Primitives

qiskit.primitives

The primitives are computational building blocks to be used in larger applications whose input units, called primitive unified blocs (PUBs), require quantum resources to efficiently produce outputs for.

Currently there are two types of primitives whose abstractions, in their latest versions, are defined by BaseSamplerV2 and BaseEstimatorV2. Samplers are responsible for accepting quantum circuits (or sweeps of values over parameterized circuits) and sampling from their classical output registers. Estimators accept combinations of circuits and observables (or sweeps thereof) to estimate expectation values of the observables.

Qiskit implements a reference implementation for each of these abstractions, StatevectorSampler and StatevectorEstimator.

Overview of EstimatorV2

BaseEstimatorV2 is a primitive that estimates expectation values for provided quantum circuit and observable combinations.

Following construction, an estimator is used by calling its run() method with a list of pubs (Primitive Unified Blocs). Each pub contains three values that, together, define a computation unit of work for the estimator to complete:

• a single QuantumCircuit, possibly parametrized, whose final state we define as $\psi(\theta)$,
• one or more observables (specified as any ObservablesArrayLike, including Pauli, SparsePauliOp, str) that specify which expectation values to estimate, denoted $H_j$, and
• a collection parameter value sets to bind the circuit against, $\theta_k$.

Running an estimator returns a BasePrimitiveJob object, where calling the method result() results in expectation value estimates and metadata for each pub:

The observables and parameter values portion of a pub can be array-valued with arbitrary dimensions, where standard broadcasting rules are applied, so that, in turn, the estimated result for each pub is in general array-valued as well. For more information, please check here.

Here is an example of how an estimator is used.

from qiskit.primitives import StatevectorEstimator as Estimator
from qiskit.circuit.library import RealAmplitudes
from qiskit.quantum_info import SparsePauliOp
 
psi1 = RealAmplitudes(num_qubits=2, reps=2)
psi2 = RealAmplitudes(num_qubits=2, reps=3)
 
H1 = SparsePauliOp.from_list([("II", 1), ("IZ", 2), ("XI", 3)])
H2 = SparsePauliOp.from_list([("IZ", 1)])
H3 = SparsePauliOp.from_list([("ZI", 1), ("ZZ", 1)])
 
theta1 = [0, 1, 1, 2, 3, 5]
theta2 = [0, 1, 1, 2, 3, 5, 8, 13]
theta3 = [1, 2, 3, 4, 5, 6]
 
estimator = Estimator()
 
# calculate [ <psi1(theta1)|H1|psi1(theta1)> ]
job = estimator.run([(psi1, H1, [theta1])])
job_result = job.result() # It will block until the job finishes.
print(f"The primitive-job finished with result {job_result}")
 
# calculate [ [<psi1(theta1)|H1|psi1(theta1)>,
#              <psi1(theta3)|H3|psi1(theta3)>],
#             [<psi2(theta2)|H2|psi2(theta2)>] ]
job2 = estimator.run(
    [
        (psi1, [H1, H3], [theta1, theta3]),
        (psi2, H2, theta2)
    ],
    precision=0.01
)
job_result = job2.result()
print(f"The primitive-job finished with result {job_result}")

Overview of SamplerV2

BaseSamplerV2 is a primitive that samples outputs of quantum circuits.

Following construction, a sampler is used by calling its run() method with a list of pubs (Primitive Unified Blocs). Each pub contains values that, together, define a computational unit of work for the sampler to complete:

• A single QuantumCircuit, possibly parameterized.
• A collection parameter value sets to bind the circuit against if it is parametric.
• Optionally, the number of shots to sample, determined in the run method if not set.

Running an estimator returns a BasePrimitiveJob object, where calling the method result() results in output samples and metadata for each pub.

Here is an example of how a sampler is used.

from qiskit.primitives import StatevectorSampler as Sampler
from qiskit import QuantumCircuit
from qiskit.circuit.library import RealAmplitudes
 
# create a Bell circuit
bell = QuantumCircuit(2)
bell.h(0)
bell.cx(0, 1)
bell.measure_all()
 
# create two parameterized circuits
pqc = RealAmplitudes(num_qubits=2, reps=2)
pqc.measure_all()
pqc2 = RealAmplitudes(num_qubits=2, reps=3)
pqc2.measure_all()
 
theta1 = [0, 1, 1, 2, 3, 5]
theta2 = [0, 1, 2, 3, 4, 5, 6, 7]
 
# initialization of the sampler
sampler = Sampler()
 
# collect 128 shots from the Bell circuit
job = sampler.run([bell], shots=128)
job_result = job.result()
print(f"The primitive-job finished with result {job_result}")
 
# run a sampler job on the parameterized circuits
job2 = sampler.run([(pqc, theta1), (pqc2, theta2)])
job_result = job2.result()
print(f"The primitive-job finished with result {job_result}")

Overview of EstimatorV1

Estimator class estimates expectation values of quantum circuits and observables.

