Source code for pennylane.workflow.qnode
# Copyright 2018-2021 Xanadu Quantum Technologies Inc.
# 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.
"""
This module contains the QNode class and qnode decorator.
"""
# pylint: disable=too-many-instance-attributes,too-many-arguments,protected-access,unnecessary-lambda-assignment, too-many-branches, too-many-statements
import copy
import functools
import inspect
import warnings
from collections.abc import Sequence
from typing import Union, Optional
import logging
import pennylane as qml
from pennylane import Device
from pennylane.measurements import CountsMP, MidMeasureMP, Shots
from pennylane.tape import QuantumTape, QuantumScript
from .execution import INTERFACE_MAP, SUPPORTED_INTERFACES
logger = logging.getLogger(__name__)
logger.addHandler(logging.NullHandler())
def _convert_to_interface(res, interface):
"""
Recursively convert res to the given interface.
"""
interface = INTERFACE_MAP[interface]
if interface in ["Numpy"]:
return res
if isinstance(res, (list, tuple)):
return type(res)(_convert_to_interface(r, interface) for r in res)
if isinstance(res, dict):
return {k: _convert_to_interface(v, interface) for k, v in res.items()}
return qml.math.asarray(res, like=interface if interface != "tf" else "tensorflow")
# pylint: disable=protected-access
def _get_device_shots(device) -> Shots:
if isinstance(device, Device):
if device._shot_vector:
return Shots(device._raw_shot_sequence)
return Shots(device.shots)
return device.shots
def _make_execution_config(
circuit: Optional["QNode"], diff_method=None
) -> "qml.devices.ExecutionConfig":
if diff_method is None or isinstance(diff_method, str):
_gradient_method = diff_method
else:
_gradient_method = "gradient-transform"
grad_on_execution = getattr(circuit, "execute_kwargs", {}).get("grad_on_execution")
if getattr(circuit, "interface", "") == "jax":
grad_on_execution = False
elif grad_on_execution == "best":
grad_on_execution = None
return qml.devices.ExecutionConfig(
interface=getattr(circuit, "interface", None),
gradient_method=_gradient_method,
grad_on_execution=grad_on_execution,
use_device_jacobian_product=getattr(circuit, "execute_kwargs", {"device_vjp": False})[
"device_vjp"
],
)
def _to_qfunc_output_type(
results: qml.typing.Result, qfunc_output, has_partitioned_shots
) -> qml.typing.Result:
if has_partitioned_shots:
return tuple(_to_qfunc_output_type(r, qfunc_output, False) for r in results)
# Special case of single Measurement in a list
if isinstance(qfunc_output, list) and len(qfunc_output) == 1:
results = [results]
# If the return type is not tuple (list or ndarray) (Autograd and TF backprop removed)
if isinstance(qfunc_output, (tuple, qml.measurements.MeasurementProcess)):
return results
return type(qfunc_output)(results)
[docs]class QNode:
"""Represents a quantum node in the hybrid computational graph.
A *quantum node* contains a :ref:`quantum function <intro_vcirc_qfunc>`
(corresponding to a :ref:`variational circuit <glossary_variational_circuit>`)
and the computational device it is executed on.
The QNode calls the quantum function to construct a :class:`~.QuantumTape` instance representing
the quantum circuit.
Args:
func (callable): a quantum function
device (~.Device): a PennyLane-compatible device
interface (str): The interface that will be used for classical backpropagation.
This affects the types of objects that can be passed to/returned from the QNode. See
``qml.workflow.SUPPORTED_INTERFACES`` for a list of all accepted strings.
* ``"autograd"``: Allows autograd to backpropagate
through the QNode. The QNode accepts default Python types
(floats, ints, lists, tuples, dicts) as well as NumPy array arguments,
and returns NumPy arrays.
* ``"torch"``: Allows PyTorch to backpropagate
through the QNode. The QNode accepts and returns Torch tensors.
* ``"tf"``: Allows TensorFlow in eager mode to backpropagate
through the QNode. The QNode accepts and returns
TensorFlow ``tf.Variable`` and ``tf.tensor`` objects.
* ``"jax"``: Allows JAX to backpropagate
through the QNode. The QNode accepts and returns
JAX ``Array`` objects.
* ``None``: The QNode accepts default Python types
(floats, ints, lists, tuples, dicts) as well as NumPy array arguments,
and returns NumPy arrays. It does not connect to any
machine learning library automatically for backpropagation.
* ``"auto"``: The QNode automatically detects the interface from the input values of
the quantum function.
diff_method (str or .TransformDispatcher): The method of differentiation to use in
the created QNode. Can either be a :class:`~.TransformDispatcher`, which includes all
quantum gradient transforms in the :mod:`qml.gradients <.gradients>` module, or a string. The following
strings are allowed:
* ``"best"``: Best available method. Uses classical backpropagation or the
device directly to compute the gradient if supported, otherwise will use
the analytic parameter-shift rule where possible with finite-difference as a fallback.
