Added custom simulator option #51
Conversation
Codecov Report
@@ Coverage Diff @@
## master #51 +/- ##
==========================================
+ Coverage 99.07% 99.11% +0.04%
==========================================
Files 8 8
Lines 323 338 +15
==========================================
+ Hits 320 335 +15
Misses 3 3
Continue to review full report at Codecov.
|
pennylane_cirq/cirq_device.py
Outdated
| _observable_map = { | ||
| "PauliX": None, | ||
| "PauliY": None, | ||
| "PauliZ": None, | ||
| "Hadamard": None, | ||
| "PauliX": CirqOperation(lambda: cirq.X), | ||
| "PauliY": CirqOperation(lambda: cirq.Y), | ||
| "PauliZ": CirqOperation(lambda: cirq.Z), | ||
| "Hadamard": CirqOperation(lambda: cirq.H), | ||
| "Hermitian": None, | ||
| "Identity": None, | ||
| "Identity": CirqOperation(lambda: cirq.I), | ||
| } |
There was a problem hiding this comment.
This dictionary is actually useful now
|
|
||
| cirq_op = self._observable_map[observable.name] | ||
| if cirq_op is None: | ||
| raise NotImplementedError(f"{observable.name} is not currently supported.") |
There was a problem hiding this comment.
I assume this catches qml.Hermitian?
Out of curiosity, are there any observables cirq supports that PL doesn't?
There was a problem hiding this comment.
We could use qml.utils.decompose_hamiltonian() to convert the Hermitian matrix to a Pauli string, but not sure if it's worth it
pennylane_cirq/simulator_device.py
Outdated
| return self._simulator.simulate_expectation_values( | ||
| self.circuit, self.to_paulistring(observable) | ||
| )[0] |
There was a problem hiding this comment.
Nice, I suppose this leads to a performance improvement?
There was a problem hiding this comment.
Yes, it essentially moves the responsibility of calculating the expectation value to the cirq device rather than just pennylane. The downstream device can then optimize how it's calculated.
There was a problem hiding this comment.
Awesome. A lot of PennyLane plugin patterns are repeated from a time (a long way!) back when other frameworks didn't provide (accessible) expectation APIs. Now that this is commonplace, we should take this into account more when writing plugins.
There was a problem hiding this comment.
Agreed, especially for more future remote services (both simulation and real QC).
There was a problem hiding this comment.
Yep. It also avoids needing to carry around a potentially large probability array longer than we need to.
| qml.PauliX(0) | ||
| return qml.expval(qml.PauliX(0)) | ||
|
|
||
| assert circuit() == 0.123 |
|
|
||
| def __init__(self, wires, shots=1000, analytic=True, qubits=None): | ||
| # pylint: disable=too-many-arguments | ||
| def __init__(self, wires, shots=1000, analytic=True, qubits=None, simulator=None): |
There was a problem hiding this comment.
Oh, also don't forget to add the new keyword argument to the docstring
There was a problem hiding this comment.
I would but there wasn't a docstring before. I can add one now tho.
There was a problem hiding this comment.
There's one above, below the class : line :)
Allows users to use custom cirq simulators with the existing cirq plugin.
Main reason for this is to support the new Floq integration. To use floq, users simply now can do
And that's it!