View on QuantumAI | Run in Google Colab | View source on GitHub | Download notebook |
try:
import cirq
except ImportError:
print("installing cirq...")
!pip install --quiet cirq
import cirq
print("installed cirq.")
Conceptual overview
The primary representation of quantum programs in Cirq is the Circuit
class. A Circuit
is a collection of Moments
. A Moment
is a collection of Operations
that all act during the same abstract time slice. An Operation
is some effect that operates on a specific subset of Qubits; the most common type of Operation
is a GateOperation
.
Let's unpack this.
At the base of this construction is the notion of a qubit. In Cirq, qubits and other quantum objects are identified by instances of subclasses of the cirq.Qid
base class. Different subclasses of Qid
can be used for different purposes. For example, the qubits that Google’s devices use are often arranged on the vertices of a square grid. For this, the class cirq.GridQubit
subclasses cirq.Qid
. For example, you can create a 3 by 3 grid of qubits using:
qubits = cirq.GridQubit.square(3)
print(qubits[0])
print(qubits)
q(0, 0) [cirq.GridQubit(0, 0), cirq.GridQubit(0, 1), cirq.GridQubit(0, 2), cirq.GridQubit(1, 0), cirq.GridQubit(1, 1), cirq.GridQubit(1, 2), cirq.GridQubit(2, 0), cirq.GridQubit(2, 1), cirq.GridQubit(2, 2)]
The next level up is the notion of cirq.Gate
. A cirq.Gate
represents a physical process that occurs on a qubit. The important property of a gate is that it can be applied to one or more qubits. This can be done via the gate.on(*qubits)
method itself or via gate(*qubits)
. Doing this turns a cirq.Gate
into a cirq.Operation
.
# This is a Pauli X gate. It is an object instance.
x_gate = cirq.X
# Applying it to the qubit at location (0, 0) (defined above)
# turns it into an operation.
x_op = x_gate(qubits[0])
print(x_op)
X(q(0, 0))
A cirq.Moment
is simply a collection of operations, each of which operates on a different set of qubits, and which conceptually represents these operations as occurring during this abstract time slice. The Moment
structure itself is not required to be related to the actual scheduling of the operations on a quantum computer or via a simulator, though it can be. For example, here is a Moment
in which Pauli X
and a CZ
gate operate on three qubits:
cz = cirq.CZ(qubits[0], qubits[1])
x = cirq.X(qubits[2])
moment = cirq.Moment(x, cz)
print(moment)
╷ 0 1 2 ╶─┼─────── 0 │ @─@ X │
The above is not the only way one can construct moments, nor even the typical method, but illustrates that a Moment
is just a collection of operations on disjoint sets of qubits.
Finally, at the top level, a cirq.Circuit
is an ordered series of cirq.Moment
objects. The first Moment
in this series contains the first Operations
that will be applied. Here, for example, is a simple circuit made up of two moments:
cz01 = cirq.CZ(qubits[0], qubits[1])
x2 = cirq.X(qubits[2])
cz12 = cirq.CZ(qubits[1], qubits[2])
moment0 = cirq.Moment([cz01, x2])
moment1 = cirq.Moment([cz12])
circuit = cirq.Circuit((moment0, moment1))
print(circuit)
(0, 0): ───@─────── │ (0, 1): ───@───@─── │ (0, 2): ───X───@───
Note that the above is one of the many ways to construct a Circuit
, which illustrates the concept that a Circuit
is an iterable of Moment
objects.
Constructing circuits
Constructing Circuits as a series of hand-crafted Moment
objects is tedious. Instead, Cirq provides a variety of different ways to create a Circuit
.
One of the most useful ways to construct a Circuit
is by appending onto the Circuit
with the Circuit.append
method.
q0, q1, q2 = [cirq.GridQubit(i, 0) for i in range(3)]
circuit = cirq.Circuit()
circuit.append([cirq.CZ(q0, q1), cirq.H(q2)])
print(circuit)
(0, 0): ───@─── │ (1, 0): ───@─── (2, 0): ───H───
This appended a new moment to the qubit, which you can continue to do:
circuit.append([cirq.H(q0), cirq.CZ(q1, q2)])
print(circuit)
(0, 0): ───@───H─── │ (1, 0): ───@───@─── │ (2, 0): ───H───@───
These two examples appended full moments. What happens when you append all of these at once?
circuit = cirq.Circuit()
circuit.append([cirq.CZ(q0, q1), cirq.H(q2), cirq.H(q0), cirq.CZ(q1, q2)])
print(circuit)
(0, 0): ───@───H─── │ (1, 0): ───@───@─── │ (2, 0): ───H───@───
This has again created two Moment
objects. How did Circuit
know how to do this? The Circuit.append
method (and its cousin, Circuit.insert
) both take an argument called strategy
of type cirq.InsertStrategy
. By default, InsertStrategy
is InsertStrategy.EARLIEST
.