An estimator is initialized with an empty parameter set. The estimator is used to create a JobV1, via the qiskit.primitives.Estimator.run() method. This method is called with the following parameters

• quantum circuits ($\psi_i(\theta)$): list of (parameterized) quantum circuits (a list of QuantumCircuit objects).
• observables ($H_j$): a list of SparsePauliOp objects.
• parameter values ($\theta_k$): list of sets of values to be bound to the parameters of the quantum circuits (list of list of float).

The method returns a JobV1 object, calling qiskit.providers.JobV1.result() yields the a list of expectation values plus optional metadata like confidence intervals for the estimation.

Here is an example of how the estimator is used.

from qiskit.primitives import Estimator
from qiskit.circuit.library import RealAmplitudes
from qiskit.quantum_info import SparsePauliOp
 
psi1 = RealAmplitudes(num_qubits=2, reps=2)
psi2 = RealAmplitudes(num_qubits=2, reps=3)
 
H1 = SparsePauliOp.from_list([("II", 1), ("IZ", 2), ("XI", 3)])
H2 = SparsePauliOp.from_list([("IZ", 1)])
H3 = SparsePauliOp.from_list([("ZI", 1), ("ZZ", 1)])
 
theta1 = [0, 1, 1, 2, 3, 5]
theta2 = [0, 1, 1, 2, 3, 5, 8, 13]
theta3 = [1, 2, 3, 4, 5, 6]
 
estimator = Estimator()
 
# calculate [ <psi1(theta1)|H1|psi1(theta1)> ]
job = estimator.run([psi1], [H1], [theta1])
job_result = job.result() # It will block until the job finishes.
print(f"The primitive-job finished with result {job_result}")
 
# calculate [ <psi1(theta1)|H1|psi1(theta1)>,
#             <psi2(theta2)|H2|psi2(theta2)>,
#             <psi1(theta3)|H3|psi1(theta3)> ]
job2 = estimator.run(
    [psi1, psi2, psi1],
    [H1, H2, H3],
    [theta1, theta2, theta3]
)
job_result = job2.result()
print(f"The primitive-job finished with result {job_result}")

Overview of SamplerV1

Sampler class calculates probabilities or quasi-probabilities of bitstrings from quantum circuits.

A sampler is initialized with an empty parameter set. The sampler is used to create a JobV1, via the qiskit.primitives.Sampler.run() method. This method is called with the following parameters

• quantum circuits ($\psi_i(\theta)$): list of (parameterized) quantum circuits. (a list of QuantumCircuit objects)
• parameter values ($\theta_k$): list of sets of parameter values to be bound to the parameters of the quantum circuits. (list of list of float)

The method returns a JobV1 object, calling qiskit.providers.JobV1.result() yields a SamplerResult object, which contains probabilities or quasi-probabilities of bitstrings, plus optional metadata like error bars in the samples.

Here is an example of how sampler is used.

from qiskit.primitives import Sampler
from qiskit import QuantumCircuit
from qiskit.circuit.library import RealAmplitudes
 
# a Bell circuit
bell = QuantumCircuit(2)
bell.h(0)
bell.cx(0, 1)
bell.measure_all()
 
# two parameterized circuits
pqc = RealAmplitudes(num_qubits=2, reps=2)
pqc.measure_all()
pqc2 = RealAmplitudes(num_qubits=2, reps=3)
pqc2.measure_all()
 
theta1 = [0, 1, 1, 2, 3, 5]
theta2 = [0, 1, 2, 3, 4, 5, 6, 7]
 
# initialization of the sampler
sampler = Sampler()
 
# Sampler runs a job on the Bell circuit
job = sampler.run(
    circuits=[bell], parameter_values=[[]], parameters=[[]]
)
job_result = job.result()
print([q.binary_probabilities() for q in job_result.quasi_dists])
 
