Bugfix - conversions in _copy_and_shift_params (#4477)

* Comment qml.math.cast_like.

* Cast to float64 _copy_and_shift_params.

* convert/cast new_params[p_idx]

* Cast new_params to float64.

* Add test for qml.gradients.param_shift when inputting integer parameters.

* Fix import_should_record_backprop.

* Fix cast and update _copy_and_shift_params.

* Unwrap before casting.

* Skip autograd astype.

* Revert general_shift_rules.py to master.

* Skip conversion for integral types.

* Test more types in test_integer_parameters.

* Update changelog.

* Update tests/gradients/parameter_shift/test_parameter_shift.py

Co-authored-by: Christina Lee <christina@xanadu.ai>

* Update pennylane/gradients/general_shift_rules.py

Co-authored-by: Christina Lee <christina@xanadu.ai>

* Reformat

* Fix TF test.

* Revert to numbers.Integral solution.

* no cover import_should_record_backprop

---------

Co-authored-by: Christina Lee <christina@xanadu.ai>

doc/releases/changelog-dev.md

@@ -351,6 +351,9 @@ array([False, False])
 
 <h3>Bug fixes 🐛</h3>
 
+* `_copy_and_shift_params` does not cast or convert integral types, just relying on `+` and `*`'s casting rules in this case.
+  [(#4477)](https://github.com/PennyLaneAI/pennylane/pull/4477)
+
 * `qml.Projector` is pickle-able again.
   [(#4452)](https://github.com/PennyLaneAI/pennylane/pull/4452)
 

pennylane/gradients/general_shift_rules.py

@@ -16,6 +16,7 @@
 shifted parameters."""
 import functools
 import itertools
+import numbers
 import warnings
 
 import numpy as np
@@ -396,10 +397,11 @@ def _copy_and_shift_params(tape, indices, shifts, multipliers, cast=False):
 
         # Shift copied parameter
         new_params = list(op.data)
-        multiplier = qml.math.convert_like(multiplier, new_params[p_idx])
-        multiplier = qml.math.cast_like(multiplier, new_params[p_idx])
-        shift = qml.math.convert_like(shift, new_params[p_idx])
-        shift = qml.math.cast_like(shift, new_params[p_idx])
+        if not isinstance(new_params[p_idx], numbers.Integral):
+            multiplier = qml.math.convert_like(multiplier, new_params[p_idx])
+            multiplier = qml.math.cast_like(multiplier, new_params[p_idx])
+            shift = qml.math.convert_like(shift, new_params[p_idx])
+            shift = qml.math.cast_like(shift, new_params[p_idx])
         new_params[p_idx] = new_params[p_idx] * multiplier
         new_params[p_idx] = new_params[p_idx] + shift
         if cast:

pennylane/math/utils.py

@@ -397,6 +397,22 @@ def function(x):
     return False
 
 
+def import_should_record_backprop():  # pragma: no cover
+    """Return should_record_backprop or an equivalent function from TensorFlow."""
+    import tensorflow.python as tfpy
+
+    if hasattr(tfpy.eager.tape, "should_record_backprop"):
+        from tensorflow.python.eager.tape import should_record_backprop
+    elif hasattr(tfpy.eager.tape, "should_record"):
+        from tensorflow.python.eager.tape import should_record as should_record_backprop
+    elif hasattr(tfpy.eager.record, "should_record_backprop"):
+        from tensorflow.python.eager.record import should_record_backprop
+    else:
+        raise ImportError("Cannot import should_record_backprop from TensorFlow.")
+
+    return should_record_backprop
+
+
 def requires_grad(tensor, interface=None):
     """Returns True if the tensor is considered trainable.
 
@@ -454,14 +470,7 @@ def requires_grad(tensor, interface=None):
     if interface == "tensorflow":
         import tensorflow as tf
 
-        try:
-            try:
-                from tensorflow.python.eager.record import should_record_backprop
-            except ImportError:  # pragma: no cover
-                from tensorflow.python.eager.tape import should_record_backprop
-        except ImportError:  # pragma: no cover
-            from tensorflow.python.eager.tape import should_record as should_record_backprop
-
+        should_record_backprop = import_should_record_backprop()
         return should_record_backprop([tf.convert_to_tensor(tensor)])
 
     if interface == "autograd":
@@ -510,14 +519,7 @@ def in_backprop(tensor, interface=None):
     if interface == "tensorflow":
         import tensorflow as tf
 
-        try:
-            try:
-                from tensorflow.python.eager.record import should_record_backprop
-            except ImportError:  # pragma: no cover
-                from tensorflow.python.eager.tape import should_record_backprop
-        except ImportError:  # pragma: no cover
-            from tensorflow.python.eager.tape import should_record as should_record_backprop
-
+        should_record_backprop = import_should_record_backprop()
         return should_record_backprop([tf.convert_to_tensor(tensor)])
 
     if interface == "autograd":

tests/gradients/parameter_shift/test_parameter_shift.py

@@ -1567,6 +1567,24 @@ def test_single_expectation_value(self, tol):
         assert np.allclose(res[0], expected[0], atol=tol, rtol=0)
         assert np.allclose(res[1], expected[1], atol=tol, rtol=0)
 
+    @pytest.mark.parametrize(
+        "par", [0, 1, 2, 3, np.int8(1), np.int16(1), np.int32(1), np.int64(1)]
+    )  # integers, zero
+    def test_integer_parameters(self, tol, par):
+        """Test that the gradient of the RY gate matches the exact analytic formula."""
+        dev = qml.device("default.qubit", wires=2)
+
+        tape = qml.tape.QuantumScript([qml.RY(par, wires=[0])], [qml.expval(qml.PauliX(0))])
+        tape.trainable_params = {0}
+
+        # gradients
+        exact = np.cos(par)
+        gtapes, fn = qml.gradients.param_shift(tape)
+        grad_PS = fn(qml.execute(gtapes, dev, gradient_fn=None))
+
+        # different methods must agree
+        assert np.allclose(grad_PS, exact, atol=tol, rtol=0)
+
     def test_multiple_expectation_values(self, tol):
         """Tests correct output shape and evaluation for a tape
         with multiple expval outputs"""