Added max_circuits_per_job and removed deepcopy dependency of the qua…

…ntum kernel trainer fixing #701 and #600 (#772)

* Added an option for num_circuits per job for kernels to fix #701

* Updated documentation and format the style.

* Removed deepcopy dependency in quantum_kernel_trainer.py

* Added release notes

* quick fix for spell test

* Added unit tests for max_circuits_per_job

* Update fix-701-max_circuits_per_job-and-600-deepcopy-dependency-e6eda2e5b986c1be.yaml

Small release note bugfix

* Update fix-701-max_circuits_per_job-and-600-deepcopy-dependency-e6eda2e5b986c1be.yaml

* Minor modifications for the unit test

* Removed copy of TrainableKernel

---------

Co-authored-by: oscar-wallis <[email protected]>
(cherry picked from commit 2f49e9e)

qiskit_machine_learning/kernels/algorithms/quantum_kernel_trainer.py

@@ -1,6 +1,6 @@
 # This code is part of a Qiskit project.
 #
-# (C) Copyright IBM 2021, 2023.
+# (C) Copyright IBM 2021, 2024.
 #
 # This code is licensed under the Apache License, Version 2.0. You may
 # obtain a copy of this license in the LICENSE.txt file in the root directory
@@ -13,7 +13,6 @@
 """Quantum Kernel Trainer"""
 from __future__ import annotations
 
-import copy
 from functools import partial
 from typing import Sequence
 
@@ -96,7 +95,8 @@ def __init__(
     ):
         """
         Args:
-            quantum_kernel: a trainable quantum kernel to be trained.
+            quantum_kernel: a trainable quantum kernel to be trained. The
+                :attr:`~.TrainableKernel.parameter_values` will be modified in place after the training.
             loss: A loss function available via string is "svc_loss" which is the same as
                 :class:`~qiskit_machine_learning.utils.loss_functions.SVCLoss`. If a string is
                 passed as the loss function, then the underlying
@@ -179,7 +179,7 @@ def fit(
     ) -> QuantumKernelTrainerResult:
         """
         Train the QuantumKernel by minimizing loss over the kernel parameters. The input
-        quantum kernel will not be altered, and an optimized quantum kernel will be returned.
+        quantum kernel will be altered.
 
         Args:
             data (numpy.ndarray): ``(N, D)`` array of training data, where ``N`` is the
@@ -198,9 +198,6 @@ def fit(
             msg = "Quantum kernel cannot be fit because there are no user parameters specified."
             raise ValueError(msg)
 
-        # Bind inputs to objective function
-        output_kernel = copy.deepcopy(self._quantum_kernel)
-
         # Randomly initialize the initial point if one was not passed
         if self._initial_point is None:
             self._initial_point = algorithm_globals.random.random(num_params)
@@ -222,11 +219,13 @@ def fit(
         result.optimizer_evals = opt_results.nfev
         result.optimal_value = opt_results.fun
         result.optimal_point = opt_results.x
-        result.optimal_parameters = dict(zip(output_kernel.training_parameters, opt_results.x))
+        result.optimal_parameters = dict(
+            zip(self.quantum_kernel.training_parameters, opt_results.x)
+        )
 
         # Return the QuantumKernel in optimized state
-        output_kernel.assign_training_parameters(result.optimal_parameters)
-        result.quantum_kernel = output_kernel
+        self.quantum_kernel.assign_training_parameters(result.optimal_parameters)
+        result.quantum_kernel = self.quantum_kernel
 
         return result
 

qiskit_machine_learning/kernels/fidelity_quantum_kernel.py

@@ -1,6 +1,6 @@
 # This code is part of a Qiskit project.
 #
-# (C) Copyright IBM 2022, 2023.
+# (C) Copyright IBM 2022, 2024.
 #
 # This code is licensed under the Apache License, Version 2.0. You may
 # obtain a copy of this license in the LICENSE.txt file in the root directory
@@ -46,6 +46,7 @@ def __init__(
         fidelity: BaseStateFidelity | None = None,
         enforce_psd: bool = True,
         evaluate_duplicates: str = "off_diagonal",
+        max_circuits_per_job: int = None,
     ) -> None:
         """
         Args:
@@ -73,6 +74,8 @@ def __init__(
                     - ``none`` when training the diagonal is set to `1` and if two identical samples
                       are found in the dataset the corresponding matrix element is set to `1`.
                       When inferring, matrix elements for identical samples are set to `1`.
+            max_circuits_per_job: Maximum number of circuits per job for the backend. Please
+               check the backend specifications. Use ``None`` for all entries per job. Default ``None``.
         Raises:
             ValueError: When unsupported value is passed to `evaluate_duplicates`.
         """
@@ -84,10 +87,15 @@ def __init__(
                 f"Unsupported value passed as evaluate_duplicates: {evaluate_duplicates}"
             )
         self._evaluate_duplicates = eval_duplicates
-
         if fidelity is None:
             fidelity = ComputeUncompute(sampler=Sampler())
         self._fidelity = fidelity
+        if max_circuits_per_job is not None:
+            if max_circuits_per_job < 1:
+                raise ValueError(
+                    f"Unsupported value passed as max_circuits_per_job: {max_circuits_per_job}"
+                )
+        self.max_circuits_per_job = max_circuits_per_job
 
