Skip to main content

QCalibrate remote interface

Project description

A client library for runnig a control optimization with qruise calibrate software

Note

The API is experimental and subject to change without a prior notice

Description

The qcalibrateremote package provides interface to QCalibrate optimization service, providing algorithms helping to define optimal control for quantum system. The actual optimization algorithm runs on the server, supplying the set of parameters and/or PWC pulses to client side. The client code evaluates the parameters by deriving and applying control signals to a system under control, performing measurement, calcualates and returns the figure of merit (infidelity) value. The algorithm tries different parameters in order to find optimal values to achieve minimal infidelity. Live progress of optimization can be observed in the in web UI, given opportunity to finish the optimization, before the optimization stopping criteria are achieved.

Currently two optimization modes are supported.

  • pure parameter optimization

  • Random chopped base PWC function optimization (Fourier and Sigmoid bases)

Installing

Install with pip:

$ pip install qcalibrateremote

for conda environment install grcpio explicitly

$ conda install grpcio

Usage

Prerequisits

  • Python 3.8+ (developed and tested with 3.8.5)

  • Qruise Calibrate account (contact r.razilov@fz-juelich.de for details)

  • Direct internet connection to server (may require VPN access)

Experiment

Experiment defines is a set of meta-parameters controlling the optimization Use use API or Web UI to create an experiment and define optimization parameters. Use online help to get learn about details. The evaluation of figure of merit class can be supplied as class evaluate_fom_class=.. or object evaluate_fom_object=..

Pure parameter optimization example

# import dependencies
from typing import Dict

from qcalibrateremote import (
    EvaluateFigureOfMerit,
    FigureOfMerit,
    create_optimizer_client,
)

# setup client connection (copy form web UI: https://www.qcalibrate.staging.optimal-control.net:31603)
experiment_id="0xabcd"
token=("ey...")

optimizer_client = create_optimizer_client(
    host="grpc.qcalibrate.staging.optimal-control.net", port=31603, token=token)

# define infidelity evaluation class
class DistanceFom(EvaluateFigureOfMerit):

    def __init__(self, *args, **kwargs) -> None:
        super().__init__()

    def infidelity(self, param1, param2) -> float:
        return (param1 - 0.55)**2 + (param2 - 0.33)**2

    def evaluate(self, parameters: Dict[str, float], **kwargs) -> FigureOfMerit:
        """Abstract method for figure of merit evaluation"""
        # print(parameters)
        return FigureOfMerit(self.infidelity(**parameters), '')

# run optimization
optimization_result = optimizer_client.run(experiment_id=experiment_id, evaluate_fom_class=DistanceFom)

# best fitting parameters
optimization_result.top[0].parameters

Rather than create a completely new configuration, one can update an existing experiment configuration

with optimizer_client.update_experiment(experiment_id) as experiment_builder:
    with experiment_builder.configuration() as configuration:
        with configuration.parameter("param1") as param1:
            param1.initial_value = 0.7


optimization_result = optimizer_client.run(
    experiment_id=experiment_id, evaluate_fom_class=DistanceFom)

# best fitting parameters
optimization_result.top[0].parameters

Pulse optimization example

# import dependencies
from typing import Dict

from qcalibrateremote import (
    EvaluateFigureOfMerit,
    FigureOfMerit,
    create_optimizer_client,
    Pulse,
)

# setup client connection (copy form web UI: https://www.qcalibrate.staging.optimal-control.net:31603)
token=("ey...")

optimizer_client = create_optimizer_client(
    host="grpc.qcalibrate.staging.optimal-control.net", port=31603, token=token)

experiment_builder = optimizer_client.create_pulse_optimization_experiment("Pulse optimization", "Created by " + __file__)

# define configuration

with experiment_builder.configuration() as configuration:

    with configuration.time("time1") as time1:
        time1.initial_value = 1
        time1.optimize = False

    with configuration.pulse("pulse1") as pulse1:
        pulse1.time_name = "time1"
        pulse1.lower_limit = -1
        pulse1.upper_limit = 1
        pulse1.bins_number = 21
        with pulse1.fourier_basis() as fourier_basis:
            fourier_basis.basis_vector_number = 5
            with fourier_basis.uniform_super_parameter_distribution() as uniform_super_parameter_distribution:
                uniform_super_parameter_distribution.lower_limit = 0.01
                uniform_super_parameter_distribution.upper_limit = 5

        with pulse1.initial_guess() as initial_guess:
            initial_guess.function = "lambda t: 1"
        with pulse1.scaling_function() as scaling_function:
            scaling_function.function = "lambda t: np.exp(-(t - 0.5)**2/(2*0.2**2))"

    with configuration.dcrab_settings() as dcrab_settings:
        dcrab_settings.maximum_iterations_per_super_iteration = 50
        dcrab_settings.super_iteration_number = 6

experiment_id = optimizer_client.add_experiment(experiment_builder)

# define infidelity evaluation class
def expected_pulse(t):
    return np.sin(2*np.pi*t)**4

class SineFom(EvaluateFigureOfMerit):

    def evaluate(self, parameters: Dict[str, float], pulses: Dict[str, Pulse], **kwargs) -> FigureOfMerit:
        pulse1 = pulses["pulse1"]

        inf = np.sum((expected_pulse(pulse1.times) - pulse1.values)**2)

        return FigureOfMerit(inf, '{}')

# run optimization
optimization_result = optimizer_client.run(experiment_id=experiment_id, evaluate_fom_object=SineFom())

# plot best fitting pulse
pulse1 = optimization_result.top[0].pulses["pulse1"]
import matplotlib.pyplot as plt

plt.plot(pulse1.times, expected_pulse(pulse1.times))
plt.plot(pulse1.times, pulse1.values)

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

qcalibrateremote-0.2.3.tar.gz (2.1 MB view hashes)

Uploaded Source

Built Distribution

qcalibrateremote-0.2.3-py2.py3-none-any.whl (36.4 kB view hashes)

Uploaded Python 2 Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page