gradient_kwargs is now a positional keyword argument to QNode

doc/development/deprecations.rst

@@ -9,6 +9,12 @@ deprecations are listed below.
 Pending deprecations
 --------------------
 
+* Specifying gradient keyword arguments is deprecated and will be removed in v0.42. 
+  Instead, please specify all arguments through the ``gradient_kwargs`` argument.
+
+  - Deprecated in v0.41
+  - Will be removed in v0.42
+
 * The ``qsvt_legacy`` function has been deprecated.
   Instead, use ``qml.qsvt``. The new functionality takes an input polynomial instead of angles.
 

doc/releases/changelog-dev.md

@@ -10,6 +10,10 @@
 
 <h3>Deprecations 👋</h3>
 
+* Specifying gradient keyword arguments is deprecated and will be removed in v0.42. 
+  Instead, please specify all arguments through the ``gradient_kwargs`` argument.
+  [(#6828)](https://github.com/PennyLaneAI/pennylane/pull/6828)
+
 <h3>Documentation 📝</h3>
 
 * Updated documentation for vibrational Hamiltonians

pennylane/workflow/qnode.py

@@ -285,9 +285,7 @@
             as the name suggests. If not provided,
             the device will determine the best choice automatically. For usage details, please refer to the
             :doc:`dynamic quantum circuits page </introduction/dynamic_quantum_circuits>`.
-
-    Keyword Args:
-        **kwargs: Any additional keyword arguments provided are passed to the differentiation
+        gradient_kwargs (dict): Any additional keyword arguments provided are passed to the differentiation
             method. Please refer to the :mod:`qml.gradients <.gradients>` module for details
             on supported options for your chosen gradient transform.
 
@@ -492,7 +490,7 @@
     """
 
     # pylint: disable=too-many-arguments
     def __init__(
         self,
         func: Callable,
         device: SupportedDeviceAPIs,
@@ -506,8 +504,15 @@
         device_vjp: Union[None, bool] = False,
         postselect_mode: Literal[None, "hw-like", "fill-shots"] = None,
         mcm_method: Literal[None, "deferred", "one-shot", "tree-traversal"] = None,
-        **gradient_kwargs,
+        gradient_kwargs: dict = {},
+        **kwargs,
     ):
+        if kwargs:
+            warnings.warn(
+                f"Specifying gradient keyword arguments {list(kwargs.keys())} is deprecated and will be removed in v0.42. Instead, please specify all arguments in the gradient_kwargs argument.",
+                qml.PennyLaneDeprecationWarning,
+            )
+            gradient_kwargs = kwargs
 
         if logger.isEnabledFor(logging.DEBUG):
             logger.debug(

tests/test_qnode.py

@@ -162,7 +162,9 @@ def test_expansion_strategy_error(self):
 
         with pytest.raises(ValueError, match=r"'expansion_strategy' is no longer"):
 
-            @qml.qnode(qml.device("default.qubit"), expansion_strategy="device")
+            @qml.qnode(
+                qml.device("default.qubit"), gradient_kwargs={"expansion_strategy": "device"}
+            )
             def _():
                 return qml.state()
 
@@ -171,7 +173,7 @@ def test_max_expansion_error(self):
 
         with pytest.raises(ValueError, match="'max_expansion' is no longer a valid"):
 
-            @qml.qnode(qml.device("default.qubit"), max_expansion=1)
+            @qml.qnode(qml.device("default.qubit"), gradient_kwargs={"max_expansion": 1})
             def _():
                 qml.state()
 
@@ -425,7 +427,7 @@ def test_unrecognized_kwargs_raise_warning(self):
 
         with warnings.catch_warnings(record=True) as w:
 
-            @qml.qnode(dev, random_kwarg=qml.gradients.finite_diff)
+            @qml.qnode(dev, gradient_kwargs={"random_kwarg": qml.gradients.finite_diff})
             def circuit(params):
                 qml.RX(params[0], wires=0)
                 return qml.expval(qml.PauliZ(0)), qml.var(qml.PauliZ(0))
@@ -440,16 +442,19 @@ def test_incorrect_diff_method_kwargs_raise_warning(self):
         dev = qml.device("default.qubit", wires=2)
 
         with warnings.catch_warnings(record=True) as w:
+            with pytest.warns(
+                qml.PennyLaneDeprecationWarning, match="Specifying gradient keyword arguments"
+            ):
 
