def test_embedding_to_one_node(self):
"""an embedding that maps everything to one node should result in a singleton graph"""
target_adj = nx.barbell_graph(16, 7)
embedding = {'a': set(target_adj)} # all map to 'a'
source_adj = eutil.target_to_source(target_adj, embedding)
self.assertEqual(source_adj, {'a': set()})
embedding = {'a': {0, 1}} # not every node is assigned to a chain
source_adj = eutil.target_to_source(target_adj, embedding)
self.assertEqual(source_adj, {'a': set()})
def test_embedding_overlap(self):
"""overlapping embeddings should raise an error"""
target_adj = nx.complete_graph(5)
embedding = {'a': {0, 1}, 'b': {1, 2}} # overlap
with self.assertRaises(ValueError):
source_adj = eutil.target_to_source(target_adj, embedding)
def load_embedding(target_nodelist, target_edgelist, embedding, embedding_tag):
target_adjacency = {v: set() for v in target_nodelist}
for u, v in target_edgelist:
target_adjacency[u].add(v)
target_adjacency[v].add(u)
source_adjacency = embutil.target_to_source(target_adjacency, embedding)
source_nodelist = sorted(source_adjacency)
source_edgelist = sorted(
sorted(edge) for edge in _adjacency_to_edges(source_adjacency))
with cache_connect() as cur:
insert_embedding(cur, source_nodelist, source_edgelist,
target_nodelist, target_edgelist, embedding,
embedding_tag)
def test_self_embedding(self):
"""a 1-to-1 embedding should not change the adjacency"""
target_adj = dnx.chimera_graph(4)
embedding = {v: {v} for v in target_adj}
source_adj = eutil.target_to_source(target_adj, embedding)
# print(source_adj)
# test the adjacencies are equal (source_adj is a dict and target_adj is a networkx graph)
for v in target_adj:
self.assertIn(v, source_adj)
for u in target_adj[v]:
self.assertIn(u, source_adj[v])
for v in source_adj:
self.assertIn(v, target_adj)
for u in source_adj[v]:
self.assertIn(u, target_adj[v])
def __init__(self,
sampler,
embedding,
chain_strength=None,
flux_biases=None,
flux_bias_num_reads=1000,
flux_bias_max_age=3600):
self.children = [sampler]
self.parameters = parameters = {'apply_flux_bias_offsets': []}
parameters.update(sampler.parameters)
self.properties = {'child_properties': sampler.properties.copy()}
#
# Get the adjacency of the child sampler (this is the target for our embedding)
#
try:
target_nodelist, target_edgelist, target_adjacency = sampler.structure
except:
# todo, better exception catching
raise
#
# Validate the chain strength, or obtain it from J-range if chain strength is not provided.
#
self.chain_strength = self._validate_chain_strength(chain_strength)
#
# We want to track the persistent embedding so that we can map input problems
# to the child sampler.
#
if isinstance(embedding, str):
embedding = get_embedding_from_tag(embedding, target_nodelist,
target_edgelist)
elif not isinstance(embedding, dict):
raise TypeError("expected input `embedding` to be a dict.")
self.embedding = embedding
#
# Derive the structure of our composed from the target graph and the embedding
#
source_adjacency = embutil.target_to_source(target_adjacency,
embedding)
try:
nodelist = sorted(source_adjacency)
edgelist = sorted(_adjacency_to_edges(source_adjacency))
except TypeError:
# python3 does not allow sorting of unlike types, so if nodes have
# different type names just choose an arbitrary order
nodelist = list(source_adjacency)
edgelist = list(_adjacency_to_edges(source_adjacency))
self.nodelist = nodelist
self.edgelist = edgelist
self.adjacency = source_adjacency
#
# If the sampler accepts flux bias offsets, we'll want to set them
#
if flux_biases is None and FLUX_BIAS_KWARG in sampler.parameters:
# If nothing is provided, then we either get them from the cache or generate them
flux_biases = get_flux_biases(sampler,
embedding,
num_reads=flux_bias_num_reads,
max_age=flux_bias_max_age)
elif flux_biases:
if FLUX_BIAS_KWARG not in sampler.accepted_kwargs:
raise ValueError(
"Given child sampler does not accept flux_biases.")
# something provided, error check
if not isinstance(flux_biases, list):
flux_biases = list(flux_biases) # cast to a list
else:
# disabled, empty or not available for this sampler so do nothing
flux_biases = None
self.flux_biases = flux_biases
