Clients

The solvers that provide sampling for solving Ising and QUBO problems, such as a D-Wave 2000Q QPU or a software sampler such as the dimod simulated annealing sampler, are typically remote resources. The D-Wave Cloud Client Client class manages such remote solver resources.

Preferred use is with a context manager—a with Client.from_config(...) as construct—to ensure proper closure of all resources. The following example snippet creates a client based on an auto-detected configuration file and instantiates a solver.

>>> with Client.from_config() as client:   # doctest: +SKIP
...     solver = client.get_solver(num_qubits__gt=2000)

Alternatively, the following example snippet creates a client for software resources that it later explicitly closes.

>>> client = Client.from_config(software=True)   # doctest: +SKIP
>>> # code that uses client
>>> client.close()    # doctest: +SKIP

Typically you use the Client class. By default, it instantiates a QPU client. You can also use the specialized QPU and CPU/GPU clients directly.

Client (Base Client)

D-Wave API clients handle communications with solver resources: problem submittal, monitoring, samples retrieval, etc.

Examples

This example creates a client using the local system’s default D-Wave Cloud Client configuration file, which is configured to access a D-Wave 2000Q QPU, submits a QUBO problem (a Boolean NOT gate represented by a penalty model), and samples 5 times.

>>> from dwave.cloud import Client
>>> Q = {(0, 0): -1, (0, 4): 0, (4, 0): 2, (4, 4): -1}
>>> with Client.from_config() as client:  # doctest: +SKIP
...     solver = client.get_solver()
...     computation = solver.sample_qubo(Q, num_reads=5)
...
>>> for i in range(5):     # doctest: +SKIP
...     print(computation.samples[i][0], computation.samples[i][4])
...
(1, 0)
(1, 0)
(0, 1)
(0, 1)
(0, 1)

Class

class dwave.cloud.client.Client(endpoint=None, token=None, solver=None, proxy=None, permissive_ssl=False, request_timeout=60, polling_timeout=None, connection_close=False, headers=None, **kwargs)

Base client class for all D-Wave API clients. Used by QPU and software sampler classes.

Manages workers and handles thread pools for submitting problems, cancelling tasks, polling problem status, and retrieving results.

Parameters:
  • endpoint (str) – D-Wave API endpoint URL.
  • token (str) – Authentication token for the D-Wave API.
  • solver (dict/str) – Default solver features (or simply solver name).
  • proxy (str) – Proxy URL to be used for accessing the D-Wave API.
  • permissive_ssl (bool, default=False) – Disables SSL verification.
  • request_timeout (float, default=60) – Connect and read timeout (in seconds) for all requests to the D-Wave API.
  • polling_timeout (float, default=None) – Problem status polling timeout (in seconds), after which polling is aborted.
  • connection_close (bool, default=False) – Force HTTP(S) connection close after each request.
  • headers (dict/str) – Additional HTTP headers.
Other Parameters:
 

Unrecognized keys (str) – All unrecognized keys are passed through to the appropriate client class constructor as string keyword arguments.

An explicit key value overrides an identical user-defined key value loaded from a configuration file.

Examples

This example directly initializes a Client. Direct initialization uses class constructor arguments, the minimum being a value for token.

>>> from dwave.cloud import Client
>>> client = Client(token='secret')
>>> # code that uses client
>>> client.close()

Methods

client.Client.from_config([config_file, …]) Client factory method to instantiate a client instance from configuration.
client.Client.solvers([refresh]) Deprecated in favor of get_solvers().
client.Client.get_solver([name, refresh]) Load the configuration for a single solver.
client.Client.get_solvers([refresh, order_by]) Return a filtered list of solvers handled by this client.
client.Client.is_solver_handled(solver) Determine if the specified solver should be handled by this client.
client.Client.close() Perform a clean shutdown.

Specialized Clients

Typically you use the Client class. By default, it instantiates a QPU client. You can also instantiate a QPU or CPU/GPU client directly.

QPU Client

An implementation of the REST API for D-Wave Solver API (SAPI) servers.

SAPI servers provide authentication, queuing, and scheduling services, and provide a network interface to solvers. This API enables you submit a binary quadratic (Ising or QUBO) model and receive samples from a distribution over the model as defined by a selected solver.

SAPI server workflow is roughly as follows:

  1. Submitted problems enter an input queue. Each user has an input queue per solver.
  2. Drawing from all input queues for a solver, problems are scheduled.
  3. Results are cached for retrieval by the client.

Class

class dwave.cloud.qpu.Client(endpoint=None, token=None, solver=None, proxy=None, permissive_ssl=False, request_timeout=60, polling_timeout=None, connection_close=False, headers=None, **kwargs)

D-Wave API client specialized to work with QPU solvers.

This class is instantiated by default, or explicitly when client=qpu, with the typical base client instantiation with Client.from_config() as client: of a client. (You should not instantiate this class with client=sw or use it with solver feature constraint software=True.)

Examples

This example explicitly instantiates a dwave.cloud.qpu.client based on the local system`s default D-Wave Cloud Client configuration file to sample a random Ising problem tailored to fit the client`s default solver`s graph.

import random
from dwave.cloud.qpu import Client

# Use context manager to ensure resources (thread pools used by Client) are released
with Client.from_config() as client:

    solver = client.get_solver()

    # Build problem to exactly fit the solver graph
    linear = {index: random.choice([-1, 1]) for index in solver.nodes}
    quad = {key: random.choice([-1, 1]) for key in solver.undirected_edges}

    # Sample 100 times and print out the first sample
    computation = solver.sample_ising(linear, quad, num_reads=100)
    print(computation.samples[0])

Methods

qpu.Client.is_solver_handled(solver) Determine if the specified solver should be handled by this client.

Software-Samplers Client

Class

class dwave.cloud.sw.Client(endpoint=None, token=None, solver=None, proxy=None, permissive_ssl=False, request_timeout=60, polling_timeout=None, connection_close=False, headers=None, **kwargs)

D-Wave API client specialized to work with remote software solvers (samplers).

This class is instantiated by default, or explicitly when client=sw, with the typical base client instantiation with Client.from_config() as client: of a client. (You should not instantiate this class with client=qpu or use it with solver feature constraint qpu=True.)

Examples

This example indirectly instantiates a dwave.cloud.sw.client based on the local system`s default D-Wave Cloud Client configuration file to sample a random Ising problem tailored to fit the client`s default solver`s graph.

import random
from dwave.cloud import Client

# Use context manager to ensure resources (thread pools used by Client) are released
with Client.from_config(solver={"software": True}) as client:

    solver = client.get_solver()

    # Build problem to exactly fit the solver graph
    linear = {index: random.choice([-1, 1]) for index in solver.nodes}
    quad = {key: random.choice([-1, 1]) for key in solver.undirected_edges}

    # Sample 100 times and print out the first sample
    computation = solver.sample_ising(linear, quad, num_reads=100)
    print(computation.samples[0])

Methods

sw.Client.is_solver_handled(solver) Determine if the specified solver should be handled by this client.