def sample(self,
bqm,
chain_strength=1.0,
chain_break_method=None,
chain_break_fraction=True,
embedding_parameters=None,
return_embedding=False,
**parameters):
"""Sample from the provided binary quadratic model.
Args:
bqm (:obj:`dimod.BinaryQuadraticModel`):
Binary quadratic model to be sampled from.
chain_strength (float, optional, default=1.0):
Magnitude of the quadratic bias (in SPIN-space) applied between
variables to create chains. The energy penalty of chain breaks
is 2 * `chain_strength`.
chain_break_method (function, optional):
Method used to resolve chain breaks during sample unembedding.
See :func:`~dwave.embedding.unembed_sampleset`.
chain_break_fraction (bool, optional, default=True):
Add a `chain_break_fraction` field to the unembedded response with
the fraction of chains broken before unembedding.
embedding_parameters (dict, optional):
If provided, parameters are passed to the embedding method as
keyword arguments. Overrides any `embedding_parameters` passed
to the constructor.
return_embedding (bool, optional, default=False):
If True, the embedding is added to :attr:`dimod.SampleSet.info`
of the returned sample set.
**parameters:
Parameters for the sampling method, specified by the child
sampler.
Returns:
:obj:`dimod.SampleSet`
Examples:
See the example in :class:`EmbeddingComposite`.
"""
# solve the problem on the child system
child = self.child
# apply the embedding to the given problem to map it to the child sampler
__, target_edgelist, target_adjacency = self.target_structure
# add self-loops to edgelist to handle singleton variables
source_edgelist = list(bqm.quadratic) + [(v, v) for v in bqm.linear]
# get the embedding
if embedding_parameters is None:
embedding_parameters = self.embedding_parameters
else:
# we want the parameters provided to the constructor, updated with
# the ones provided to the sample method. To avoid the extra copy
# we do an update, avoiding the keys that would overwrite the
# sample-level embedding parameters
embedding_parameters.update(
(key, val) for key, val in self.embedding_parameters
if key not in embedding_parameters)
embedding = self.find_embedding(source_edgelist, target_edgelist,
**embedding_parameters)
if bqm and not embedding:
raise ValueError("no embedding found")
bqm_embedded = embed_bqm(bqm,
embedding,
target_adjacency,
chain_strength=chain_strength,
smear_vartype=dimod.SPIN)
if 'initial_state' in parameters:
# if initial_state was provided in terms of the source BQM, we want
# to modify it to now provide the initial state for the target BQM.
# we do this by spreading the initial state values over the
# chains
state = parameters['initial_state']
parameters['initial_state'] = {
u: state[v]
for v, chain in embedding.items() for u in chain
}
if self.scale_aware and 'ignored_interactions' in child.parameters:
ignored = []
for chain in embedding.values():
# just use 0 as a null value because we don't actually need
# the biases, just the interactions
ignored.extend(chain_to_quadratic(chain, target_adjacency, 0))
parameters['ignored_interactions'] = ignored
response = child.sample(bqm_embedded, **parameters)
return unembed_sampleset(response,
embedding,
source_bqm=bqm,
chain_break_method=chain_break_method,
chain_break_fraction=chain_break_fraction,
return_embedding=return_embedding)
def sample(self,
bqm,
chain_strength=None,
chain_break_method=None,
chain_break_fraction=True,
embedding_parameters=None,
return_embedding=None,
warnings=None,
**parameters):
"""Sample from the provided binary quadratic model.
Args:
bqm (:obj:`dimod.BinaryQuadraticModel`):
Binary quadratic model to be sampled from.
chain_strength (float/mapping/callable, optional):
Magnitude of the quadratic bias (in SPIN-space) applied between
variables to create chains. The energy penalty of chain breaks
is 2 * `chain_strength`. If a mapping is passed, a
chain-specific strength is applied. If a callable is passed, it
will be called on `chain_strength(bqm, embedding)` and should
return a float or mapping, to be interpreted as above. By default,
`chain_strength` is calculated with
:func:`~dwave.embedding.chain_strength.uniform_torque_compensation`.
chain_break_method (function/list, optional):
Method or methods used to resolve chain breaks. If multiple
methods are given, the results are concatenated and a new field
called "chain_break_method" specifying the index of the method
is appended to the sample set.
