When running GST with a FOGI model, the output model's labels get messed up. It seems like every iteration, the model gets rebuilt and during the label construction it includes a combination of the previous labels. Creating a nested pattern of labels which grows at an exponential rate, and causing huge memory consumption. Pointed out by @jordanh6
Steps to reproduce the behavior:
Create a FOGI model
import pygsti
from pygsti.modelpacks import smq1Q_XY as std
import numpy as np
from FOGI.fogi import create_fogiASMP_smq1Q_XY_model
fogi_model = create_fogiASMP_smq1Q_XY_model(pygsti_model=True)
sserita
commented
2 months ago
When running GST with a FOGI model, the output model's labels get messed up. It seems like every iteration, the model gets rebuilt and during the label construction it includes a combination of the previous labels. Creating a nested pattern of labels which grows at an exponential rate, and causing huge memory consumption. Pointed out by @jordanh6
Steps to reproduce the behavior:
Create a FOGI model
import pygsti from pygsti.modelpacks import smq1Q_XY as std import numpy as np from FOGI.fogi import create_fogiASMP_smq1Q_XY_model fogi_model = create_fogiASMP_smq1Q_XY_model(pygsti_model=True)
Generate data to run GST
datagen_model = std.target_model("GLND") errgen_vector = np.zeros(datagen_model.num_params) errgen_vector[4] = .1 datagen_model.from_vector(errgen_vector) design = std.create_gst_experiment_design(32) dataset = pygsti.data.simulate_data(datagen_model, design.all_circuits_needing_data, num_samples=100000, seed=20230217, sample_error='none')
data = pygsti.protocols.ProtocolData(design, dataset)
Run GST
protoOpt = pygsti.protocols.GateSetTomography(fogi_model, verbosity=0, gaugeopt_suite=None)
Expected behavior The labels of the result model should be unchanged
Environment: