def GetTopicModel(self, model=None, args=None, class_ids=None, topic_names=None, use_sparse_format=None,
request_type=None):
if args is None:
args = messages_pb2.GetTopicModelArgs()
if model is not None:
args.model_name = model.name()
if class_ids is not None:
args.ClearField('class_id')
for class_id in class_ids:
args.class_id.append(class_id)
if topic_names is not None:
args.ClearField('topic_name')
for topic_name in topic_names:
args.topic_name.append(topic_name)
if use_sparse_format is not None:
args.use_sparse_format=use_sparse_format
if request_type is not None:
args.request_type = request_type
args_blob = args.SerializeToString()
length = HandleErrorCode(self.lib_, self.lib_.ArtmRequestTopicModel(self.id_, len(args_blob), args_blob))
topic_model_blob = ctypes.create_string_buffer(length)
HandleErrorCode(self.lib_, self.lib_.ArtmCopyRequestResult(length, topic_model_blob))
topic_model = messages_pb2.TopicModel()
topic_model.ParseFromString(topic_model_blob)
return topic_model
def GetTopicModel(self, model):
length = HandleErrorCode(
self.lib_, self.lib_.ArtmRequestTopicModel(self.id_, model.name()))
topic_model_blob = ctypes.create_string_buffer(length)
HandleErrorCode(
self.lib_, self.lib_.ArtmCopyRequestResult(length,
topic_model_blob))
topic_model = messages_pb2.TopicModel()
topic_model.ParseFromString(topic_model_blob)
return topic_model
def Overwrite(self, topic_model, commit=True):
copy_ = messages_pb2.TopicModel()
copy_.CopyFrom(topic_model)
copy_.name = self.name()
blob = copy_.SerializeToString()
blob_p = ctypes.create_string_buffer(blob)
HandleErrorCode(self.lib_,
self.lib_.ArtmOverwriteTopicModel(self.master_id_, len(blob), blob_p))
if commit:
self.master_component.WaitIdle()
self.Synchronize(decay_weight=0.0, apply_weight=1.0, invoke_regularizers=False)
model_config = messages_pb2.ModelConfig()
model_config.topics_count = topics_count
model_config.inner_iterations_count = inner_iterations_count
model_config.score_name.append(perplexity_score_name)
model_config.score_name.append(sparsity_theta_score_name)
model_config.score_name.append(sparsity_phi_score_name)
model_config.score_name.append(topic_kernel_score_name)
# model_config.regularizer_name.append(regularizer_name_theta)
# model_config.regularizer_tau.append(0.1)
# model_config.regularizer_name.append(regularizer_name_decor)
# model_config.regularizer_tau.append(200000)
model = master_component.CreateModel(model_config)
initial_topic_model = messages_pb2.TopicModel();
initial_topic_model.topics_count = topics_count;
initial_topic_model.name = model.name()
random.seed(123)
for i in range(0, len(unique_tokens.entry)):
token = unique_tokens.entry[i].key_token
initial_topic_model.token.append(token);
weights = initial_topic_model.token_weights.add();
for topic_index in range(0, topics_count):
weights.value.append(random.random())
model.Overwrite(initial_topic_model)
for iter in range(0, outer_iteration_count):
master_component.InvokeIteration(1)
master_component.WaitIdle(120000);