# Copyright 2025 Qilimanjaro Quantum Tech
#
# 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
#
# http://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.
from __future__ import annotations
from abc import ABC
from typing import TYPE_CHECKING, Callable, TypeVar, cast, overload
import numpy as np
from qilisdk.analog import Schedule
from qilisdk.core import QTensor, expect_val, reset_qubits
from qilisdk.digital import Circuit
from qilisdk.functionals.functional_result import FunctionalResult
from qilisdk.functionals.quantum_reservoirs import QuantumReservoir
from qilisdk.functionals.quantum_reservoirs_result import QuantumReservoirResult
from qilisdk.functionals.sampling import Sampling
from qilisdk.functionals.time_evolution import TimeEvolution
from qilisdk.functionals.variational_program import VariationalProgram
from qilisdk.functionals.variational_program_result import VariationalProgramResult
from qilisdk.settings import get_settings
if TYPE_CHECKING:
from qilisdk.core.types import Number
from qilisdk.functionals.functional import Functional, PrimitiveFunctional
from qilisdk.functionals.sampling_result import SamplingResult
from qilisdk.functionals.time_evolution_result import TimeEvolutionResult
[docs]
TResult = TypeVar("TResult", bound=FunctionalResult)
[docs]
class Backend(ABC):
def __init__(self) -> None:
self._handlers: dict[type[Functional], Callable[[Functional], FunctionalResult]] = {
Sampling: lambda f: self._execute_sampling(cast("Sampling", f)),
TimeEvolution: lambda f: self._execute_time_evolution(cast("TimeEvolution", f)),
QuantumReservoir: lambda f: self._execute_quantum_reservoir(cast("QuantumReservoir", f)),
VariationalProgram: lambda f: self._execute_variational_program(cast("VariationalProgram", f)),
}
@overload
[docs]
def execute(self, functional: Sampling) -> SamplingResult: ...
@overload
def execute(self, functional: TimeEvolution) -> TimeEvolutionResult: ...
@overload
def execute(self, functional: VariationalProgram[Sampling]) -> VariationalProgramResult[SamplingResult]: ...
@overload
def execute(
self, functional: VariationalProgram[TimeEvolution]
) -> VariationalProgramResult[TimeEvolutionResult]: ...
@overload
def execute(self, functional: PrimitiveFunctional[TResult]) -> TResult: ...
def execute(self, functional: Functional) -> FunctionalResult:
try:
handler = self._handlers[type(functional)]
except KeyError as exc:
raise NotImplementedError(
f"{type(self).__qualname__} does not support {type(functional).__qualname__}"
) from exc
return handler(functional)
def _execute_sampling(self, functional: Sampling) -> SamplingResult:
raise NotImplementedError(f"{type(self).__qualname__} has no Sampling implementation")
def _execute_time_evolution(self, functional: TimeEvolution) -> TimeEvolutionResult:
raise NotImplementedError(f"{type(self).__qualname__} has no TimeEvolution implementation")
def _execute_quantum_reservoir(self, functional: QuantumReservoir) -> QuantumReservoirResult:
state = functional.initial_state.to_density_matrix()
expected_values: list[list[Number]] = []
intermediate_states: list[QTensor] = []
cache: dict[Circuit, tuple[tuple[float, ...], QTensor]] = {}
for input_dict in functional.input_per_layer:
functional.reservoir_layer.set_parameters(input_dict)
for step in functional.reservoir_layer:
if isinstance(step, Circuit):
param_signature = tuple(step.get_parameter_values())
cached = cache.get(step)
if cached is None or cached[0] != param_signature:
U = step.to_qtensor()
cache[step] = (param_signature, U)
else:
U = cached[1]
state = U @ state @ U.adjoint()
elif isinstance(step, Schedule):
res = self._execute_time_evolution(TimeEvolution(step, [], state, functional.nshots))
if not res.final_state:
raise ValueError("Reservoir Runtime Error: Time Evolution Failed.")
state = res.final_state
try:
state = state.to_density_matrix()
except ValueError as exc:
raise ValueError(
"Reservoir Runtime Error: state repair failed before expectation value computation. "
f"{exc} "
"Try improving simulation precision (e.g., smaller dt, more integrator substeps, or higher precision)."
) from exc
if functional.store_intermediate_states:
intermediate_states.append(state)
expected_values.append(
[expect_val(operator=obs, state=state) for obs in functional.reservoir_layer.observables_as_qtensor]
)
if functional.reservoir_layer.qubits_to_reset:
state = reset_qubits(state, functional.reservoir_layer.qubits_to_reset)
return QuantumReservoirResult(
expected_values=np.array(expected_values),
final_expected_values=np.array(expected_values[-1]),
final_state=state.to_density_matrix() if functional.store_final_state else None,
intermediate_states=intermediate_states if functional.store_intermediate_states else None,
)
def _execute_variational_program(
self, functional: VariationalProgram[PrimitiveFunctional[TResult]]
) -> VariationalProgramResult[TResult]:
"""Optimize a :class:`~qilisdk.functionals.variational_program.VariationalProgram`.
Args:
functional (VariationalProgram): Variational program to optimize.
Returns:
VariationalProgramResult[TResult]: Optimizer output and final functional execution.
Raises:
ValueError: If the wrapped functional has no trainable parameters.
"""
def evaluate_sample(parameters: list[float]) -> float:
param_names = functional.functional.get_parameter_names(where=lambda param: param.is_trainable)
param_bounds = functional.functional.get_parameter_bounds()
new_param_dict = {}
for i, param in enumerate(parameters):
name = param_names[i]
lower_bound, upper_bound = param_bounds[name]
if lower_bound != upper_bound:
new_param_dict[name] = param
err = functional.check_parameter_constraints(new_param_dict)
if err > 0:
return err
functional.functional.set_parameters(new_param_dict)
results = self.execute(functional.functional)
final_results = functional.cost_function.compute_cost(results)
if isinstance(final_results, float):
return final_results
if isinstance(final_results, complex) and abs(final_results.imag) < get_settings().atol:
return final_results.real
raise ValueError(f"Unsupported result type {type(final_results)}.")
if len(functional.functional.get_parameters(where=lambda param: param.is_trainable)) == 0:
raise ValueError("Functional provided does not contain trainable parameters.")
optimizer_result = functional.optimizer.optimize(
cost_function=evaluate_sample,
init_parameters=list(functional.functional.get_parameters(where=lambda param: param.is_trainable).values()),
bounds=list(functional.functional.get_parameter_bounds(where=lambda param: param.is_trainable).values()),
store_intermediate_results=functional.store_intermediate_results,
)
param_names = functional.functional.get_parameter_names(where=lambda param: param.is_trainable)
optimal_parameter_dict = {param_names[i]: param for i, param in enumerate(optimizer_result.optimal_parameters)}
err = functional.check_parameter_constraints(optimal_parameter_dict)
if err > 0:
raise ValueError(
"Optimizer Failed at finding an optimal solution. Check the parameter constraints or try with a different optimization method."
)
functional.functional.set_parameters(optimal_parameter_dict)
optimal_results: TResult = self.execute(functional.functional)
return VariationalProgramResult(optimizer_result=optimizer_result, result=optimal_results)
