QPUs and Stacks

The two modules iqm.iqm_sim_qpus and iqm.iqm_sim_stacks are being used to call the IQM QPU objects either with or without the inclusion of a transpiler.

The module iqm.iqm_post_stacks is being used to post-process results which have not been obtained in a single pass of the Qaptiva execution stack and which need post-processing, e.g. if LazySynthesis has been used.

IQM QPUs

class iqm.iqm_sim_qpus.IQMSimQPU(use_noisy=True, use_remap=False, sim_method='deterministic', n_samples=100000, bond_dimension=None, fidelity=None, use_GPU=False, use_average_noise_parameters=False, mpo_parameters=None)

Bases: QPUHandler

A base class that provides the basic functions for the IQM QPU’s.

Parameters

use_noisy (Optional[bool]) – Choose between a noisy (True) and noiseless (False) QPU, defaults to True.
use_remap (Optional[bool]) – If set to true, the used qubits of each circuit will be remapped onto an initial segment of integers. Unused qubits are removed from the circuit, defaults to False
sim_method (Optional[str]) – noisy simulation method (deterministic|stochastic|MPO). Defaults to deterministic. MPO is a tensor network method for approximate simulation. Parameter is ignored if use_noisy is False.
n_samples (Optional[int]) – number of samples for stochastic simulation. Defaults to 100000.
bond_dimension (Optional[int]) – The maximum bond dimension size allowed throughout the tensor network simulation. Parameter is ignored if sim_method is not MPO.
fidelity (Optional[float]) – Required precision for the truncation of the density matrix. If set to None, no precision goal will be set for each truncation. If set to a value, the set bond dimension will be allowed to increase to reach this specific truncation fidelity. Parameter is ignored if sim_method is not MPO.
use_GPU (Optional[bool]) – variable to off load simulations to a GPU, primarily useful for stochastic method and ignored for noiseless and MPO simulations, defaults to False
use_average_noise_parameters (Optional[bool]) – variable that determines if average noise parameters (True) for all qubits or qubitwise noise parameters (False) are taken for noise model. Defaults to False
mpo_parameters (Optional[dict]) – Set parameters for the MPO simulator such as the strategy, defaults to None

submit_job(job)

Returns a Result structure corresponding to the execution of a Job.

Note

Note that the job is assumed to be already compiled/optimized

Parameters: job (Job) – the job to execute
Return type: Result

get_specs()

Returns a HardwareSpecs object with the number of qubits, topology and gateset.

Return type: HardwareSpecs

estimate_resources_for_batch(batch)

Placeholder: subclasses must implement this.

QPU classes implement this method to locally compute a ResourceModel which tells what resources they will need to compute a batch.

Parameters: batch (Batch) –

class iqm.iqm_sim_qpus.IQM5SimQPU(use_noisy=True, use_remap=False, sim_method='deterministic', n_samples=100000, bond_dimension=None, fidelity=None, use_GPU=False, use_average_noise_parameters=False, mpo_parameters=None)

Bases: IQMSimQPU

The 5 qubit IQM QPU without any transpiler.

Parameters

use_noisy (Optional[bool]) – Choose between a noisy (True) and noiseless (False) QPU, defaults to True.
use_remap (Optional[bool]) – If set to true, the used qubits of each circuit will be remapped onto an initial segment of integers. Unused qubits are removed from the circuit, defaults to False
sim_method (Optional[str]) – noisy simulation method (deterministic|stochastic|MPO). Defaults to deterministic. MPO is a tensor network method for approximate simulation. Parameter is ignored if use_noisy is False.
n_samples (Optional[int]) – number of samples for stochastic simulation. Defaults to 100000.
bond_dimension (Optional[int]) – The maximum bond dimension size allowed throughout the tensor network simulation. Parameter is ignored if sim_method is not MPO.
fidelity (Optional[float]) – Required precision for the truncation of the density matrix. If set to None, no precision goal will be set for each truncation. If set to a value, the set bond dimension will be allowed to increase to reach this specific truncation fidelity. Parameter is ignored if sim_method is not MPO.
use_GPU (Optional[bool]) – variable to off load simulations to a GPU, primarily useful for stochastic method and ignored for noiseless and MPO simulations, defaults to False
use_average_noise_parameters (Optional[bool]) – variable that determines if average noise parameters (True) for all qubits or qubitwise noise parameters (False) are taken for noise model. Defaults to False
mpo_parameters (Optional[dict]) – Set parameters for the MPO simulator such as the strategy, defaults to None