-            @qml.qnode(dev, grad_method=qml.gradients.finite_diff)
-            def circuit0(params):
-                qml.RX(params[0], wires=0)
-                return qml.expval(qml.PauliZ(0)), qml.var(qml.PauliZ(0))
+                @qml.qnode(dev, grad_method=qml.gradients.finite_diff)
+                def circuit0(params):
+                    qml.RX(params[0], wires=0)
+                    return qml.expval(qml.PauliZ(0)), qml.var(qml.PauliZ(0))
 
-            @qml.qnode(dev, gradient_fn=qml.gradients.finite_diff)
-            def circuit2(params):
-                qml.RX(params[0], wires=0)
-                return qml.expval(qml.PauliZ(0)), qml.var(qml.PauliZ(0))
+                @qml.qnode(dev, gradient_fn=qml.gradients.finite_diff)
+                def circuit2(params):
+                    qml.RX(params[0], wires=0)
+                    return qml.expval(qml.PauliZ(0)), qml.var(qml.PauliZ(0))
 
         assert len(w) == 2
         assert "Use diff_method instead" in str(w[0].message)
@@ -858,7 +863,7 @@ def test_single_expectation_value_with_argnum_one(self, diff_method, tol):
         y = pnp.array(-0.654, requires_grad=True)
 
         @qnode(
-            dev, diff_method=diff_method, argnum=[1]
+            dev, diff_method=diff_method, gradient_kwargs={"argnum": [1]}
         )  # <--- we only choose one trainable parameter
         def circuit(x, y):
             qml.RX(x, wires=[0])
@@ -1307,11 +1312,11 @@ def ansatz0():
             return qml.expval(qml.X(0))
 
         with pytest.raises(ValueError, match="'shots' is not a valid gradient_kwarg."):
-            qml.QNode(ansatz0, dev, shots=100)
+            qml.QNode(ansatz0, dev, gradient_kwargs={"shots": 100})
 
         with pytest.raises(ValueError, match="'shots' is not a valid gradient_kwarg."):
 
-            @qml.qnode(dev, shots=100)
+            @qml.qnode(dev, gradient_kwargs={"shots": 100})
             def _():
                 return qml.expval(qml.X(0))
 

tests/test_qnode_legacy.py

@@ -372,7 +372,7 @@ def test_unrecognized_kwargs_raise_warning(self):
 
         with warnings.catch_warnings(record=True) as w:
 
-            @qml.qnode(dev, random_kwarg=qml.gradients.finite_diff)
+            @qml.qnode(dev, gradient_kwargs={"random_kwarg": qml.gradients.finite_diff})
             def circuit(params):
                 qml.RX(params[0], wires=0)
                 return qml.expval(qml.PauliZ(0)), qml.var(qml.PauliZ(0))
@@ -387,16 +387,19 @@ def test_incorrect_diff_method_kwargs_raise_warning(self):
         dev = DefaultQubitLegacy(wires=2)
 
         with warnings.catch_warnings(record=True) as w:
+            with pytest.warns(
+                qml.PennyLaneDeprecationWarning, match="Specifying gradient keyword arguments"
+            ):
 
-            @qml.qnode(dev, grad_method=qml.gradients.finite_diff)
-            def circuit0(params):
-                qml.RX(params[0], wires=0)
-                return qml.expval(qml.PauliZ(0)), qml.var(qml.PauliZ(0))
+                @qml.qnode(dev, grad_method=qml.gradients.finite_diff)
+                def circuit0(params):
+                    qml.RX(params[0], wires=0)
+                    return qml.expval(qml.PauliZ(0)), qml.var(qml.PauliZ(0))
 
-            @qml.qnode(dev, gradient_fn=qml.gradients.finite_diff)
-            def circuit2(params):
-                qml.RX(params[0], wires=0)
-                return qml.expval(qml.PauliZ(0)), qml.var(qml.PauliZ(0))
+                @qml.qnode(dev, gradient_fn=qml.gradients.finite_diff)
+                def circuit2(params):
+                    qml.RX(params[0], wires=0)
+                    return qml.expval(qml.PauliZ(0)), qml.var(qml.PauliZ(0))
 
         assert len(w) == 2
         assert "Use diff_method instead" in str(w[0].message)
@@ -798,7 +801,7 @@ def test_single_expectation_value_with_argnum_one(self, diff_method, tol):
         y = pnp.array(-0.654, requires_grad=True)
 
         @qnode(
-            dev, diff_method=diff_method, argnum=[1]
+            dev, diff_method=diff_method, gradient_kwargs={"argnum": [1]}
         )  # <--- we only choose one trainable parameter
         def circuit(x, y):
             qml.RX(x, wires=[0])