From b1a1c1c545ca65a3a20e11276da26c674cc8f889 Mon Sep 17 00:00:00 2001 From: eliottrosenberg <61400172+eliottrosenberg@users.noreply.github.com> Date: Wed, 13 Mar 2024 01:47:10 -0500 Subject: [PATCH] Update T1 experiment (#6487) * T1 experiment: add A and B constants to fit, make wait times log-spaced * lowercase variable names * update tests * update tests * update tests * update tests * update tests * update tests * update tests * informative assert statements in tests * use float wait time * typecheck * nits --- .../cirq/experiments/t1_decay_experiment.py | 21 ++++++---- .../experiments/t1_decay_experiment_test.py | 40 +++++++------------ 2 files changed, 28 insertions(+), 33 deletions(-) diff --git a/cirq-core/cirq/experiments/t1_decay_experiment.py b/cirq-core/cirq/experiments/t1_decay_experiment.py index 0d44db7b412..a8ee85e70b5 100644 --- a/cirq-core/cirq/experiments/t1_decay_experiment.py +++ b/cirq-core/cirq/experiments/t1_decay_experiment.py @@ -12,7 +12,7 @@ # See the License for the specific language governing permissions and # limitations under the License. -from typing import Any, Optional, TYPE_CHECKING +from typing import Any, Optional, Sequence, TYPE_CHECKING, cast import warnings import pandas as pd @@ -77,7 +77,12 @@ def t1_decay( var = sympy.Symbol('delay_ns') - sweep = study.Linspace(var, start=min_delay_nanos, stop=max_delay_nanos, length=num_points) + if min_delay_nanos == 0: + min_delay_nanos = 0.4 + sweep_vals_ns = np.unique( + np.round(np.logspace(np.log10(min_delay_nanos), np.log10(max_delay_nanos), num_points)) + ) + sweep = study.Points(var, cast(Sequence[float], sweep_vals_ns)) circuit = circuits.Circuit( ops.X(qubit), ops.wait(qubit, nanos=var), ops.measure(qubit, key='output') @@ -118,8 +123,8 @@ def data(self) -> pd.DataFrame: def constant(self) -> float: """The t1 decay constant.""" - def exp_decay(x, t1): - return np.exp(-x / t1) + def exp_decay(x, t1, a, b): + return a * np.exp(-x / t1) + b xs = self._data['delay_ns'] ts = self._data['true_count'] @@ -132,8 +137,8 @@ def exp_decay(x, t1): # Fit to exponential decay to find the t1 constant try: - popt, _ = optimize.curve_fit(exp_decay, xs, probs, p0=[t1_guess]) - t1 = popt[0] + self.popt, _ = optimize.curve_fit(exp_decay, xs, probs, p0=[t1_guess, 1.0, 0.0]) + t1 = self.popt[0] return t1 except RuntimeError: warnings.warn("Optimal parameters could not be found for curve fit", RuntimeWarning) @@ -166,7 +171,9 @@ def plot( ax.plot(xs, ts / (fs + ts), 'ro-', **plot_kwargs) if include_fit and not np.isnan(self.constant): - ax.plot(xs, np.exp(-xs / self.constant), label='curve fit') + t1 = self.constant + t1, a, b = self.popt + ax.plot(xs, a * np.exp(-xs / t1) + b, label='curve fit') plt.legend() ax.set_xlabel(r"Delay between initialization and measurement (nanoseconds)") diff --git a/cirq-core/cirq/experiments/t1_decay_experiment_test.py b/cirq-core/cirq/experiments/t1_decay_experiment_test.py index 64d220dd5c1..acbd7526433 100644 --- a/cirq-core/cirq/experiments/t1_decay_experiment_test.py +++ b/cirq-core/cirq/experiments/t1_decay_experiment_test.py @@ -53,7 +53,7 @@ def noisy_moment(self, moment, system_qubits): repetitions=10, max_delay=cirq.Duration(nanos=500), ) - results.plot() + results.plot(include_fit=True) def