Example:

We can call the noisy 5-qubit QPU, return the hardware specifications and submit a job via:

from qat.lang import Program
from qat.lang.AQASM import CSIGN
from qat.qpus import IQM5SimQPU
from qat.iqm import PRX
import numpy as np

quantum_program = Program()
qubits = quantum_program.qalloc(3)
quantum_program.apply(PRX(np.pi/2., -np.pi/2.), qubits[0])
quantum_program.apply(PRX(np.pi/2., -np.pi/2.), qubits[2])
quantum_program.apply(CSIGN, qubits[0], qubits[2])
quantum_program.apply(PRX(np.pi/2., -3.*np.pi/2.), qubits[2])

circuit = quantum_program.to_circ()
job = circuit.to_job()
qpu = IQM5SimQPU()
hardware_specs = qpu.get_specs()
result = qpu.submit(job)

class iqm.iqm_sim_qpus.IQM20SimQPU(use_noisy=True, use_remap=False, sim_method='deterministic', n_samples=100000, bond_dimension=None, fidelity=None, use_GPU=False, use_average_noise_parameters=False, mpo_parameters=None)

Bases: IQMSimQPU

The 20 qubit IQM QPU without any transpiler.

Parameters

use_noisy (Optional[bool]) – Choose between a noisy (True) and noiseless (False) QPU, defaults to True.
use_remap (Optional[bool]) – If set to true, the used qubits of each circuit will be remapped onto an initial segment of integers. Unused qubits are removed from the circuit, defaults to False
sim_method (Optional[str]) – noisy simulation method (deterministic|stochastic|MPO). Defaults to deterministic. MPO is a tensor network method for approximate simulation. Parameter is ignored if use_noisy is False.
n_samples (Optional[int]) – number of samples for stochastic simulation. Defaults to 100000.
bond_dimension (Optional[int]) – The maximum bond dimension size allowed throughout the tensor network simulation. Parameter is ignored if sim_method is not MPO.
fidelity (Optional[float]) – Required precision for the truncation of the density matrix. If set to None, no precision goal will be set for each truncation. If set to a value, the set bond dimension will be allowed to increase to reach this specific truncation fidelity. Parameter is ignored if sim_method is not MPO.
use_GPU (Optional[bool]) – variable to off load simulations to a GPU, primarily useful for stochastic method and ignored for noiseless and MPO simulations, defaults to False
use_average_noise_parameters (Optional[bool]) – variable that determines if average noise parameters (True) for all qubits or qubitwise noise parameters (False) are taken for noise model. Defaults to False
mpo_parameters (Optional[dict]) – Set parameters for the MPO simulator such as the strategy, defaults to None

IQM Stacks

class iqm.iqm_sim_stacks.IQM5SimStack(use_noisy=True, use_nnizer=True, use_remap=False, sim_method='deterministic', n_samples=100000, bond_dimension=None, fidelity=None, nnizer_parameters=None, lazy_synthesis_parameters=None, transpiler_method='PRX', decomposition_type='ZXZ', move_rz_to_end=True, use_GPU=False, use_average_noise_parameters=False, mpo_parameters=None)

Bases: QPUHandler

The 5 qubit IQM QPU with a transpiler.