See :func:`~dwave.embedding.unembed_sampleset` and
:mod:`dwave.embedding.chain_breaks`.
chain_break_fraction (bool, optional, default=True):
Add a `chain_break_fraction` field to the unembedded response with
the fraction of chains broken before unembedding.
embedding_parameters (dict, optional):
If provided, parameters are passed to the embedding method as
keyword arguments. Overrides any `embedding_parameters` passed
to the constructor.
return_embedding (bool, optional):
If True, the embedding, chain strength, chain break method and
embedding parameters are added to :attr:`dimod.SampleSet.info`
of the returned sample set. The default behaviour is defined
by :attr:`return_embedding_default`, which itself defaults to
False.
warnings (:class:`~dwave.system.warnings.WarningAction`, optional):
Defines what warning action to take, if any. See
:mod:`~dwave.system.warnings`. The default behaviour is defined
by :attr:`warnings_default`, which itself defaults to
:class:`~dwave.system.warnings.IGNORE`
**parameters:
Parameters for the sampling method, specified by the child
sampler.
Returns:
:obj:`dimod.SampleSet`
Examples:
See the example in :class:`EmbeddingComposite`.
"""
if return_embedding is None:
return_embedding = self.return_embedding_default
# solve the problem on the child system
child = self.child
# apply the embedding to the given problem to map it to the child sampler
__, target_edgelist, target_adjacency = self.target_structure
# add self-loops to edgelist to handle singleton variables
source_edgelist = list(bqm.quadratic) + [(v, v) for v in bqm.linear]
# get the embedding
if embedding_parameters is None:
embedding_parameters = self.embedding_parameters
else:
# we want the parameters provided to the constructor, updated with
# the ones provided to the sample method. To avoid the extra copy
# we do an update, avoiding the keys that would overwrite the
# sample-level embedding parameters
embedding_parameters.update(
(key, val) for key, val in self.embedding_parameters
if key not in embedding_parameters)
embedding = self.find_embedding(source_edgelist, target_edgelist,
**embedding_parameters)
if bqm and not embedding:
raise ValueError("no embedding found")
if not hasattr(embedding, 'embed_bqm'):
embedding = EmbeddedStructure(target_edgelist, embedding)
bqm_embedded = embedding.embed_bqm(bqm,
chain_strength=chain_strength,
smear_vartype=dimod.SPIN)
if warnings is None:
warnings = self.warnings_default
elif 'warnings' in child.parameters:
parameters.update(warnings=warnings)
warninghandler = WarningHandler(warnings)
warninghandler.chain_strength(bqm, embedding.chain_strength, embedding)
warninghandler.chain_length(embedding)
if 'initial_state' in parameters:
# if initial_state was provided in terms of the source BQM, we want
# to modify it to now provide the initial state for the target BQM.
# we do this by spreading the initial state values over the
# chains
state = parameters['initial_state']
parameters['initial_state'] = {
u: state[v]
for v, chain in embedding.items() for u in chain
}
if self.scale_aware and 'ignored_interactions' in child.parameters:
ignored = []
for chain in embedding.values():
# just use 0 as a null value because we don't actually need
# the biases, just the interactions
ignored.extend(chain_to_quadratic(chain, target_adjacency, 0))
parameters['ignored_interactions'] = ignored
response = child.sample(bqm_embedded, **parameters)
def async_unembed(response):
# unembed the sampleset aysnchronously.
warninghandler.chain_break(response, embedding)
sampleset = unembed_sampleset(
response,
embedding,
source_bqm=bqm,
chain_break_method=chain_break_method,
chain_break_fraction=chain_break_fraction,
return_embedding=return_embedding)
if return_embedding:
sampleset.info['embedding_context'].update(
embedding_parameters=embedding_parameters,
chain_strength=embedding.chain_strength)
if chain_break_fraction and len(sampleset):
warninghandler.issue(
"All samples have broken chains",
func=lambda:
(sampleset.record.chain_break_fraction.all(), None))
if warninghandler.action is WarningAction.SAVE:
# we're done with the warning handler so we can just pass the list
# off, if later we want to pass in a handler or similar we should
# do a copy
sampleset.info.setdefault('warnings',
[]).extend(warninghandler.saved)
return sampleset
return dimod.SampleSet.from_future(response, async_unembed)