test_result_eq(): @@ -61,7 +61,7 @@ def test_result_eq(): eq.make_equality_group( lambda: cirq.experiments.T1DecayResult( data=pd.DataFrame( - columns=['delay_ns', 'false_count', 'true_count'], index=[0], data=[[100.0, 2, 8]] + columns=['delay_ns', 'false_count', 'true_count'], index=[0], data=[[100, 2, 8]] ) ) ) @@ -103,7 +103,7 @@ def noisy_moment(self, moment, system_qubits): data=pd.DataFrame( columns=['delay_ns', 'false_count', 'true_count'], index=range(4), - data=[[100.0, 0, 10], [400.0, 0, 10], [700.0, 10, 0], [1000.0, 10, 0]], + data=[[100.0, 0, 10], [215.0, 0, 10], [464.0, 0, 10], [1000.0, 10, 0]], ) ) @@ -117,13 +117,14 @@ def test_all_on_results(): min_delay=cirq.Duration(nanos=100), max_delay=cirq.Duration(micros=1), ) - assert results == cirq.experiments.T1DecayResult( + desired = cirq.experiments.T1DecayResult( data=pd.DataFrame( columns=['delay_ns', 'false_count', 'true_count'], index=range(4), - data=[[100.0, 0, 10], [400.0, 0, 10], [700.0, 0, 10], [1000.0, 0, 10]], + data=[[100.0, 0, 10], [215.0, 0, 10], [464.0, 0, 10], [1000.0, 0, 10]], ) ) + assert results == desired, f'{results.data=} {desired.data=}' def test_all_off_results(): @@ -135,13 +136,14 @@ def test_all_off_results(): min_delay=cirq.Duration(nanos=100), max_delay=cirq.Duration(micros=1), ) - assert results == cirq.experiments.T1DecayResult( + desired = cirq.experiments.T1DecayResult( data=pd.DataFrame( columns=['delay_ns', 'false_count', 'true_count'], index=range(4), - data=[[100.0, 10, 0], [400.0, 10, 0], [700.0, 10, 0], [1000.0, 10, 0]], + data=[[100.0, 10, 0], [215.0, 10, 0], [464.0, 10, 0], [1000.0, 10, 0]], ) ) + assert results == desired, f'{results.data=} {desired.data=}' @pytest.mark.usefixtures('closefigures') @@ -150,28 +152,14 @@ def test_curve_fit_plot_works(): data=pd.DataFrame( columns=['delay_ns', 'false_count', 'true_count'], index=range(4), - data=[[100.0, 6, 4], [400.0, 10, 0], [700.0, 10, 0], [1000.0, 10, 0]], + data=[[100.0, 6, 4], [215.0, 10, 0], [464.0, 10, 0], [1000.0, 10, 0]], ) ) good_fit.plot(include_fit=True) -@pytest.mark.usefixtures('closefigures') -def test_curve_fit_plot_warning(): - bad_fit = cirq.experiments.T1DecayResult( - data=pd.DataFrame( - columns=['delay_ns', 'false_count', 'true_count'], - index=range(4), - data=[[100.0, 10, 0], [400.0, 10, 0], [700.0, 10, 0], [1000.0, 10, 0]], - ) - ) - - with pytest.warns(RuntimeWarning, match='Optimal parameters could not be found for curve fit'): - bad_fit.plot(include_fit=True) - - -@pytest.mark.parametrize('t1', [200, 500, 700]) +@pytest.mark.parametrize('t1', [200.0, 500.0, 700.0]) def test_noise_model_continous(t1): class GradualDecay(cirq.NoiseModel): def __init__(self, t1: float): @@ -196,10 +184,10 @@ def noisy_moment(self, moment, system_qubits): results = cirq.experiments.t1_decay( sampler=cirq.DensityMatrixSimulator(noise=GradualDecay(t1)), qubit=cirq.GridQubit(0, 0), - num_points=4, + num_points=10, repetitions=10, - min_delay=cirq.Duration(nanos=100), - max_delay=cirq.Duration(micros=1), + min_delay=cirq.Duration(nanos=1), + max_delay=cirq.Duration(micros=10), ) assert np.isclose(results.constant, t1, 50)