Parameters

use_noisy (Optional[bool]) – Choose between a noisy (True) and noiseless (False) QPU, defaults to True.
use_nnizer (Optional[bool]) – Choose between Nnizer (True) and LazySynthesis (False) as the topology adaption method, defaults to True
use_remap (Optional[bool]) – If set to true, the used qubits of each circuit will be remapped onto an initial segment of integers. Unused qubits are removed from the circuit, defaults to False
sim_method (Optional[str]) – noisy simulation method (deterministic|stochastic|MPO). Defaults to deterministic. MPO is a tensor network method for approximate simulation. Parameter is ignored if use_noisy is False.
n_samples (Optional[int]) – number of samples for stochastic simulation. Defaults to 100000.
bond_dimension (Optional[int]) – The maximum bond dimension size allowed throughout the tensor network simulation. Parameter is ignored if sim_method is not MPO.
fidelity (Optional[float]) – Required precision for the truncation of the density matrix. If set to None, no precision goal will be set for each truncation. If set to a value, the set bond dimension will be allowed to increase to reach this specific truncation fidelity. Parameter is ignored if sim_method is not MPO.
nnizer_parameters (Optional[dict]) – Set parameters for the Nnizer such as the specific method. Default: {“method”: “atos”}
lazy_synthesis_parameters (Optional[dict]) – Set parameters for lazy synthesis such as the search depth. Default: {“depth”: 2}
transpiler_method (Optional[str]) – Choose between the PRX, VZ and NISQ transpiler, defaults to ‘PRX’
decomposition_type (Optional[str]) – Choose a decomposition type for the KAK compression, defaults to ‘ZXZ’ (VZ transpiler only)
move_rz_to_end (Optional[bool]) – Choose whether RZ gates should be collected at the end (default) or beginning of the circuit (VZ transpiler only)
use_GPU (Optional[bool]) – variable to off load simulations to a GPU, primarily useful for stochastic method and ignored for noiseless and MPO simulations, defaults to False
use_average_noise_parameters (Optional[bool]) – variable that determines if average noise parameters (True) for all qubits or qubitwise noise parameters (False) are taken for noise model. Defaults to False
mpo_parameters (Optional[dict]) – Set parameters for the MPO simulator such as the strategy, defaults to None

Example:

Similarly to the QPU, we can call the noisy 5-qubit stack, which includes the transpiler and the QPU:

from qat.lang import Program
from qat.lang.AQASM import H, CNOT
from qat.qpus import IQM5SimStack

quantum_program = Program()
qubits = quantum_program.qalloc(2)
quantum_program.apply(H, qubits[0])
quantum_program.apply(CNOT, qubits[0], qubits[1])

circuit = quantum_program.to_circ()
job = circuit.to_job()
stack = IQM5SimStack()
hardware_specs = stack.get_specs()
result = stack.submit(job)

submit_job(job)

Returns a Result structure corresponding to the execution of a Job.

Note

Note that the job is assumed to be already compiled/optimized

Parameters: job (Job) – the job to execute
Return type: Result

get_specs()

Returns a HardwareSpecs object with the number of qubits, topology and gateset.

Return type: HardwareSpecs

estimate_resources_for_batch(batch)

Placeholder: subclasses must implement this.

QPU classes implement this method to locally compute a ResourceModel which tells what resources they will need to compute a batch.

Parameters: batch (Batch) –

class iqm.iqm_sim_stacks.IQM20SimStack(use_noisy=True, use_nnizer=True, use_remap=False, sim_method='deterministic', n_samples=100000, bond_dimension=None, fidelity=None, nnizer_parameters=None, lazy_synthesis_parameters=None, transpiler_method='PRX', decomposition_type='ZXZ', move_rz_to_end=True, use_GPU=False, use_average_noise_parameters=False, mpo_parameters=None)

Bases: QPUHandler

The 20 qubit IQM QPU with a transpiler.