     def evaluate(self, x_vec: np.ndarray, y_vec: np.ndarray | None = None) -> np.ndarray:
         x_vec, y_vec = self._validate_input(x_vec, y_vec)
@@ -208,17 +216,38 @@ def _get_kernel_entries(
         back from the async job.
         """
         num_circuits = left_parameters.shape[0]
+        kernel_entries = []
+        # Check if it is trivial case, only identical samples
         if num_circuits != 0:
-            job = self._fidelity.run(
-                [self._feature_map] * num_circuits,
-                [self._feature_map] * num_circuits,
-                left_parameters,
-                right_parameters,
-            )
-            kernel_entries = job.result().fidelities
-        else:
-            # trivial case, only identical samples
-            kernel_entries = []
+            if self.max_circuits_per_job is None:
+                job = self._fidelity.run(
+                    [self._feature_map] * num_circuits,
+                    [self._feature_map] * num_circuits,
+                    left_parameters,
+                    right_parameters,
+                )
+                kernel_entries = job.result().fidelities
+            else:
+                # Determine the number of chunks needed
+                num_chunks = (
+                    num_circuits + self.max_circuits_per_job - 1
+                ) // self.max_circuits_per_job
+                for i in range(num_chunks):
+                    # Determine the range of indices for this chunk
+                    start_idx = i * self.max_circuits_per_job
+                    end_idx = min((i + 1) * self.max_circuits_per_job, num_circuits)
+                    # Extract the parameters for this chunk
+                    chunk_left_parameters = left_parameters[start_idx:end_idx]
+                    chunk_right_parameters = right_parameters[start_idx:end_idx]
+                    # Execute this chunk
+                    job = self._fidelity.run(
+                        [self._feature_map] * (end_idx - start_idx),
+                        [self._feature_map] * (end_idx - start_idx),
+                        chunk_left_parameters,
+                        chunk_right_parameters,
+                    )
+                    # Extend the kernel_entries list with the results from this chunk
+                    kernel_entries.extend(job.result().fidelities)
         return kernel_entries
 
     def _is_trivial(

releasenotes/notes/fix-701-max_circuits_per_job-and-600-deepcopy-dependency-e6eda2e5b986c1be.yaml

@@ -0,0 +1,14 @@
+---
+fixes:
+  - |
+    Added a `max_circuits_per_job` parameter to the :class:`.FidelityQuantumKernel` used
+    in the case that if more circuits are submitted than the job limit for the 
+    backend, the circuits are split up and run through separate jobs.
+  - |    
+    Removed :class:`.QuantumKernelTrainer` dependency on `copy.deepcopy` that was
+    throwing an error with real backends. Now, it modifies the :class:`.TrainableKernel`
+    in place. If you would like to use the initial kernel, please call 
+    :meth:`~.TrainableKernel.assign_training_parameters` of the :class:`~.TrainableKernel`
+    using the :attr:`~.QuantumKernelTrainer.initial_point` attribute of 
+    :class:`~.QuantumKernelTrainer`.
+ 

test/kernels/test_fidelity_qkernel.py

@@ -106,10 +106,27 @@ def test_defaults(self):
 
         self.assertGreaterEqual(score, 0.5)
 
+    def test_max_circuits_per_job(self):
+        """Test max_circuits_per_job parameters."""
+        kernel_all = FidelityQuantumKernel(feature_map=self.feature_map, max_circuits_per_job=None)
+        kernel_matrix_all = kernel_all.evaluate(x_vec=self.sample_train)
+        with self.subTest("Check when max_circuits_per_job > left_parameters"):
+            kernel_more = FidelityQuantumKernel(
+                feature_map=self.feature_map, max_circuits_per_job=20
+            )
+            kernel_matrix_more = kernel_more.evaluate(x_vec=self.sample_train)
+            np.testing.assert_equal(kernel_matrix_all, kernel_matrix_more)
+        with self.subTest("Check when max_circuits_per_job = 1"):
+            kernel_1 = FidelityQuantumKernel(feature_map=self.feature_map, max_circuits_per_job=1)
+            kernel_matrix_1 = kernel_1.evaluate(x_vec=self.sample_train)
+            np.testing.assert_equal(kernel_matrix_all, kernel_matrix_1)
+
     def test_exceptions(self):
         """Test quantum kernel raises exceptions and warnings."""
         with self.assertRaises(ValueError, msg="Unsupported value of 'evaluate_duplicates'."):
             _ = FidelityQuantumKernel(evaluate_duplicates="wrong")
+        with self.assertRaises(ValueError, msg="Unsupported value of 'max_circuits_per_job'."):
+            _ = FidelityQuantumKernel(max_circuits_per_job=-1)
 
     @idata(
         # params, fidelity, feature map, enforce_psd, duplicate