refactor: avoid self.params; assign self._model outside _create_model

annif/backend/nn_ensemble.py

@@ -137,18 +137,17 @@ def _merge_hits_from_sources(self, hits_from_sources, params):
  np.expand_dims(score_vector.transpose(), 0))
  return VectorSuggestionResult(results[0])
 
- def _create_model(self, sources):
+ def _create_model(self, sources, params):
  self.info("creating NN ensemble model")
 
  inputs = Input(shape=(len(self.project.subjects), len(sources)))
 
  flat_input = Flatten()(inputs)
  drop_input = Dropout(
- rate=float(
- self.params['dropout_rate']))(flat_input)
- hidden = Dense(int(self.params['nodes']),
+ rate=float(params['dropout_rate']))(flat_input)
+ hidden = Dense(int(params['nodes']),
  activation="relu")(drop_input)
- drop_hidden = Dropout(rate=float(self.params['dropout_rate']))(hidden)
+ drop_hidden = Dropout(rate=float(params['dropout_rate']))(hidden)
  delta = Dense(len(self.project.subjects),
  kernel_initializer='zeros',
  bias_initializer='zeros')(drop_hidden)
@@ -157,21 +156,22 @@ def _create_model(self, sources):
 
  predictions = Add()([mean, delta])
 
- self._model = Model(inputs=inputs, outputs=predictions)
- self._model.compile(optimizer=self.params['optimizer'],
-  loss='binary_crossentropy',
-  metrics=['top_k_categorical_accuracy'])
- if 'lr' in self.params:
- self._model.optimizer.learning_rate.assign(
- float(self.params['lr']))
+ model = Model(inputs=inputs, outputs=predictions)
+ model.compile(optimizer=params['optimizer'],
+ loss='binary_crossentropy',
+ metrics=['top_k_categorical_accuracy'])
+ if 'lr' in params:
+ model.optimizer.learning_rate.assign(
+ float(params['lr']))
 
  summary = []
- self._model.summary(print_fn=summary.append)
+ model.summary(print_fn=summary.append)
  self.debug("Created model: \n" + "\n".join(summary))
+ return model
 
  def _train(self, corpus, params, jobs=0):
- sources = annif.util.parse_sources(self.params['sources'])
- self._create_model(sources)
+ sources = annif.util.parse_sources(params['sources'])
+ self._model = self._create_model(sources, params)
  self._fit_model(
  corpus,
  epochs=int(params['epochs']),

tests/test_backend_nn_ensemble.py

@@ -56,8 +56,8 @@ def test_nn_ensemble_can_set_lr(registry):
  backend_id='nn_ensemble',
  config_params={'epochs': 1, 'lr': 0.002},
  project=project)
- nn_ensemble._create_model(['dummy-en'])
- assert nn_ensemble._model.optimizer.learning_rate.value() == 0.002
+ model = nn_ensemble._create_model(['dummy-en'], nn_ensemble.params)
+ assert model.optimizer.learning_rate.value() == 0.002
 
 
 def test_set_lmdb_map_size(registry, tmpdir):