* ``"device"``: Queries the device directly for the gradient.
Only allowed on devices that provide their own gradient computation.
* ``"backprop"``: Use classical backpropagation. Only allowed on
simulator devices that are classically end-to-end differentiable,
for example :class:`default.qubit <~.DefaultQubit>`. Note that
the returned QNode can only be used with the machine-learning
framework supported by the device.
* ``"adjoint"``: Uses an `adjoint method <https://arxiv.org/abs/2009.02823>`__ that
reverses through the circuit after a forward pass by iteratively applying the inverse
(adjoint) gate. Only allowed on supported simulator devices such as
:class:`default.qubit <~.DefaultQubit>`.
* ``"parameter-shift"``: Use the analytic parameter-shift
rule for all supported quantum operation arguments, with finite-difference
as a fallback.
* ``"hadamard"``: Use the analytic hadamard gradient test
rule for all supported quantum operation arguments. More info is in the documentation
:func:`qml.gradients.hadamard_grad <.gradients.hadamard_grad>`.
* ``"finite-diff"``: Uses numerical finite-differences for all quantum operation
arguments.
* ``"spsa"``: Uses a simultaneous perturbation of all operation arguments to approximate
the derivative.
* ``None``: QNode cannot be differentiated. Works the same as ``interface=None``.
expansion_strategy (str): The strategy to use when circuit expansions or decompositions
are required.
- ``gradient``: The QNode will attempt to decompose
the internal circuit such that all circuit operations are supported by the gradient
method. Further decompositions required for device execution are performed by the
device prior to circuit execution.
- ``device``: The QNode will attempt to decompose the internal circuit
such that all circuit operations are natively supported by the device.
The ``gradient`` strategy typically results in a reduction in quantum device evaluations
required during optimization, at the expense of an increase in classical preprocessing.
max_expansion (int): The number of times the internal circuit should be expanded when
executed on a device. Expansion occurs when an operation or measurement is not
supported, and results in a gate decomposition. If any operations in the decomposition
remain unsupported by the device, another expansion occurs.
grad_on_execution (bool, str): Whether the gradients should be computed on the execution or not.
Only applies if the device is queried for the gradient; gradient transform
functions available in ``qml.gradients`` are only supported on the backward
pass. The 'best' option chooses automatically between the two options and is default.
cache="auto" (str or bool or dict or Cache): Whether to cache evalulations.
``"auto"`` indicates to cache only when ``max_diff > 1``. This can result in
a reduction in quantum evaluations during higher order gradient computations.
If ``True``, a cache with corresponding ``cachesize`` is created for each batch
execution. If ``False``, no caching is used. You may also pass your own cache
to be used; this can be any object that implements the special methods
``__getitem__()``, ``__setitem__()``, and ``__delitem__()``, such as a dictionary.
cachesize (int): The size of any auto-created caches. Only applies when ``cache=True``.
max_diff (int): If ``diff_method`` is a gradient transform, this option specifies
the maximum number of derivatives to support. Increasing this value allows
for higher order derivatives to be extracted, at the cost of additional
(classical) computational overhead during the backwards pass.
device_vjp (bool): Whether or not to use the device-provided Vector Jacobian Product (VJP).
A value of ``None`` indicates to use it if the device provides it, but use the full jacobian otherwise.
Keyword Args:
**kwargs: 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.
**Example**
QNodes can be created by decorating a quantum function:
>>> dev = qml.device("default.qubit", wires=1)
>>> @qml.qnode(dev)
... def circuit(x):
... qml.RX(x, wires=0)
... return qml.expval(qml.Z(0))
or by instantiating the class directly:
>>> def circuit(x):
... qml.RX(x, wires=0)
... return qml.expval(qml.Z(0))
>>> dev = qml.device("default.qubit", wires=1)
>>> qnode = qml.QNode(circuit, dev)
.. details::
:title: Parameter broadcasting
:href: parameter-broadcasting
QNodes can be executed simultaneously for multiple parameter settings, which is called
*parameter broadcasting* or *parameter batching*.
We start with a simple example and briefly look at the scenarios in which broadcasting is
possible and useful. Finally we give rules and conventions regarding the usage of
broadcasting, together with some more complex examples.
Also see the :class:`~.pennylane.operation.Operator` documentation for implementation
details.
