def fit(self, train_generator=None, validation_generator=None, **kwargs):
'''
Pass through method to external model after running through pipeline
Optionally overwrite normal method to pass in the generator directly. Used
to speedup training by caching the transformed input before training the
model - avoids downloading, reading, encoding images in every batch
'''
if self.pipeline is None:
raise ModelError('Must set pipeline before fitting')
if self.state['fitted']:
LOGGER.warning('Cannot refit model, skipping operation')
return self
if train_generator is None:
# Explicitly fit only on train split
train_generator = self.pipeline.transform(
X=None,
dataset_split=TRAIN_SPLIT,
return_y=True,
infinite_loop=True,
**self.get_params())
validation_generator = self.pipeline.transform(
X=None,
dataset_split=VALIDATION_SPLIT,
return_y=True,
infinite_loop=True,
**self.get_params())
self._fit(train_generator, validation_generator)
# Mark the state so it doesnt get refit and can now be saved
self.state['fitted'] = True
return self
def predict(self, X, **kwargs):
'''
Pass through method to external model after running through pipeline
'''
if not self.state['fitted']:
raise ModelError('Must fit model before predicting')
transformed = self.pipeline.transform(X, **kwargs)
return self.external_model.predict(transformed)
def save(self, **kwargs):
'''
Extend parent function with a few additional save routines
1) save params
2) save feature metadata
'''
if self.pipeline is None:
raise ModelError('Must set pipeline before saving')
if not self.state['fitted']:
raise ModelError('Must fit model before saving')
self.params = self.get_params(**kwargs)
self.feature_metadata = self.get_feature_metadata(**kwargs)
super(BaseModel, self).save(**kwargs)
# Sqlalchemy updates relationship references after save so reload class
self.pipeline.load(load_externals=False)
def fit(self, **kwargs):
'''
Pass through method to external model after running through pipeline
'''
if self.pipeline is None:
raise ModelError('Must set pipeline before fitting')
if self.state['fitted']:
LOGGER.warning('Cannot refit model, skipping operation')
return self
# Explicitly fit only on train split
X, y = self.pipeline.transform(X=None,
dataset_split=TRAIN_SPLIT,
return_y=True)
# Reduce dimensionality of y if it is only 1 column
self.external_model.fit(X, y.squeeze(), **kwargs)
self.state['fitted'] = True
return self
def assert_fitted(self, msg=""):
"""
Helper method to raise an error if model isn't fit
"""
if not self.fitted:
raise ModelError(msg)
def assert_pipeline(self, msg=""):
"""
Helper method to raise an error if pipeline isn't present and configured
"""
if self.pipeline is None or not self.pipeline.fitted:
raise ModelError(msg)
def assert_fitted(self, msg=''):
'''
Helper method to raise an error if model isn't fit
'''
if not self.fitted:
raise ModelError(msg)