InsertStrategies
cirq.InsertStrategy
defines how Operations
are placed in a Circuit
when requested to be inserted at a given location. Here, a location is identified by the index of the Moment
(in the Circuit
) where the insertion is requested to be placed at (in the case of Circuit.append
, this means inserting at the Moment
, at an index one greater than the maximum moment index in the Circuit
).
There are four such strategies: InsertStrategy.EARLIEST
, InsertStrategy.NEW
, InsertStrategy.INLINE
and InsertStrategy.NEW_THEN_INLINE
.
InsertStrategy.EARLIEST
, which is the default, is defined as:
Scans backward from the insert location until a moment with operations touching qubits affected by the operation to insert is found. The operation is added to the moment just after that location.
For example, if you first create an Operation
in a single moment, and then use InsertStrategy.EARLIEST
, the Operation
can slide back to this first Moment
if there is space:
from cirq.circuits import InsertStrategy
circuit = cirq.Circuit()
circuit.append([cirq.CZ(q0, q1)])
circuit.append([cirq.H(q0), cirq.H(q2)], strategy=InsertStrategy.EARLIEST)
print(circuit)
(0, 0): ───@───H─── │ (1, 0): ───@─────── (2, 0): ───H───────
After creating the first moment with a CZ
gate, the second append uses the InsertStrategy.EARLIEST
strategy. The H
on q0
cannot slide back, while the H
on q2
can and so ends up in the first Moment
.
Contrast this with InsertStrategy.NEW
that is defined as:
Every operation that is inserted is created in a new moment.
circuit = cirq.Circuit()
circuit.append([cirq.H(q0), cirq.H(q1), cirq.H(q2)], strategy=InsertStrategy.NEW)
print(circuit)
(0, 0): ───H─────────── (1, 0): ───────H─────── (2, 0): ───────────H───
Here every operator processed by the append ends up in a new moment. InsertStrategy.NEW
is most useful when you are inserting a single operation and do not want it to interfere with other Moments
.
Another strategy is InsertStrategy.INLINE
:
Attempts to add the operation to insert into the moment just before the desired insert location. But, if there’s already an existing operation affecting any of the qubits touched by the operation to insert, a new moment is created instead.
circuit = cirq.Circuit()
circuit.append([cirq.CZ(q1, q2)])
circuit.append([cirq.CZ(q1, q2)])
circuit.append([cirq.H(q0), cirq.H(q1), cirq.H(q2)], strategy=InsertStrategy.INLINE)
print(circuit)
(0, 0): ───────H─────── (1, 0): ───@───@───H─── │ │ (2, 0): ───@───@───H───
After two initial CZ
between the second and third qubit, the example inserts three H
operations. The H
on the first qubit is inserted into the previous Moment
, but the H
on the second and third qubits cannot be inserted into the previous Moment
, so a new Moment
is created.
Finally, InsertStrategy.NEW_THEN_INLINE
is a useful strategy to start a new moment and then continue
inserting from that point onwards.
Creates a new moment at the desired insert location for the first operation, but then switches to inserting operations according to InsertStrategy.INLINE
.
circuit = cirq.Circuit()
circuit.append([cirq.H(q0)])
circuit.append([cirq.CZ(q1, q2), cirq.H(q0)], strategy=InsertStrategy.NEW_THEN_INLINE)
print(circuit)
(0, 0): ───H───H─── (1, 0): ───────@─── │ (2, 0): ───────@───
The first append creates a single moment with an H
on the first qubit. Then, the append with the InsertStrategy.NEW_THEN_INLINE
strategy begins by inserting the CZ
in a new Moment
(the InsertStrategy.NEW
in InsertStrategy.NEW_THEN_INLINE
). Subsequent appending is done InsertStrategy.INLINE
, so the next H
on the first qubit is appending in the just created Moment
.
Here is a picture showing simple examples of appending 1 and then 2 using the different strategies
Patterns for arguments to append and insert
In the above examples, you used a series of Circuit.append
calls with a list of different Operations
added to the circuit. However, the argument where you have supplied a list can also take more than just list values. For instance:
def my_layer():
yield cirq.CZ(q0, q1)
yield [cirq.H(q) for q in (q0, q1, q2)]
yield [cirq.CZ(q1, q2)]
yield [cirq.H(q0), [cirq.CZ(q1, q2)]]
circuit = cirq.Circuit()
circuit.append(my_layer())
for x in my_layer():
print(x)
CZ(q(0, 0), q(1, 0)) [cirq.H(cirq.GridQubit(0, 0)), cirq.H(cirq.GridQubit(1, 0)), cirq.H(cirq.GridQubit(2, 0))] [cirq.CZ(cirq.GridQubit(1, 0), cirq.GridQubit(2, 0))] [cirq.H(cirq.GridQubit(0, 0)), [cirq.CZ(cirq.GridQubit(1, 0), cirq.GridQubit(2, 0))]]
print(circuit)
(0, 0): ───@───H───H─────── │ (1, 0): ───@───H───@───@─── │ │ (2, 0): ───H───────@───@───
Recall that Python functions with a yield
are generators. Generators are functions that act as iterators. The above example iterates over my_layer()
with a for
loop. In this case, each of the yield
s produces:
Operations
,- lists of
Operations
, - or lists of
Operations
mixed with lists ofOperations
.