**Example**
Again consider the following ``circuit``:
>>> dev = qml.device("default.qubit", wires=1)
>>> @qml.qnode(dev)
... def circuit(x):
... qml.RX(x, wires=0)
... return qml.expval(qml.Z(0))
If we want to execute it at multiple values ``x``,
we may pass those as a one-dimensional array to the QNode:
>>> x = np.array([np.pi / 6, np.pi * 3 / 4, np.pi * 7 / 6])
>>> circuit(x)
tensor([ 0.8660254 , -0.70710678, -0.8660254 ], requires_grad=True)
The resulting array contains the QNode evaluations at the single values:
>>> [circuit(x_val) for x_val in x]
[tensor(0.8660254, requires_grad=True),
tensor(-0.70710678, requires_grad=True),
tensor(-0.8660254, requires_grad=True)]
In addition to the results being stacked into one ``tensor`` already, the broadcasted
execution actually is performed in one simulation of the quantum circuit, instead of
three sequential simulations.
**Benefits & Supported QNodes**
Parameter broadcasting can be useful to simplify the execution syntax with QNodes. More
importantly though, the simultaneous execution via broadcasting can be significantly
faster than iterating over parameters manually. If we compare the execution time for the
above QNode ``circuit`` between broadcasting and manual iteration for an input size of
``100``, we find a speedup factor of about :math:`30`.
This speedup is a feature of classical simulators, but broadcasting may reduce
the communication overhead for quantum hardware devices as well.
A QNode supports broadcasting if all operators that receive broadcasted parameters do so.
(Operators that are used in the circuit but do not receive broadcasted inputs do not need
to support it.) A list of supporting operators is available in
:obj:`~.pennylane.ops.qubit.attributes.supports_broadcasting`.
Whether or not broadcasting delivers an increased performance will depend on whether the
used device is a classical simulator and natively supports this. The latter can be checked
with the capabilities of the device:
>>> dev.capabilities()["supports_broadcasting"]
True
If a device does not natively support broadcasting, it will execute broadcasted QNode calls
by expanding the input arguments into separate executions. That is, every device can
execute QNodes with broadcasting, but only supporting devices will benefit from it.
**Usage**
The first example above is rather simple. Broadcasting is possible in more complex
scenarios as well, for which it is useful to understand the concept in more detail.
The following rules and conventions apply:
*There is at most one broadcasting axis*
The broadcasted input has (exactly) one more axis than the operator(s) which receive(s)
it would usually expect. For example, most operators expect a single scalar input and the
*broadcasted* input correspondingly is a 1D array:
>>> x = np.array([1., 2., 3.])
>>> op = qml.RX(x, wires=0) # Additional axis of size 3.
An operator ``op`` that supports broadcasting indicates the expected number of
axes--or dimensions--in its attribute ``op.ndim_params``. This attribute is a tuple with
one integer per argument of ``op``. The batch size of a broadcasted operator is stored
in ``op.batch_size``:
>>> op.ndim_params # RX takes one scalar input.
(0,)
>>> op.batch_size # The broadcasting axis has size 3.
3
The broadcasting axis is always the leading axis of an argument passed to an operator:
>>> from scipy.stats import unitary_group
>>> U = np.stack([unitary_group.rvs(4) for _ in range(3)])
>>> U.shape # U stores three two-qubit unitaries, each of shape 4x4
(3, 4, 4)
>>> op = qml.QubitUnitary(U, wires=[0, 1])
>>> op.batch_size
3
Stacking multiple broadcasting axes is *not* supported.
*Multiple operators are broadcasted simultaneously*
It is possible to broadcast multiple parameters simultaneously. In this case, the batch
size of the broadcasting axes must match, and the parameters are combined like in Python's
``zip`` function. Non-broadcasted parameters do not need
to be augmented manually but can simply be used as one would in individual QNode
executions:
.. code-block:: python
dev = qml.device("default.qubit", wires=4)
@qml.qnode(dev)
def circuit(x, y, U):
qml.QubitUnitary(U, wires=[0, 1, 2, 3])
qml.RX(x, wires=0)
qml.RY(y, wires=1)
qml.RX(x, wires=2)
qml.RY(y, wires=3)
return qml.expval(qml.Z(0) @ qml.X(1) @ qml.Z(2) @ qml.Z(3))
x = np.array([0.4, 2.1, -1.3])
y = 2.71
U = np.stack([unitary_group.rvs(16) for _ in range(3)])
This circuit takes three arguments, and the first two are used twice each. ``x`` and
``U`` will lead to a batch size of ``3`` for the ``RX`` rotations and the multi-qubit
unitary, respectively. The input ``y`` is a ``float`` value and will be used together with
all three values in ``x`` and ``U``. We obtain three output values:
>>> circuit(x, y, U)
tensor([-0.06939911, 0.26051235, -0.20361048], requires_grad=True)
This is equivalent to iterating over all broadcasted arguments using ``zip``:
>>> [circuit(x_val, y, U_val) for x_val, U_val in zip(x, U)]
[tensor(-0.06939911, requires_grad=True),
tensor(0.26051235, requires_grad=True),
