Add calibration-to-noise pipeline

cirq-google/cirq_google/engine/calibration_to_noise_properties.py

@@ -0,0 +1,178 @@
+# Copyright 2022 The Cirq Developers
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     https://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+"""Tools for converting Calibrations to NoiseProperties.
+
+Given a Calibration "cal", a user can simulate noise approximating that
+calibration using the following pipeline:
+
+    noise_props = cg.noise_properties_from_calibration(cal)
+    noise_model = cg.NoiseModelFromGoogleNoiseProperties(noise_props)
+    simulator = cirq.Simulator(noise=noise_model)
+    simulator.simulate(circuit)
+"""
+
+from typing import Dict, Tuple, Type, TYPE_CHECKING
+import numpy as np
+
+from cirq import ops
+from cirq.devices import noise_utils
+from cirq_google import engine
+from cirq_google import ops as cg_ops
+from cirq_google.devices import google_noise_properties
+
+if TYPE_CHECKING:
+    import cirq
+
+
+def _unpack_1q_from_calibration(
+    metric_name: str, calibration: engine.Calibration
+) -> Dict['cirq.Qid', float]:
+    """Converts a single-qubit metric from Calibration to dict format."""
+    if metric_name not in calibration:
+        return {}
+    return {
+        engine.Calibration.key_to_qubit(key): engine.Calibration.value_to_float(val)
+        for key, val in calibration[metric_name].items()
+    }
+
+
+def _unpack_2q_from_calibration(
+    metric_name: str, calibration: engine.Calibration
+) -> Dict[Tuple['cirq.Qid', ...], float]:
+    """Converts a two-qubit metric from Calibration to dict format."""
+    if metric_name not in calibration:
+        return {}
+    return {
+        engine.Calibration.key_to_qubits(key): engine.Calibration.value_to_float(val)
+        for key, val in calibration[metric_name].items()
+    }
+
+
+def noise_properties_from_calibration(
+    calibration: engine.Calibration,
+) -> google_noise_properties.GoogleNoiseProperties:
+    """Translates between a Calibration object and a NoiseProperties object.
+
+    The NoiseProperties object can then be used as input to the NoiseModelFromNoiseProperties
+    class (cirq.devices.noise_properties) to create a NoiseModel that can be used with a simulator.
+
+    To manually override noise properties, call `override` on the output:
+
+        # Set all gate durations to 37ns.
+        noise_properties_from_calibration(cal).override(gate_times_ns=37)
+
+    See GoogleNoiseProperties for details.
+
+    Args:
+        calibration: a Calibration object with hardware metrics.
+    """
+
+    # TODO: acquire this based on the target device.
+    # Default map of gates to their durations.
+    default_gate_ns: Dict[Type['cirq.Gate'], float] = {
+        ops.ZPowGate: 25.0,
+        ops.MeasurementGate: 4000.0,
+        ops.ResetChannel: 250.0,
+        ops.PhasedXZGate: 25.0,
+        ops.FSimGate: 32.0,
+        ops.ISwapPowGate: 32.0,
+        ops.CZPowGate: 32.0,
+        # ops.WaitGate is a special case.
+    }
+
+    # Unpack all values from Calibration object
+    # 1. Extract T1 for all qubits
+    T1_micros = _unpack_1q_from_calibration('single_qubit_idle_t1_micros', calibration)
+    t1_ns = {q: T1_micro * 1000 for q, T1_micro in T1_micros.items()}
+
+    # 2. Extract Tphi for all qubits
+    rb_incoherent_errors = _unpack_1q_from_calibration(
+        'single_qubit_rb_incoherent_error_per_gate', calibration
+    )
+    tphi_ns = {}
+    if rb_incoherent_errors:
+        microwave_time_ns = default_gate_ns[ops.PhasedXZGate]
+        for qubit, q_t1_ns in t1_ns.items():
+            tphi_err = rb_incoherent_errors[qubit] - microwave_time_ns / (3 * q_t1_ns)
+            q_tphi_ns = 1e10 if tphi_err <= 0 else microwave_time_ns / (3 * tphi_err)
+            tphi_ns[qubit] = q_tphi_ns
+
+    # 3a. Extract Pauli error for single-qubit gates.
+    rb_pauli_errors = _unpack_1q_from_calibration(
+        'single_qubit_rb_pauli_error_per_gate', calibration
+    )
+    gate_pauli_errors = {
+        noise_utils.OpIdentifier(gate, q): pauli_err
+        for q, pauli_err in rb_pauli_errors.items()
+        for gate in google_noise_properties.SINGLE_QUBIT_GATES
+    }
+
+    # 3b. Extract Pauli error for two-qubit gates.
+    gate_prefix_pairs: Dict[Type['cirq.Gate'], str] = {
+        cg_ops.SycamoreGate: 'two_qubit_parallel_sycamore_gate',
+        ops.ISwapPowGate: 'two_qubit_parallel_sqrt_iswap_gate',
+    }
+    for gate, prefix in gate_prefix_pairs.items():
+        pauli_error = _unpack_2q_from_calibration(
+            prefix + '_xeb_pauli_error_per_cycle', calibration
+        )
+        gate_pauli_errors.update(
+            {
+                k: v
+                for qs, pauli_err in pauli_error.items()
+                for k, v in {
+                    noise_utils.OpIdentifier(gate, *qs): pauli_err,
+                    noise_utils.OpIdentifier(gate, *qs[::-1]): pauli_err,
+                }.items()
+            }
+        )
+
+    # 4. Extract readout fidelity for all qubits.
+    p00 = _unpack_1q_from_calibration('single_qubit_p00_error', calibration)
+    p11 = _unpack_1q_from_calibration('single_qubit_p11_error', calibration)
+    readout_errors = {
+        q: np.array([p00.get(q, 0), p11.get(q, 0)]) for q in set(p00.keys()) | set(p11.keys())
+    }
+
+    # 5. Extract entangling angle errors.
+    fsim_errors = {}
+    for gate, prefix in gate_prefix_pairs.items():
+        theta_errors = _unpack_2q_from_calibration(
+            prefix + '_xeb_entangler_theta_error_per_cycle',
+            calibration,
+        )
+        phi_errors = _unpack_2q_from_calibration(
+            prefix + '_xeb_entangler_phi_error_per_cycle',
+            calibration,
+        )
+        angle_keys = set(theta_errors.keys()) | set(phi_errors.keys())
+        for qubits in angle_keys:
+            theta = theta_errors.get(qubits, 0)
+            phi = phi_errors.get(qubits, 0)
+            op_id = noise_utils.OpIdentifier(gate, *qubits)
+            fsim_errors[op_id] = ops.PhasedFSimGate(theta=theta, phi=phi)
+            op_id_reverse = noise_utils.OpIdentifier(gate, *qubits[::-1])
+            fsim_errors[op_id_reverse] = ops.PhasedFSimGate(theta=theta, phi=phi)
+
+    # Known false positive: https://github.com/PyCQA/pylint/issues/5857
+    return google_noise_properties.GoogleNoiseProperties(  # pylint: disable=unexpected-keyword-arg
+        gate_times_ns=default_gate_ns,
+        t1_ns=t1_ns,
+        tphi_ns=tphi_ns,
+        readout_errors=readout_errors,
+        gate_pauli_errors=gate_pauli_errors,
+        fsim_errors=fsim_errors,
+    )