When you pass an iterator to the append
method, Circuit
is able to flatten all of these and pass them as one giant list to Circuit.append
(this also works for Circuit.insert
).
The above idea uses the concept of cirq.OP_TREE
. An OP_TREE
is not a class, but a contract. The basic idea is that, if the input can be iteratively flattened into a list of operations, then the input is an OP_TREE
.
A very nice pattern emerges from this structure: define generators for sub-circuits, which can vary by size or Operation
parameters.
Another useful method to construct a Circuit
fully formed from an OP_TREE
is to pass the OP_TREE
into Circuit
when initializing it:
circuit = cirq.Circuit(cirq.H(q0), cirq.H(q1))
print(circuit)
(0, 0): ───H─── (1, 0): ───H───
Slicing and iterating over circuits
Circuits can be iterated over and sliced. When they are iterated, each item in the iteration is a moment:
circuit = cirq.Circuit(cirq.H(q0), cirq.CZ(q0, q1))
for moment in circuit:
print(moment)
╷ 0 ╶─┼─── 0 │ H │ ╷ 0 ╶─┼─── 0 │ @ │ │ 1 │ @ │
Slicing a Circuit
, on the other hand, produces a new Circuit
with only the moments corresponding to the slice:
circuit = cirq.Circuit(cirq.H(q0), cirq.CZ(q0, q1), cirq.H(q1), cirq.CZ(q0, q1))
print(circuit[1:3])
(0, 0): ───@─────── │ (1, 0): ───@───H───
Two especially useful applications of this are dropping the last moment (which are often just measurements): circuit[:-1]
, and reversing a circuit: circuit[::-1]
.
Nesting circuits with CircuitOperation
Circuits can be nested inside one another with cirq.CircuitOperation
. This is useful for concisely defining large, repetitive circuits, as the repeated section can be defined once and then be reused elsewhere. Circuits that need to be serialized especially benefit from this, as loops and functions used in the Python construction of a circuit are otherwise not captured in serialization.
The subcircuit must first be "frozen" to indicate that no further changes will be made to it.
subcircuit = cirq.Circuit(cirq.H(q1), cirq.CZ(q0, q1), cirq.CZ(q2, q1), cirq.H(q1))
subcircuit_op = cirq.CircuitOperation(subcircuit.freeze())
circuit = cirq.Circuit(cirq.H(q0), cirq.H(q2), subcircuit_op)
print(circuit)
[ (0, 0): ───────@─────────── ] [ │ ] (0, 0): ───H───[ (1, 0): ───H───@───@───H─── ]─── [ │ ] [ (2, 0): ───────────@─────── ] │ (1, 0): ───────#2──────────────────────────────── │ (2, 0): ───H───#3────────────────────────────────
Frozen circuits can also be constructed directly, for convenience.
circuit = cirq.Circuit(
cirq.CircuitOperation(
cirq.FrozenCircuit(cirq.H(q1), cirq.CZ(q0, q1), cirq.CZ(q2, q1), cirq.H(q1))
)
)
print(circuit)
[ (0, 0): ───────@─────────── ] [ │ ] (0, 0): ───[ (1, 0): ───H───@───@───H─── ]─── [ │ ] [ (2, 0): ───────────@─────── ] │ (1, 0): ───#2──────────────────────────────── │ (2, 0): ───#3────────────────────────────────
A CircuitOperation
is sort of like a function: by default, it will behave like the circuit it contains, but you can also pass arguments to it that alter the qubits it operates on, the number of times it repeats, and other properties. CircuitOperation
s can also be referenced multiple times within the same "outer" circuit for conciseness.