tensor(-0.20361048, requires_grad=True)]
In the same way it is possible to broadcast multiple arguments of a single operator,
for example:
>>> qml.Rot.ndim_params # Rot takes three scalar arguments
(0, 0, 0)
>>> x = np.array([0.4, 2.3, -0.1]) # Broadcast the first argument with size 3
>>> y = 1.6 # Do not broadcast the second argument
>>> z = np.array([1.2, -0.5, 2.5]) # Broadcast the third argument with size 3
>>> op = qml.Rot(x, y, z, wires=0)
>>> op.batch_size
3
*Broadcasting does not modify classical processing*
Note that classical processing in QNodes will happen *before* broadcasting is taken into
account. This means, that while *operators* always interpret the first axis as the
broadcasting axis, QNodes do not necessarily do so:
.. code-block:: python
@qml.qnode(dev)
def circuit_unpacking(x):
qml.RX(x[0], wires=0)
qml.RY(x[1], wires=1)
qml.RZ(x[2], wires=1)
return qml.expval(qml.Z(0) @ qml.X(1))
x = np.array([[1, 2], [3, 4], [5, 6]])
The prepared parameter ``x`` has shape ``(3, 2)``, corresponding to the three operations
and a batch size of ``2``:
>>> circuit_unpacking(x)
tensor([0.02162852, 0.30239696], requires_grad=True)
If we were to iterate manually over the parameter settings, we probably would put the
batching axis in ``x`` first. This is not the behaviour with parameter broadcasting
because it does not modify the unpacking step within the QNode, so that ``x`` is
unpacked *first* and the unpacked elements are expected to contain the
broadcasted parameters for each operator individually;
if we attempted to put the broadcasting axis of size ``2`` first, the
indexing of ``x`` would fail in the ``RZ`` rotation within the QNode.
"""
def __init__(
self,
func,
device: Union[Device, "qml.devices.Device"],
interface="auto",
diff_method="best",
expansion_strategy="gradient",
max_expansion=10,
grad_on_execution="best",
cache="auto",
cachesize=10000,
max_diff=1,
device_vjp=False,
**gradient_kwargs,
):
if logger.isEnabledFor(logging.DEBUG):
logger.debug(
"""Creating QNode(func=%s, device=%s, interface=%s, diff_method=%s, expansion_strategy=%s, max_expansion=%s, grad_on_execution=%s, cache=%s, cachesize=%s, max_diff=%s, gradient_kwargs=%s""",
(
func
if not (logger.isEnabledFor(qml.logging.TRACE) and inspect.isfunction(func))
else "\n" + inspect.getsource(func)
),
repr(device),
interface,
diff_method,
expansion_strategy,
max_expansion,
grad_on_execution,
cache,
cachesize,
max_diff,
gradient_kwargs,
)
if interface not in SUPPORTED_INTERFACES:
raise qml.QuantumFunctionError(
f"Unknown interface {interface}. Interface must be "
f"one of {SUPPORTED_INTERFACES}."
)
if not isinstance(device, (Device, qml.devices.Device)):
raise qml.QuantumFunctionError(
"Invalid device. Device must be a valid PennyLane device."
)
if "shots" in inspect.signature(func).parameters:
warnings.warn(
"Detected 'shots' as an argument to the given quantum function. "
"The 'shots' argument name is reserved for overriding the number of shots "
"taken by the device. Its use outside of this context should be avoided.",
UserWarning,
)
self._qfunc_uses_shots_arg = True
else:
self._qfunc_uses_shots_arg = False
for kwarg in gradient_kwargs:
if kwarg in ["gradient_fn", "grad_method"]:
warnings.warn(
f"It appears you may be trying to set the method of differentiation via the kwarg "
f"{kwarg}. This is not supported in qnode and will default to backpropogation. Use "
f"diff_method instead."
)
elif kwarg not in qml.gradients.SUPPORTED_GRADIENT_KWARGS:
warnings.warn(
f"Received gradient_kwarg {kwarg}, which is not included in the list of standard qnode "
f"gradient kwargs."
)
# input arguments
self.func = func
self.device = device
self._interface = interface
self.diff_method = diff_method
self.expansion_strategy = expansion_strategy
self.max_expansion = max_expansion
cache = (max_diff > 1) if cache == "auto" else cache
# execution keyword arguments
self.execute_kwargs = {
"grad_on_execution": grad_on_execution,
"cache": cache,
"cachesize": cachesize,
"max_diff": max_diff,
"max_expansion": max_expansion,
"device_vjp": device_vjp,
}
if self.expansion_strategy == "device":
self.execute_kwargs["expand_fn"] = None
# internal data attributes
self._tape = None
self._qfunc_output = None
self._user_gradient_kwargs = gradient_kwargs
self._original_device = device
self.gradient_fn = None
self.gradient_kwargs = {}
self._tape_cached = False
self._transform_program = qml.transforms.core.TransformProgram()
self._update_gradient_fn()
functools.update_wrapper(self, func)
def __copy__(self):
copied_qnode = QNode.__new__(QNode)
for attr, value in vars(self).items():
if attr not in {"execute_kwargs", "_transform_program", "gradient_kwargs"}:
setattr(copied_qnode, attr, value)
copied_qnode.execute_kwargs = dict(self.execute_kwargs)
copied_qnode._transform_program = qml.transforms.core.TransformProgram(
self.transform_program
) # pylint: disable=protected-access
copied_qnode.gradient_kwargs = dict(self.gradient_kwargs)
return copied_qnode
def __repr__(self):
"""String representation."""