Parameters

use_noisy (Optional[bool]) – Choose between a noisy (True) and noiseless (False) QPU, defaults to True.
use_nnizer (Optional[bool]) – Choose between Nnizer (True) and LazySynthesis (False) as the topology adaption method, defaults to True
use_remap (Optional[bool]) – If set to true, the used qubits of each circuit will be remapped onto an initial segment of integers. Unused qubits are removed from the circuit, defaults to False
sim_method (Optional[str]) – noisy simulation method (deterministic|stochastic|MPO). Defaults to deterministic. MPO is a tensor network method for approximate simulation. Parameter is ignored if use_noisy is False.
n_samples (Optional[int]) – number of samples for stochastic simulation. Defaults to 100000.
bond_dimension (Optional[int]) – The maximum bond dimension size allowed throughout the tensor network simulation. Parameter is ignored if sim_method is not MPO.
fidelity (Optional[float]) – Required precision for the truncation of the density matrix. If set to None, no precision goal will be set for each truncation. If set to a value, the set bond dimension will be allowed to increase to reach this specific truncation fidelity. Parameter is ignored if sim_method is not MPO.
nnizer_parameters (Optional[dict]) – Set parameters for the Nnizer such as the specific method. Default: {“method”: “atos”}
lazy_synthesis_parameters (Optional[dict]) – Set parameters for lazy synthesis such as the search depth. Default: {“depth”: 2}
transpiler_method (Optional[str]) – Choose between the PRX, VZ and NISQ transpiler, defaults to ‘PRX’
decomposition_type (Optional[str]) – Choose a decomposition type for the KAK compression, defaults to ‘ZXZ’ (VZ transpiler only)
move_rz_to_end (Optional[bool]) – Choose whether RZ gates should be collected at the end (default) or beginning of the circuit (VZ transpiler only)
use_GPU (Optional[bool]) – variable to off load simulations to a GPU, primarily useful for stochastic method and ignored for noiseless and MPO simulations, defaults to False
use_average_noise_parameters (Optional[bool]) – variable that determines if average noise parameters (True) for all qubits or qubitwise noise parameters (False) are taken for noise model. Defaults to False
mpo_parameters (Optional[dict]) – Set parameters for the MPO simulator such as the strategy, defaults to None

submit_job(job)

Returns a Result structure corresponding to the execution of a Job.

Note

Note that the job is assumed to be already compiled/optimized

Parameters: job (Job) – the job to execute
Return type: Result

get_specs()

Returns a HardwareSpecs object with the number of qubits, topology and gateset.

Return type: HardwareSpecs

estimate_resources_for_batch(batch)

Placeholder: subclasses must implement this.

QPU classes implement this method to locally compute a ResourceModel which tells what resources they will need to compute a batch.

Parameters: batch (Batch) –

Postprocessing Stacks

class iqm.iqm_post_stacks.IQM5PostStack(result, use_noisy=True, use_nnizer=True, use_remap=False, sim_method='deterministic', n_samples=100000, bond_dimension=None, fidelity=None, nnizer_parameters=None, lazy_synthesis_parameters=None, transpiler_method='PRX', decomposition_type='ZXZ', move_rz_to_end=True, use_GPU=False, use_average_noise_parameters=False, mpo_parameters=None)

Bases: QPUHandler

A stack for postprocessing results obtained from the 5 qubit IQM QPU outside of a proper Qaptiva execution stack. Note that all parameters have to be the same as used for creating the results which should be postprocessed.