subcircuit_op = cirq.CircuitOperation(cirq.FrozenCircuit(cirq.CZ(q0, q1)))
# Create a copy of subcircuit_op that repeats twice...
repeated_subcircuit_op = subcircuit_op.repeat(2)
# ...and another copy that replaces q0 with q2 to perform CZ(q2, q1).
moved_subcircuit_op = subcircuit_op.with_qubit_mapping({q0: q2})
circuit = cirq.Circuit(repeated_subcircuit_op, moved_subcircuit_op)
print(circuit)
[ (0, 0): ───@─── ] (0, 0): ───[ │ ]──────────────────────────────────────────────────────────────── [ (1, 0): ───@─── ](loops=2) │ (1, 0): ───#2─────────────────────────────#2────────────────────────────────────────────────── │ [ (0, 0): ───@─── ] (2, 0): ──────────────────────────────────[ │ ]───────────────────────────────── [ (1, 0): ───@─── ](qubit_map={q(0, 0): q(2, 0)})
For the most part, a CircuitOperation
behaves just like a regular Operation
: its qubits are the qubits of the contained circuit (after applying any provided mapping), and it can be placed inside any Moment
that doesn't already contain operations on those qubits. This means that CircuitOperation
s can be used to represent more complex operation timing, such as three operations on one qubit in parallel with two operations on another:
subcircuit_op = cirq.CircuitOperation(cirq.FrozenCircuit(cirq.H(q0)))
circuit = cirq.Circuit(
subcircuit_op.repeat(3), subcircuit_op.repeat(2).with_qubit_mapping({q0: q1})
)
print(circuit)
(0, 0): ───[ (0, 0): ───H─── ](loops=3)───────────────────────────────── (1, 0): ───[ (0, 0): ───H─── ](qubit_map={q(0, 0): q(1, 0)}, loops=2)───
In the above example, even though the top CircuitOperation
is iterated three times and the bottom one is iterated two times, they still reside within the same Moment
, meaning they can be thought of conceptually as executing simultaneously in the same time step. However, this may not hold when the circuit is run on hardware or a simulator.
CircuitOperation
s can also be nested within each other to arbitrary depth.
qft_1 = cirq.CircuitOperation(cirq.FrozenCircuit(cirq.H(q0)))
qft_2 = cirq.CircuitOperation(cirq.FrozenCircuit(cirq.H(q1), cirq.CZ(q0, q1) ** 0.5, qft_1))
qft_3 = cirq.CircuitOperation(
cirq.FrozenCircuit(cirq.H(q2), cirq.CZ(q1, q2) ** 0.5, cirq.CZ(q0, q2) ** 0.25, qft_2)
)
# etc.
Finally, the mapped_circuit
method will return the circuit that a CircuitOperation
represents after all repetitions and remappings have been applied. By default, this only "unrolls" a single layer of CircuitOperation
s. To recursively unroll all layers, you can pass deep=True
to this method.
# A large CircuitOperation with other sub-CircuitOperations.
print('Original qft_3 CircuitOperation')
print(qft_3)
# Unroll the outermost CircuitOperation to a normal circuit.
print('Single layer unroll:')
print(qft_3.mapped_circuit(deep=False))
# Unroll all of the CircuitOperations recursively.
print('Recursive unroll:')
print(qft_3.mapped_circuit(deep=True))
Original qft_3 CircuitOperation [ [ (0, 0): ───────@───────[ (0, 0): ───H─── ]─── ] ] [ (0, 0): ───────────────@────────[ │ ]─── ] [ │ [ (1, 0): ───H───@^0.5───────────────────────── ] ] [ │ │ ] [ (1, 0): ───────@───────┼────────#2────────────────────────────────────────────────── ] [ │ │ ] [ (2, 0): ───H───@^0.5───@^0.25─────────────────────────────────────────────────────── ] Single layer unroll: [ (0, 0): ───────@───────[ (0, 0): ───H─── ]─── ] (0, 0): ───────────────@────────[ │ ]─── │ [ (1, 0): ───H───@^0.5───────────────────────── ] │ │ (1, 0): ───────@───────┼────────#2────────────────────────────────────────────────── │ │ (2, 0): ───H───@^0.5───@^0.25─────────────────────────────────────────────────────── Recursive unroll: (0, 0): ───────────────@────────────@───────H─── │ │ (1, 0): ───────@───────┼────────H───@^0.5─────── │ │ (2, 0): ───H───@^0.5───@^0.25───────────────────
Summary
Circuit
s are sliceable iterables of Moments
, which are sliceable iterables of Operation
s, which are Gate
s applied to Qubit
s. Cirq provides intuitive and flexible ways to construct, add to, divide, and nest Circuits
, with generators, slicing, and CircuitOperation
s.
Knowing how to create a circuit is useful, but what you put into a circuit is the interesting part. Read more about these constituent structures at:
- Qubits - the software representation of quantum bits
- Gates - quantum gates that become operations when applied to qubits
- Operators - more complicated structures to put in circuits
Once you've built your circuit, read these to learn what you can do with it:
- Simulation - run your circuit on the Cirq simulator to see what it does
- Quantum Virtual Machine - run your circuit on a specialized simulator that mimics actual quantum hardware, including noise.
- Transform circuits - run transformer functions to change your circuit in different ways
If you need to import or export circuits into or out of Cirq, see:
- Import/export circuits - features to serialize/deserialize circuits into/from different formats