# Sampler runs a job on the parameterized circuits
job2 = sampler.run(
    circuits=[pqc, pqc2],
    parameter_values=[theta1, theta2],
    parameters=[pqc.parameters, pqc2.parameters])
job_result = job2.result()
print([q.binary_probabilities() for q in job_result.quasi_dists])

Migration from Primitives V1 to V2

The formal distinction between the Primitives V1 and V2 APIs are the base classes from which primitives implementations inherit, which are all listed at the bottom of the page. At a conceptual level, however, here are some notable differences keep in mind when migrating from V1 to V2:

1. The V2 primitives favour vectorized inputs, where single circuits can be grouped with vector-valued (or more generally, array-valued) specifications. Each group is called a primitive unified bloc (pub), and each pub gets its own result. For example, in the estimator, you can compare the following differences:

   # Favoured V2 pattern. There is only one pub here, but there could be more.
   job = estimator_v2.run([(circuit, [obs1, obs2, obs3, obs4])])
   evs = job.result()[0].data.evs
    
   # V1 equivalent, where the same circuit must be provided four times.
   job = estimator_v1.run([circuit] * 4, [obs1, obs2, obs3, obs4])
   evs = job.result().values

   Not shown in the above example, for brevity, is that the circuit can be parametric, with arrays of parameter value sets broadcasted against the array of observables. The sampler is similar, but with no observables:

   # Favoured V2 pattern. There is only one pub here, but there could be more.
   job = sampler_v2.run([(circuit, [vals1, vals2, vals3])])
   samples = job.result()[0].data
    
   # V1 equivalent, where the same circuit must be provided three times.
   sampler_v1.run([circuit] * 3, [vals1, vals2, vals3])
   quasi_dists = job.result().quasi_dists

2. The V2 sampler returns samples of classical outcomes, preserving the shot order in which they were measured. This is in contrast to the V1 sampler that outputs quasi-probability distributions which are instead an estimate of the distribution over classical outcomes. Moreover, the V2 sampler result objects organize data in terms of their input circuits’ classical register names, which provides natural compatibility with dynamic circuits.

   The closest analog of quasi-probability distributions in the V2 interface is the get_counts() method, shown in the example below. However, we emphasize that for utility scale experiments (100+ qubits), the chances of measuring the same bitstring twice are small, so that binning like counts in a dictionary format will not typically be an efficient data processing strategy.

   circuit = QuantumCircuit(QuantumRegister(2, "qreg"), ClassicalRegister(2, "alpha"))
   circuit.h(0)
   circuit.cx(0, 1)
   circuit.measure([0, 1], [0, 1])
    
   # V1 sampler usage
   result = sampler_v1.run([circuit]).result()
   quasi_dist = result.quasi_dists[0]
    
   # V2 sampler usage
   result = sampler_v2.run([circuit]).result()
   # these are the bit values from the alpha register, over all shots
   bitvals = result[0].data.alpha
   # we can use it to generate a Counts mapping, which is similar to a quasi prob distribution
   counts = bitvals.get_counts()
   # which can in turn be converted to the V1 type through normalization
   quasi_dist = QuasiDistribution({outcome: freq / shots for outcome, freq in counts.items()})

3. The V2 primitives have brought the concept of sampling overhead, inherent to all quantum systems via their inherent probabilistic nature, out of the options and into the API itself. For the sampler, this means that the shots argument is now part of the run() signature, and moreover that each pub is able to specify its own value for shots, which takes precedence over any value given to the method. The sampler has an analogous precision argument that specifies the error bars that the primitive implementation should target for expectation values estimates.

   This concept is not present in the API of the V1 primitives, though all implementations of the V1 primitives have related settings somewhere in their options.

   # Sample two circuits at 128 shots each.
   sampler_v2.run([circuit1, circuit2], shots=128)
    
   # Sample two circuits at different amounts of shots. The "None"s are necessary as placeholders
   # for the lack of parameter values in this example.
   sampler_v2.run([(circuit1, None, 123), (circuit2, None, 456)])
    
   # Estimate expectation values for two pubs, both with 0.05 precision.
   estimator_v2.run([(circuit1, obs_array1), (circuit2, obs_array_2)], precision=0.05)

Primitives API

Estimator V2

Sampler V2

Results V2

Estimator V1

Sampler V1