if isinstance(self.device, qml.devices.Device):
return f"<QNode: device='{self.device}', interface='{self.interface}', diff_method='{self.diff_method}'>"
detail = "<QNode: wires={}, device='{}', interface='{}', diff_method='{}'>"
return detail.format(
self.device.num_wires,
self.device.short_name,
self.interface,
self.diff_method,
)
@property
def interface(self):
"""The interface used by the QNode"""
return self._interface
@interface.setter
def interface(self, value):
if value not in SUPPORTED_INTERFACES:
raise qml.QuantumFunctionError(
f"Unknown interface {value}. Interface must be one of {SUPPORTED_INTERFACES}."
)
self._interface = INTERFACE_MAP[value]
self._update_gradient_fn(shots=self.device.shots)
@property
def transform_program(self):
"""The transform program used by the QNode."""
return self._transform_program
[docs] def add_transform(self, transform_container):
"""Add a transform (container) to the transform program.
.. warning:: This is a developer facing feature and is called when a transform is applied on a QNode.
"""
self._transform_program.push_back(transform_container=transform_container)
def _update_gradient_fn(self, shots=None, tape=None):
if self.diff_method is None:
self._interface = None
self.gradient_fn = None
self.gradient_kwargs = {}
return
if tape is None and shots:
tape = qml.tape.QuantumScript([], [], shots=shots)
diff_method = self.diff_method
if (
self.device.name == "lightning.qubit"
and qml.metric_tensor in self.transform_program
and self.diff_method == "best"
):
diff_method = "parameter-shift"
self.gradient_fn, self.gradient_kwargs, self.device = self.get_gradient_fn(
self._original_device, self.interface, diff_method, tape=tape
)
self.gradient_kwargs.update(self._user_gradient_kwargs or {})
def _update_original_device(self):
# FIX: If the qnode swapped the device, increase the num_execution value on the original device.
# In the long run, we should make sure that the user's device is the one
# actually run so she has full control. This could be done by changing the class
# of the user's device before and after executing the tape.
if self.device is not self._original_device:
if not self._tape_cached:
self._original_device._num_executions += 1 # pylint: disable=protected-access
# Update for state vector simulators that have the _pre_rotated_state attribute
if hasattr(self._original_device, "_pre_rotated_state"):
self._original_device._pre_rotated_state = self.device._pre_rotated_state
# Update for state vector simulators that have the _state attribute
if hasattr(self._original_device, "_state"):
self._original_device._state = self.device._state
# pylint: disable=too-many-return-statements
[docs] @staticmethod
def get_gradient_fn(
device, interface, diff_method="best", tape: Optional["qml.tape.QuantumTape"] = None
):
"""Determine the best differentiation method, interface, and device
for a requested device, interface, and diff method.
Args:
device (.Device): PennyLane device
interface (str): name of the requested interface
diff_method (str or .TransformDispatcher): The requested method of differentiation.
If a string, allowed options are ``"best"``, ``"backprop"``, ``"adjoint"``,
``"device"``, ``"parameter-shift"``, ``"hadamard"``, ``"finite-diff"``, or ``"spsa"``.
A gradient transform may also be passed here.
tape (Optional[.QuantumTape]): the circuit that will be differentiated. Should include shots information.
Returns:
tuple[str or .TransformDispatcher, dict, .Device: Tuple containing the ``gradient_fn``,
``gradient_kwargs``, and the device to use when calling the execute function.
"""
config = _make_execution_config(None, diff_method)
if isinstance(device, qml.devices.Device):
if device.supports_derivatives(config, circuit=tape):
new_config = device.preprocess(config)[1]
return new_config.gradient_method, {}, device
if diff_method in {"backprop", "adjoint", "device"}: # device-only derivatives
raise qml.QuantumFunctionError(
f"Device {device} does not support {diff_method} with requested circuit."
