Problem using reverse annealing through AWS
Apologies in advance for a very tedious question. I'm accessing the hardware through Amazon Web Services for a project and the reverse annealing feature seems to not be working. In particular, when I run:
device = "arn:aws:braket:::device/qpu/d-wave/Advantage_system1"
qpu = BraketDWaveSampler(("amazon-braket-test-test", "test"), device)
J = {coupler: random.choice([-1, 1]) for coupler in qpu.edgelist}
initial = {qubit: random.randint(0, 1) for qubit in qpu.nodelist}
reverse_schedule = [[0.0, 1.0], [5, 0.45], [99, 0.45], [100, 1.0]]
reverse_anneal_params = dict(anneal_schedule=reverse_schedule,initial_state=initial,reinitialize_state=True)
#sampleset = qpu.sample_ising({}, J, num_reads=10)
sampleset = qpu.sample_ising({}, J, num_reads=10, **reverse_anneal_params)
I get an error (output posted at the end of this message since it is long), however when I comment the last line and uncomment the second to last line it works correctly, suggesting that the reverse annealing protocol is where the issue is (note this example follows very closely to the one at https://docs.dwavesys.com/docs/latest/c_solver_parameters.html#anneal-schedule )
The error output is:
--------------------------------------------------------------------------- ValidationError Traceback (most recent call last) <ipython-input-69-40b538d38917> in <module> 6 reverse_anneal_params = dict(anneal_schedule=reverse_schedule,initial_state=initial,reinitialize_state=True) 7 #sampleset = qpu.sample_ising({}, J, num_reads=10) ----> 8 sampleset = qpu.sample_ising({}, J, num_reads=10, **reverse_anneal_params) ~/.venvs/venv2021_1/lib/python3.8/site-packages/braket/ocean_plugin/braket_dwave_sampler.py in sample_ising(self, h, J, **kwargs) 140 {0: 1, 1: -1} 141 """ --> 142 return super().sample_ising(h, J, **kwargs) 143 144 def sample_ising_quantum_task( ~/.venvs/venv2021_1/lib/python3.8/site-packages/braket/ocean_plugin/braket_sampler.py in sample_ising(self, h, J, **kwargs) 186 h = dict((v, b) for v, b in enumerate(h) if b or v in self._access_optimized_nodelist()) 187 --> 188 aws_task = self.sample_ising_quantum_task(h, J, **kwargs) 189 variables = set(h).union(*J) 190 ~/.venvs/venv2021_1/lib/python3.8/site-packages/braket/ocean_plugin/braket_dwave_sampler.py in sample_ising_quantum_task(self, h, J, **kwargs) 179 {0: 1, 1: -1} 180 """ --> 181 return super().sample_ising_quantum_task(h, J, **kwargs) 182 183 def sample_qubo(self, Q: Dict[Tuple[int, int], int], **kwargs) -> SampleSet: ~/.venvs/venv2021_1/lib/python3.8/site-packages/braket/ocean_plugin/braket_sampler.py in sample_ising_quantum_task(self, h, J, **kwargs) 244 raise BinaryQuadraticModelStructureError("Problem graph incompatible with solver.") 245 --> 246 return self.solver.run( 247 Problem(ProblemType.ISING, h, J), 248 self._s3_destination_folder, ~/.venvs/venv2021_1/lib/python3.8/site-packages/braket/aws/aws_device.py in run(self, task_specification, s3_destination_folder, shots, poll_timeout_seconds, poll_interval_seconds, *aws_quantum_task_args, **aws_quantum_task_kwargs) 138 `braket.aws.aws_quantum_task.AwsQuantumTask.create()` 139 """ --> 140 return AwsQuantumTask.create( 141 self._aws_session, 142 self._arn, ~/.venvs/venv2021_1/lib/python3.8/site-packages/braket/aws/aws_quantum_task.py in create(aws_session, device_arn, task_specification, s3_destination_folder, shots, device_parameters, disable_qubit_rewiring, tags, *args, **kwargs) 122 if tags is not None: 123 create_task_kwargs.update({"tags": tags}) --> 124 return _create_internal( 125 task_specification, 126 aws_session, /usr/lib/python3.8/functools.py in wrapper(*args, **kw) 873 '1 positional argument') 874 --> 875 return dispatch(args[0].__class__)(*args, **kw) 876 877 funcname = getattr(func, '__name__', 'singledispatch function') ~/.venvs/venv2021_1/lib/python3.8/site-packages/braket/aws/aws_quantum_task.py in _(problem, aws_session, create_task_kwargs, device_arn, device_parameters, disable_qubit_rewiring, *args, **kwargs) 444 { 445 "action": problem.to_ir().json(), --> 446 "deviceParameters": DwaveDeviceParameters.parse_obj(device_parameters).json( 447 exclude_none=True 448 ), ~/.venvs/venv2021_1/lib/python3.8/site-packages/pydantic/main.cpython-38-x86_64-linux-gnu.so in pydantic.main.BaseModel.parse_obj() ~/.venvs/venv2021_1/lib/python3.8/site-packages/pydantic/main.cpython-38-x86_64-linux-gnu.so in pydantic.main.BaseModel.__init__() ValidationError: 1 validation error for DwaveDeviceParameters providerLevelParameters -> initialState value is not a valid list (type=type_error.list)
Comments
Hello,
It looks like you are using Amazon Braket, so you might need to reach out to the Amazon Support team to get more information that is specific to Braket:
https://aws.amazon.com/contact-us/
Usually the initial_state variable takes on the form of a dictionary as you can see in the documentation.
Something like the following:
{0: -1, 1: -1}The reverse_annealing Jupyter Notebook might be a useful resource for you to look at.
It is under /leap/techniques/reverse_annealing in the Jupyter Hub:
https://cloud.dwavesys.com/leap/learning/
As I had mentioned above, Braket might wrap things a little different, so you might require additional assistance from Amazon.
I hope this helps to resolve your issue!
Please let us know if you have any additional questions or follow-up.
I have spoken with someone at AWS and they have helped resolve the issue. I will post it here in case other people run into the same problem and are looking here for a solution. The issue apparently relates to https://github.com/dwavesystems/dwave-cloud-client/issues/465 and basically boils down to the fact that the AWS version wants the initial state as a list rather than a dictionary (with 3 for unused variables).
Please sign in to leave a comment.