Parameters

result (Result) – The result (obtained outside an execution stack) which should be postprocessed.
use_noisy (Optional[bool]) – Choose between a noisy (True) and noiseless (False) QPU, defaults to True.
use_nnizer (Optional[bool]) – Choose between Nnizer (True) and LazySynthesis (False) as the topology adaption method, defaults to True
use_remap (Optional[bool]) – If set to true, the used qubits of each circuit will be remapped onto an initial segment of integers. Unused qubits are removed from the circuit, defaults to False
sim_method (Optional[str]) – noisy simulation method (deterministic|stochastic|MPO). Defaults to deterministic. MPO is a tensor network method for approximate simulation. Parameter is ignored if use_noisy is False.
n_samples (Optional[int]) – number of samples for stochastic simulation. Defaults to 100000.
bond_dimension (Optional[int]) – The maximum bond dimension size allowed throughout the tensor network simulation. Parameter is ignored if sim_method is not MPO.
fidelity (Optional[float]) – Required precision for the truncation of the density matrix. If set to None, no precision goal will be set for each truncation. If set to a value, the set bond dimension will be allowed to increase to reach this specific truncation fidelity. Parameter is ignored if sim_method is not MPO.
nnizer_parameters (Optional[dict]) – Set parameters for the Nnizer such as the specific method. Default: {“method”: “atos”}
lazy_synthesis_parameters (Optional[dict]) – Set parameters for lazy synthesis such as the search depth. Default: {“depth”: 2}
transpiler_method (Optional[str]) – Choose between the PRX, VZ and NISQ transpiler, defaults to ‘PRX’
decomposition_type (Optional[str]) – Choose a decomposition type for the KAK compression, defaults to ‘ZXZ’ (VZ transpiler only)
move_rz_to_end (Optional[bool]) – Choose whether RZ gates should be collected at the end (default) or beginning of the circuit (VZ transpiler only)
use_GPU (Optional[bool]) – variable to off load simulations to a GPU, primarily useful for stochastic method and ignored for noiseless and MPO simulations, defaults to False
use_average_noise_parameters (Optional[bool]) – variable that determines if average noise parameters (True) for all qubits or qubitwise noise parameters (False) are taken for noise model. Defaults to False
mpo_parameters (Optional[dict]) – Set parameters for the MPO simulator such as the strategy, defaults to None

Example:

Similarly to the IQM5SimStack, we can call the IQM5PostStack to postprocess results which we have obtained when running transpiled code outside of the full IQM5SimStack. Note that the export/import of the result is done here only as an example and might be done differently in real use.

import json
from qat.core import Batch, Result
from qat.lang import Program
from qat.lang.AQASM import H, CNOT
from qat.qpus import get_default_qpu, IQM5SimQPU, IQM5PostStack
from qat.plugins import IQMTranspiler

quantum_program = Program()
qubits = quantum_program.qalloc(5)
quantum_program.apply(H, qubits[0])
quantum_program.apply(CNOT, [qubits[0], qubits[1]])

circ = quantum_program.to_circ()
job = circ.to_job()
batch = Batch(jobs=[job])
specs = IQM5SimQPU().get_specs()
transpiler = IQMTranspiler(use_nnizer=False, lazy_synthesis_parameters={
    'optimize_initial': False})
transpiled_batch = transpiler.compile(batch, specs)

default_qpu = get_default_qpu()
transpiled_batch_result = default_qpu.submit(transpiled_batch)
transpiled_result = transpiled_batch_result.results[0]
results_dict = {}
for sample in transpiled_result:
    results_dict[sample.state.int] = sample.probability
with open("batch_result.json", "w") as f:
    json.dump(results_dict, f)

imported_result = Result()
with open("batch_result.json", "r") as f:
    imported_results_dict = json.load(f)
for entry in imported_results_dict:
    imported_result.add_sample(state = int(entry), probability = imported_results_dict[entry])

stack_pp = IQM5PostStack(result=imported_result, use_nnizer=False, lazy_synthesis_parameters={
    'optimize_initial': False})
stack_pp_result = stack_pp.submit(batch)[0]

submit_job(job)

Returns a Result structure containing the postprocessed version of the Result passed as parameter to the constructor of the IQM5PostprocessingStack

Note

Note that the job is assumed to be already compiled/optimized

Parameters: job (Job) – the job matching the Result to be postprocessed
Return type: Result

get_specs()

Returns a HardwareSpecs object with the number of qubits, topology and gateset.