)
if diff_method == "best":
return QNode.get_best_method(device, interface, tape=tape)
if isinstance(device, qml.Device):
# handled by device.supports_derivatives with new device interface
if diff_method == "backprop":
return QNode._validate_backprop_method(device, interface, tape=tape)
if diff_method == "adjoint":
return QNode._validate_adjoint_method(device)
if diff_method == "device":
return QNode._validate_device_method(device)
if diff_method == "parameter-shift":
return QNode._validate_parameter_shift(device)
if diff_method == "finite-diff":
return qml.gradients.finite_diff, {}, device
if diff_method == "spsa":
return qml.gradients.spsa_grad, {}, device
if diff_method == "hadamard":
return qml.gradients.hadamard_grad, {}, device
if isinstance(diff_method, str):
raise qml.QuantumFunctionError(
f"Differentiation method {diff_method} not recognized. Allowed "
"options are ('best', 'parameter-shift', 'backprop', 'finite-diff', "
"'device', 'adjoint', 'spsa', 'hadamard')."
)
if isinstance(diff_method, qml.transforms.core.TransformDispatcher):
return diff_method, {}, device
raise qml.QuantumFunctionError(
f"Differentiation method {diff_method} must be a gradient transform or a string."
)
[docs] @staticmethod
def get_best_method(device, interface, tape=None):
"""Returns the 'best' differentiation method
for a particular device and interface combination.
This method attempts to determine support for differentiation
methods using the following order:
* ``"device"``
* ``"backprop"``
* ``"parameter-shift"``
* ``"finite-diff"``
The first differentiation method that is supported (going from
top to bottom) will be returned. Note that the SPSA-based and Hadamard-based gradients
are not included here.
Args:
device (.Device): PennyLane device
interface (str): name of the requested interface
shots
Returns:
tuple[str or .TransformDispatcher, dict, .Device: Tuple containing the ``gradient_fn``,
``gradient_kwargs``, and the device to use when calling the execute function.
"""
config = _make_execution_config(None, "best")
if isinstance(device, qml.devices.Device):
if device.supports_derivatives(config, circuit=tape):
new_config = device.preprocess(config)[1]
return new_config.gradient_method, {}, device
return QNode._validate_parameter_shift(device)
try:
return QNode._validate_device_method(device)
except qml.QuantumFunctionError:
try:
return QNode._validate_backprop_method(device, interface, tape=tape)
except qml.QuantumFunctionError:
try:
return QNode._validate_parameter_shift(device)
except qml.QuantumFunctionError:
return qml.gradients.finite_diff, {}, device
[docs] @staticmethod
def best_method_str(device, interface):
"""Similar to :meth:`~.get_best_method`, except return the
'best' differentiation method in human-readable format.
This method attempts to determine support for differentiation
methods using the following order:
* ``"device"``
* ``"backprop"``
* ``"parameter-shift"``
* ``"finite-diff"``
The first differentiation method that is supported (going from
top to bottom) will be returned. Note that the SPSA-based and Hadamard-based gradient
are not included here.
This method is intended only for debugging purposes. Otherwise,
:meth:`~.get_best_method` should be used instead.
Args:
device (.Device): PennyLane device
interface (str): name of the requested interface
Returns:
str: The gradient function to use in human-readable format.
"""
transform = QNode.get_best_method(device, interface)[0]
if transform is qml.gradients.finite_diff:
return "finite-diff"
if transform in (qml.gradients.param_shift, qml.gradients.param_shift_cv):
return "parameter-shift"
# only other options at this point are "backprop" or "device"
return transform
@staticmethod
def _validate_backprop_method(device, interface, tape=None):
if isinstance(device, qml.devices.Device):
raise ValueError(
"QNode._validate_backprop_method only applies to the qml.Device interface."
)
if tape.shots if tape else _get_device_shots(device):
raise qml.QuantumFunctionError("Backpropagation is only supported when shots=None.")
if tape and any(isinstance(m.obs, qml.SparseHamiltonian) for m in tape.measurements):
raise qml.QuantumFunctionError("backprop cannot differentiate a qml.SparseHamiltonian.")
mapped_interface = INTERFACE_MAP.get(interface, interface)
# determine if the device supports backpropagation
backprop_interface = device.capabilities().get("passthru_interface", None)
if backprop_interface is not None:
# device supports backpropagation natively
if mapped_interface == backprop_interface or interface == "auto":
return "backprop", {}, device
raise qml.QuantumFunctionError(
f"Device {device.short_name} only supports diff_method='backprop' when using the "
f"{backprop_interface} interface."
)
# determine if the device has any child devices that support backpropagation
backprop_devices = device.capabilities().get("passthru_devices", None)
if backprop_devices is not None:
# device is analytic and has child devices that support backpropagation natively
if interface == "auto":
return "backprop", {}, device
if mapped_interface in backprop_devices:
# no need to create another device if the child device is the same (e.g., default.mixed)
if backprop_devices[mapped_interface] == device.short_name:
return "backprop", {}, device
# TODO: need a better way of passing existing device init options
# to a new device?
expand_fn = device.expand_fn
batch_transform = device.batch_transform
debugger = device._debugger
tracker = device.tracker
new_device = qml.device(
backprop_devices[mapped_interface], wires=device.wires, shots=device.shots
)
new_device.expand_fn = expand_fn
new_device.batch_transform = batch_transform
new_device._debugger = debugger
new_device.tracker = tracker
return "backprop", {}, new_device
raise qml.QuantumFunctionError(
f"Device {device.short_name} only supports diff_method='backprop' when using the "
f"{list(backprop_devices.keys())} interfaces."
