THULab/qiskit_calibration_drift_nazimari

IBM Quantum Calibration Drift (TsFile)

Apache TsFile version of nazimari/qiskit-calibration-drift.

Overview

Continuously-updated calibration data from IBM Quantum hardware with concurrent environmental measurements. The dataset enables correlation analysis between qubit performance and atmospheric / space-weather conditions.

Property	Value
Update frequency	Every 30 minutes
Backends	ibm_fez (156 qubits), ibm_torino (133 qubits), ibm_marrakesh (156 qubits)
Total qubits	445
Collection method	Automated polling via GitHub Actions
Calibration source	IBM Quantum Runtime API
Weather source	NWS API (NOAA)
Space weather source	SWPC (NOAA)

• Rows: 155,368 calibration measurements (1 per backend/property/qubit sample).
• Calibrated period: calibrated_time ranges from 2025-07-12 to 2026-05-05 (UTC); observed_time reflects the actual polling window.
• Sampling: one record per calibration property and qubit, collected on a ~30-minute polling cadence.

Schema (TsFile structure)

• Time (INT64, milliseconds) — the IBM calibration timestamp (calibrated_time), i.e. when IBM last calibrated that specific property.
• backend, property, qubit (TAG) — the device/measurement dimensions. Query one qubit's calibration history with WHERE backend='ibm_fez' AND property='T1' AND qubit='0'.
• value (DOUBLE) — measured value of the calibration property.
• observed_time (STRING) — UTC collection timestamp of the record.
• latitude, longitude, solar_zenith_deg, temperature_c, pressure_hpa, humidity_pct (DOUBLE) — concurrent data-center environmental measurements.
• kp_index, solar_flux_sfu, dst_nt, bz_gsm_nt, neutron_flux (DOUBLE) — concurrent space-weather indices.

calibrated_time is retained as the TsFile Time; observed_time is also kept as a FIELD so the true collection period is preserved. Following the source card's guidance, calibrated_time can be stale (IBM does not recalibrate all properties simultaneously), whereas observed_time reflects the current collection period.

The source card lists a location column, but the released parquet does not contain it, so no location TAG is present.

Calibration properties

Per-qubit: T1 (relaxation time, seconds), T2 (dephasing time, seconds), readout_error (measurement error probability), prob_meas0_prep1 (P(measure 0 | prepared 1)), prob_meas1_prep0 (P(measure 1 | prepared 0)), sx_error (SX gate error).

Per-edge: cz_error_{i}_{j} (two-qubit CZ gate error for the edge between qubits i and j).

Usage

Install the Apache TsFile Python SDK (pip install tsfile) and read a converted file:

from pathlib import Path
from tsfile import TsFileReader

path = Path("qiskit_calibration_drift.tsfile")
with TsFileReader(str(path)) as reader:
    schemas = reader.get_all_table_schemas()
    print("tables:", list(schemas))
    table_name = next(iter(schemas))
    table = schemas[table_name]
    columns = [column.get_column_name() for column in table.get_columns()]
    print("columns:", columns)
    field_names = [
        column.get_column_name()
        for column in table.get_columns()
        if column.get_column_name() not in {"Time", "time"}
    ]
    if field_names:
        with reader.query_table(table_name, field_names[:3], batch_size=1024) as result:
            batch = result.read_arrow_batch()
            if batch is not None:
                print(batch.to_pandas().head())

Source & license

• Original dataset: https://huggingface.co/datasets/nazimari/qiskit-calibration-drift
• Author / publisher: nazimari (original dataset card)
• License: cc-by-4.0