Return type: HardwareSpecs

estimate_resources_for_batch(batch)

Placeholder: subclasses must implement this.

QPU classes implement this method to locally compute a ResourceModel which tells what resources they will need to compute a batch.

Parameters: batch (Batch) –

class iqm.iqm_post_stacks.IQM20PostStack(result, use_noisy=True, use_nnizer=True, use_remap=False, sim_method='deterministic', n_samples=100000, bond_dimension=None, fidelity=None, nnizer_parameters=None, lazy_synthesis_parameters=None, transpiler_method='PRX', decomposition_type='ZXZ', move_rz_to_end=True, use_GPU=False, use_average_noise_parameters=False, mpo_parameters=None)

Bases: QPUHandler

A stack for postprocessing results obtained from the 20 qubit IQM QPU outside of a proper Qaptiva execution stack. Note that all parameters have to be the same as used for creating the results which should be postprocessed.

Parameters

result (Result) – The result (obtained outside an execution stack) which should be postprocessed.
use_noisy (Optional[bool]) – Choose between a noisy (True) and noiseless (False) QPU, defaults to True.
use_nnizer (Optional[bool]) – Choose between Nnizer (True) and LazySynthesis (False) as the topology adaption method, defaults to True
use_remap (Optional[bool]) – If set to true, the used qubits of each circuit will be remapped onto an initial segment of integers. Unused qubits are removed from the circuit, defaults to False
sim_method (Optional[str]) – noisy simulation method (deterministic|stochastic|MPO). Defaults to deterministic. MPO is a tensor network method for approximate simulation. Parameter is ignored if use_noisy is False.
n_samples (Optional[int]) – number of samples for stochastic simulation. Defaults to 100000.
bond_dimension (Optional[int]) – The maximum bond dimension size allowed throughout the tensor network simulation. Parameter is ignored if sim_method is not MPO.
fidelity (Optional[float]) – Required precision for the truncation of the density matrix. If set to None, no precision goal will be set for each truncation. If set to a value, the set bond dimension will be allowed to increase to reach this specific truncation fidelity. Parameter is ignored if sim_method is not MPO.
nnizer_parameters (Optional[dict]) – Set parameters for the Nnizer such as the specific method. Default: {“method”: “atos”}
lazy_synthesis_parameters (Optional[dict]) – Set parameters for lazy synthesis such as the search depth. Default: {“depth”: 2}
transpiler_method (Optional[str]) – Choose between the PRX, VZ and NISQ transpiler, defaults to ‘PRX’
decomposition_type (Optional[str]) – Choose a decomposition type for the KAK compression, defaults to ‘ZXZ’ (VZ transpiler only)
move_rz_to_end (Optional[bool]) – Choose whether RZ gates should be collected at the end (default) or beginning of the circuit (VZ transpiler only)
use_GPU (Optional[bool]) – variable to off load simulations to a GPU, primarily useful for stochastic method and ignored for noiseless and MPO simulations, defaults to False
use_average_noise_parameters (Optional[bool]) – variable that determines if average noise parameters (True) for all qubits or qubitwise noise parameters (False) are taken for noise model. Defaults to False
mpo_parameters (Optional[dict]) – Set parameters for the MPO simulator such as the strategy, defaults to None

submit_job(job)

Returns a Result structure containing the postprocessed version of the Result passed as parameter to the constructor of the IQM5PostprocessingStack

Note

Note that the job is assumed to be already compiled/optimized

Parameters: job (Job) – the job matching the Result to be postprocessed
Return type: Result

get_specs()

Returns a HardwareSpecs object with the number of qubits, topology and gateset.

Return type: HardwareSpecs

estimate_resources_for_batch(batch)

Placeholder: subclasses must implement this.

QPU classes implement this method to locally compute a ResourceModel which tells what resources they will need to compute a batch.

Parameters: batch (Batch) –