)
raise qml.QuantumFunctionError(
f"The {device.short_name} device does not support native computations with "
"autodifferentiation frameworks."
)
@staticmethod
def _validate_adjoint_method(device):
# The conditions below provide a minimal set of requirements that we can likely improve upon in
# future, or alternatively summarize within a single device capability. Moreover, we also
# need to inspect the circuit measurements to ensure only expectation values are taken. This
# cannot be done here since we don't yet know the composition of the circuit.
if isinstance(device, qml.devices.Device):
raise ValueError(
"QNode._validate_adjoint_method only applies to the qml.Device interface."
)
required_attrs = ["_apply_operation", "_apply_unitary", "adjoint_jacobian"]
supported_device = all(hasattr(device, attr) for attr in required_attrs)
supported_device = supported_device and device.capabilities().get("returns_state")
if not supported_device:
raise ValueError(
f"The {device.short_name} device does not support adjoint differentiation."
)
if device.shots is not None:
warnings.warn(
"Requested adjoint differentiation to be computed with finite shots. "
"Adjoint differentiation always calculated exactly.",
UserWarning,
)
return "device", {"use_device_state": True, "method": "adjoint_jacobian"}, device
@staticmethod
def _validate_device_method(device):
if isinstance(device, qml.devices.Device):
raise ValueError(
"QNode._validate_device_method only applies to the qml.Device interface."
)
# determine if the device provides its own jacobian method
if device.capabilities().get("provides_jacobian", False):
return "device", {}, device
raise qml.QuantumFunctionError(
f"The {device.short_name} device does not provide a native "
"method for computing the jacobian."
)
@staticmethod
def _validate_parameter_shift(device):
if isinstance(device, qml.devices.Device):
return qml.gradients.param_shift, {}, device
model = device.capabilities().get("model", None)
if model in {"qubit", "qutrit"}:
return qml.gradients.param_shift, {}, device
if model == "cv":
return qml.gradients.param_shift_cv, {"dev": device}, device
raise qml.QuantumFunctionError(
f"Device {device.short_name} uses an unknown model ('{model}') "
"that does not support the parameter-shift rule."
)
@property
def tape(self) -> QuantumTape:
"""The quantum tape"""
return self._tape
qtape = tape # for backwards compatibility
[docs] def construct(self, args, kwargs): # pylint: disable=too-many-branches
"""Call the quantum function with a tape context, ensuring the operations get queued."""
kwargs = copy.copy(kwargs)
old_interface = self.interface
if self._qfunc_uses_shots_arg:
shots = _get_device_shots(self._original_device)
else:
shots = kwargs.pop("shots", _get_device_shots(self._original_device))
if old_interface == "auto":
self.interface = qml.math.get_interface(*args, *list(kwargs.values()))
with qml.queuing.AnnotatedQueue() as q:
self._qfunc_output = self.func(*args, **kwargs)
self._tape = QuantumScript.from_queue(q, shots)
params = self.tape.get_parameters(trainable_only=False)
self.tape.trainable_params = qml.math.get_trainable_indices(params)
if any(isinstance(m, CountsMP) for m in self.tape.measurements) and any(
qml.math.is_abstract(a) for a in args
):
raise qml.QuantumFunctionError("Can't JIT a quantum function that returns counts.")
if isinstance(self._qfunc_output, qml.numpy.ndarray):
measurement_processes = tuple(self.tape.measurements)
elif not isinstance(self._qfunc_output, Sequence):
measurement_processes = (self._qfunc_output,)
else:
measurement_processes = self._qfunc_output
if not measurement_processes or not all(
isinstance(m, qml.measurements.MeasurementProcess) for m in measurement_processes
):
raise qml.QuantumFunctionError(
"A quantum function must return either a single measurement, "
"or a nonempty sequence of measurements."
)
terminal_measurements = [
m for m in self.tape.measurements if not isinstance(m, MidMeasureMP)
]
if any(ret is not m for ret, m in zip(measurement_processes, terminal_measurements)):
raise qml.QuantumFunctionError(
"All measurements must be returned in the order they are measured."
