory of the repository:

git clone https://github.com/aws/amazon-braket-schemas-python.git
cd amazon-braket-schemas-python
pip install .

You can check your currently installed version of amazon-braket-schemas with pip show:

pip show amazon-braket-schemas

or alternatively from within Python:

>>> import braket._schemas as braket_schemas
>>> braket_schemas.__version__

Usage

There are currently two types of IR, including jacqd (JsonAwsQuantumCircuitDescription) and annealing. See below for their usage.

Serializing python structures

from braket.ir.jaqcd import CNot, H, Program, Expectation
from braket.ir.annealing import Problem, ProblemType

program = Program(instructions=[H(target=0), CNot(control=0, target=1)])
print(program.json(indent=2))

"""
{
  "braketSchemaHeader": {
    "name": "braket.ir.jaqcd.program",
    "version": "1"
  },
  "instructions": [
    {
      "target": 0,
      "type": "h"
    },
    {
      "control": 0,
      "target": 1,
      "type": "cnot"
    }
  ],
  "results": null,
  "basis_rotation_instructions": null,
}
"""

program = Program(
    instructions=[H(target=0), CNot(control=0, target=1)],
    results=[Expectation(targets=[0], observable=['x'])],
    basis_rotation_instructions=[H(target=0)]
)
print(program.json(indent=2))

"""
{
  "braketSchemaHeader": {
    "name": "braket.ir.jaqcd.program",
    "version": "1"
  },
  "instructions": [
    {
      "target": 0,
      "type": "h"
    },
    {
      "control": 0,
      "target": 1,
      "type": "cnot"
    }
  ],
  "results": [
    {
      "observable": [
        "x"
      ],
      "targets": [
        0
      ],
      "type": "expectation"
    }
  ],
  "basis_rotation_instructions": [
    {
      "target": 0,
      "type": "h"
    }
  ]
}
"""

problem = Problem(type=ProblemType.QUBO, linear={0: 0.3, 4: -0.3}, quadratic={"0,5": 0.667})
print(problem.json(indent=2))

"""
{
  "braketSchemaHeader": {
    "name": "braket.ir.annealing.problem",
    "version": "1"
  },
  "type": "QUBO",
  "linear": {0: 0.3, 4: -0.3},
  "quadratic": {"0,5": 0.667}
}
"""

Deserializing into python structures

from braket.ir.jaqcd import Program
from braket.ir.annealing import Problem

jaqcd_string = """
{
  "instructions": [
    {
      "target": 0,
      "type": "h"
    },
    {
      "control": 0,
      "target": 1,
      "type": "cnot"
    }
  ],
  "results": [
    {
      "observable": [
        "x"
      ],
      "targets": [
        0
      ],
      "type": "expectation"
    }
  ],
  "basis_rotation_instructions": [
    {
      "target": 0,
      "type": "h"
    }
  ]
}
"""

program = Program.parse_raw(jaqcd_string)
print(program)

"""
braketSchemaHeader=BraketSchemaHeader(name='braket.ir.jaqcd.program', version='1') instructions=[H(target=0, type=<Type.h: 'h'>), CNot(control=0, target=1, type=<Type.cnot: 'cnot'>)] results=[Expectation(observable=['x'], targets=[0], type=<Type.expectation: 'expectation'>)] basis_rotation_instructions=[H(target=0, type=<Type.h: 'h'>)]
"""

annealing_string = """
{
  "type": "QUBO",
  "linear": {0: 0.3, 4: -0.3},
  "quadratic": {"0,5": 0.667}
}
"""

problem = Problem.parse_raw(annealing_string)
print(problem)

"""
braketSchemaHeader=BraketSchemaHeader(name='braket.ir.annealing.problem', version='1') type=<ProblemType.QUBO: 'QUBO'>, linear={0: 0.3, 4: -0.3}, quadratic={'0,5': 0.667}
"""

Documentation

Detailed documentation, including the API reference, can be found on Read the Docs.

You can also generate the docs from source. First, install tox:

pip install tox

To build the Sphinx docs, run the following command in the root repo directory:

tox -e docs

You can then find the generated HTML files in build/documentation/html.

Testing

Make sure to install test dependencies first:

pip install -e "amazon-braket-schemas-python[test]"

To run the unit tests, linters, and doc generation:

tox

To select tests based on a keyword:

tox -e py37 -- -k 'your_test'

License

This project is licensed under the Apache-2.0 License.