Submission of Qadence Jobs to Pasqal Cloud
It is possible to submit quantum computational jobs to execute remotely on Pasqal's cloud platform from Qadence.
This feature can only be used if you have an account on the cloud platform, which has access to the Qadence workload.
The qadence module qadence.pasqal_cloud_connection offers functionality to specify the computation easily, upload the specification and retrieve the result when the computation has finished execution on the cloud platform.
In this tutorial, a simple quantum circuit will be defined as an example to showcase the submission process for remote computations.
The same process can be applied to run more complex quantum circuits on the cloud platform.
Let's first define a very simple quantum circuit that creates a Bell state.
If we want to upload this circuit to the cloud platform we need to follow 4 steps: - Authentication and connection to cloud - Defining workload specification - Submission - Retrieval of results
Authentication and connection
To setup a connection the cloud platform, use the SDK object present in qadence.pasqal_cloud_connection. The email and password are the ones used to login to the webportal. The project-id can be found in the webportal under "Projects".
from qadence.pasqal_cloud_connection import SDK
connection = SDK("john.doe@email.com", "my-password", project_id="proj1")
Defining workload specification
A workload is a quantum calculation to execute on a Pasqal cloud backend.
To create a workload specification, we need some extra information on top the circuit itself.
We need to specify the backend, chosen here to be PyQTorch.
The cloud platform currently only supports PyQTorch.
Moreover, the requested result type needs to be defined.
Based on the workload specification, the appropriate run methods (run, sample or expectation) will be called by the QuantumModel by passing them through the enum value ResultTypes argument.
Moreover, the requested result type needs to be defined.
These are provided in a list, so that multiple result types can be requested in a single submission.
from qadence import BackendName
from qadence.pasqal_cloud_connection import WorkloadSpec, ResultTypes
workload = WorkloadSpec(circuit, BackendName.PYQTORCH, [ResultTypes.SAMPLE, ResultTypes.RUN])
Using a Quantum Model
If you already have your quantum computation defined as a QuantumModel, it is possible to create a workload specification directly from the model using get_spec_from_model.
Then, the circuit and backend specifications will be extracted from the model, the other values need to be provided as extra arguments.
from qadence.pasqal_cloud_connection import get_spec_from_model
model = QuantumModel(...)
workload = get_spec_from_model(model, [ResultType.SAMPLE])
Observable Expectation Value
For the result type ResultType.EXPECTATION it is mandatory to provide an observable to the workload specification.
In the example below we use the trivial identity observable I(0) @ I(1).
workload = WorkloadSpec(circuit, BackendName.PYQTORCH, [ResultTypes.EXPECTATION], observable=I(0)*I(1))
Parametric Circuits
In the case of a parametric circuit, i.e. a circuit that contains feature parameters or variational parameters, values for these parameters need to be provided. The parameter values are defined in a dictionary, where keys are the parameter name and values are parameter value passed as torch tensors. The parameter values are defined in a dictionary, where keys are the parameter name and values are parameter value passed as torch tensors. It is possible to set multiple values by using a 1-D tensor, to the parameters, in that case the computation is executed for each value in the tensor. A mix of 0-D and 1-D tensors can be provided to keep some parameters constant and others changed during this process. However, all 1-D tensors need to have the same length.
parametric_circuit = ...
parameter_values = {"param1": tensor(0), "param2": tensor([0, 1, 2]), "param3": tensor([5, 6, 7])}
workload = WorkloadSpec(parametric_circuit, BackendName.PYQTORCH, [ResultTypes.EXPECTATION], parameter_values=parameter_values)
Submission
Submission to the cloud platform is done very easily using the submit_workload function.
The workload id will be provided by executing the function.
This id is needed later, to request the status of the given workload.
from qadence.pasqal_cloud_connection import submit_workload
workload_id = submit_workload(connection, workload)
Check Workload Status
The status of a workload can be: done, pending, running, paused, canceled, timed out or error.
The check_status function can be used to see if the workload is finished already.
If so, the results of the computation will be provided in a WorkloadResult object.
The result of the computation itself can be found in the result attribute of this object.
If the workload has not finished yet, or resulted in an error, check_status will raise an exception, either a WorkloadStoppedError or WorkloadNotDoneError.
from qadence.pasqal_cloud_connection import check_status
workload_result = check_status(connection, workload_id)
print(workload_result.result)
Retrieval of Results
If you wish to wait for the workload to be finished, before moving further with your code, you can use the get_result function.
This function checks in set intervals the status of the workload until the workload is finished or the function has timed out.
The polling rate as well as the time out duration can be set optionally.