)
num_wires = len(self.tape.wires) if not self.device.wires else len(self.device.wires)
for obj in self.tape.operations + self.tape.observables:
if (
getattr(obj, "num_wires", None) is qml.operation.WiresEnum.AllWires
and obj.wires
and len(obj.wires) != num_wires
):
# check here only if enough wires
raise qml.QuantumFunctionError(f"Operator {obj.name} must act on all wires")
# Apply the deferred measurement principle if the device doesn't
# support mid-circuit measurements natively.
# Only apply transform with old device API as postselection with
# broadcasting will split tapes.
expand_mid_measure = (
any(isinstance(op, MidMeasureMP) for op in self.tape.operations)
and not isinstance(self.device, qml.devices.Device)
and not self.device.capabilities().get("supports_mid_measure", False)
)
if expand_mid_measure or self.expansion_strategy == "device":
# Assume that tapes are not split if old device is used since postselection is not supported.
tapes, _ = qml.defer_measurements(self._tape, device=self.device)
self._tape = tapes[0]
if self.expansion_strategy == "device":
if isinstance(self.device, qml.devices.Device):
tape, _ = self.device.preprocess()[0]([self.tape])
if len(tape) != 1:
raise ValueError(
"Using 'device' for the `expansion_strategy` is not supported for batches of tapes"
)
self._tape = tape[0]
else:
self._tape = self.device.expand_fn(self.tape, max_expansion=self.max_expansion)
if old_interface == "auto":
self.interface = "auto"
def _execution_component(self, args: tuple, kwargs: dict, override_shots) -> qml.typing.Result:
"""Construct the transform program and execute the tapes. Helper function for ``__call__``
Args:
args (tuple): the arguments the QNode is called with
kwargs (dict): the keyword arguments the QNode is called with
override_shots : the shots to use for the execution.
Returns:
Result
"""
cache = self.execute_kwargs.get("cache", False)
using_custom_cache = (
hasattr(cache, "__getitem__")
and hasattr(cache, "__setitem__")
and hasattr(cache, "__delitem__")
)
self._tape_cached = using_custom_cache and self.tape.hash in cache
# Add the device program to the QNode program
if isinstance(self.device, qml.devices.Device):
config = _make_execution_config(self, self.gradient_fn)
device_transform_program, config = self.device.preprocess(execution_config=config)
full_transform_program = self.transform_program + device_transform_program
else:
config = None
full_transform_program = qml.transforms.core.TransformProgram(self.transform_program)
has_mcm_support = (
any(isinstance(op, MidMeasureMP) for op in self._tape)
and hasattr(self.device, "capabilities")
and self.device.capabilities().get("supports_mid_measure", False)
)
if has_mcm_support:
full_transform_program.add_transform(qml.dynamic_one_shot)
override_shots = 1
# Add the gradient expand to the program if necessary
if getattr(self.gradient_fn, "expand_transform", False):
full_transform_program.insert_front_transform(
qml.transform(self.gradient_fn.expand_transform),
**self.gradient_kwargs,
)
# Calculate the classical jacobians if necessary
full_transform_program.set_classical_component(self, args, kwargs)
full_transform_program.prune_dynamic_transform()
# pylint: disable=unexpected-keyword-arg
res = qml.execute(
(self._tape,),
device=self.device,
gradient_fn=self.gradient_fn,
interface=self.interface,
transform_program=full_transform_program,
config=config,
gradient_kwargs=self.gradient_kwargs,
override_shots=override_shots,
**self.execute_kwargs,
)
res = res[0]
# convert result to the interface in case the qfunc has no parameters
if (
len(self.tape.get_parameters(trainable_only=False)) == 0
and not self.transform_program.is_informative
):
res = _convert_to_interface(res, self.interface)
return _to_qfunc_output_type(
res, self._qfunc_output, self._tape.shots.has_partitioned_shots
)
[docs] def __call__(self, *args, **kwargs) -> qml.typing.Result:
old_interface = self.interface
if old_interface == "auto":
interface = qml.math.get_interface(*args, *list(kwargs.values()))
self._interface = INTERFACE_MAP[interface]
if self._qfunc_uses_shots_arg:
override_shots = False
else:
if "shots" not in kwargs:
kwargs["shots"] = _get_device_shots(self._original_device)
override_shots = kwargs["shots"]
# construct the tape
self.construct(args, kwargs)
original_grad_fn = [self.gradient_fn, self.gradient_kwargs, self.device]
self._update_gradient_fn(shots=override_shots, tape=self._tape)
try:
res = self._execution_component(args, kwargs, override_shots=override_shots)
finally:
if old_interface == "auto":
self._interface = "auto"
self._update_original_device()
_, self.gradient_kwargs, self.device = original_grad_fn
return res
qnode = lambda device, **kwargs: functools.partial(QNode, device=device, **kwargs)
qnode.__doc__ = QNode.__doc__
qnode.__signature__ = inspect.signature(QNode)
_modules/pennylane/workflow/qnode
Download Python script
Download Notebook
View on GitHub