Add calibrations with the TemplateOptimization

[eggerdj]

Summary

TemplateOptimization can be used to match templates in a quantum circuit and replace them with more efficient gate decompositions. This PR implements #5751 by providing a library of ZX-based templates and provides a new transpilation pass (named CalibrationAdder) to add the corresponding schedules (i.e. calibrations) generated from a subclass of the new abstract base class CalibrationCreator.

Details and comments

An illustrative example of this functionality can be run through the self contained code

from qiskit import QuantumCircuit, transpile, schedule, IBMQ

from qiskit.transpiler import PassManager
from qiskit.transpiler.passes import TemplateOptimization

# New contributions to the template optimization
from qiskit.transpiler.passes.scheduling import RZXCalibrationBuilder, rzx_templates

IBMQ.load_account()
provider = IBMQ.get_provider(...)

backend = provider.get_backend('ibmq_athens')


qc = QuantumCircuit(2)
qc.cx(0,1)
qc.p(0.12, 1)
qc.cx(0,1)
print('Original circuit:')
print(qc)

pass_ = TemplateOptimization(**rzx_templates(['zz2']))
qc_cz = PassManager(pass_).run(qc)
print('ZX based circuit:')
print(qc_cz)

# Add the calibrations
pass_ = RZXCalibrationBuilder(backend)
cal_qc = PassManager(pass_).run(qc_cz)

# Transpile to the backend basis gates
cal_qct = transpile(cal_qc, backend)
qct = transpile(qc, backend)

# Compare the schedule durations
print('Duration of schedule with the calibration:')
print(schedule(cal_qct, backend).duration)
print('Duration of standard with two CNOT gates:')
print(schedule(qct, backend).duration)

The output of this code (see below) shows that there is a significant reduction in schedule duration when using the native ZX interaction of the backend. In this exmaple, the pulse schedule of the unitary is generated from the instruction schedule map of the backend.

Original circuit:
                          
q_0: ──■───────────────■──
     ┌─┴─┐┌─────────┐┌─┴─┐
q_1: ┤ X ├┤ P(0.12) ├┤ X ├
     └───┘└─────────┘└───┘
ZX based circuit:
                                         ┌────────────┐             »
q_0: ────────────────────────────────────┤1           ├─────────────»
     ┌──────────┐┌──────────┐┌──────────┐│  RZX(0.12) │┌───────────┐»
q_1: ┤ RZ(-π/2) ├┤ RX(-π/2) ├┤ RZ(-π/2) ├┤0           ├┤ RX(-0.12) ├»
     └──────────┘└──────────┘└──────────┘└────────────┘└───────────┘»
«                                                    
«q_0: ───────────────────────────────────────────────
«     ┌──────────┐┌──────────┐┌──────────┐┌─────────┐
«q_1: ┤ RZ(-π/2) ├┤ RX(-π/2) ├┤ RZ(-π/2) ├┤ P(0.12) ├
«     └──────────┘└──────────┘└──────────┘└─────────┘
Duration of schedule with the calibration:
1472
Duration of standard with two CNOT gates:
3712

[lcapelluto]

LGTM. I didn't look closely at the templates, but it seems like those are tested to be the identify, which should be sufficient, yes?

[eggerdj]

Yes, the templates are tested to be the identity. Since these are parametric it requires assigning a value to the parameter. In the tests this value is 0.2 which is not some special value such as pi/2.