def test_success_trial_to_json():
hp = HyperparameterSamples({'a': 2})
repo = InMemoryHyperparamsRepository()
trial = Trial(save_trial_function=repo.save_trial,
hyperparams=hp,
main_metric_name='mse')
with trial:
given_success_trial_validation_split(trial)
trial_json = trial.to_json()
assert trial_json['status'] == TRIAL_STATUS.SUCCESS.value
assert trial_json['error'] is None
assert trial_json['error_traceback'] is None
assert trial_json['main_metric_name'] == trial.main_metric_name
assert then_success_trial_split_json_is_valid(
trial_json['validation_splits'][0])
start_time = datetime.datetime.strptime(trial_json['start_time'],
TRIAL_DATETIME_STR_FORMAT)
end_time = datetime.datetime.strptime(
trial_json['end_time'],
TRIAL_DATETIME_STR_FORMAT) + datetime.timedelta(hours=1)
assert start_time < end_time
def test_failure_trial_to_json():
hp = HyperparameterSamples({'a': 2})
repo = InMemoryHyperparamsRepository()
trial = Trial(save_trial_function=repo.save_trial,
hyperparams=hp,
main_metric_name='mse')
with trial:
trial_split = given_failed_trial_split(trial)
trial_json = trial.to_json()
assert trial_json['status'] == TRIAL_STATUS.FAILED.value
assert trial_json['error'] == str(trial_split.error)
assert trial_json['error_traceback'] == EXPECTED_ERROR_TRACEBACK
assert trial_json['main_metric_name'] == trial.main_metric_name
assert then_failed_validation_split_json_is_valid(
trial_json['validation_splits'][0], trial_split=trial_split)
start_time = datetime.datetime.strptime(trial_json['start_time'],
TRIAL_DATETIME_STR_FORMAT)
end_time = datetime.datetime.strptime(
trial_json['end_time'],
TRIAL_DATETIME_STR_FORMAT) + datetime.timedelta(hours=1)
assert start_time < end_time
def test_trial_split_is_new_best_score_should_return_true_with_a_new_best_score_after_multiple_scores(
):
hp = HyperparameterSamples({'a': 2})
repo = InMemoryHyperparamsRepository()
trial = Trial(save_trial_function=repo.save_trial,
hyperparams=hp,
main_metric_name=MAIN_METRIC_NAME)
with trial.new_validation_split(Identity()) as trial_split:
trial_split.add_metric_results_train(name=MAIN_METRIC_NAME,
score=0.5,
higher_score_is_better=False)
trial_split.add_metric_results_validation(name=MAIN_METRIC_NAME,
score=0.5,
higher_score_is_better=False)
trial_split.add_metric_results_train(name=MAIN_METRIC_NAME,
score=0.7,
higher_score_is_better=False)
trial_split.add_metric_results_validation(name=MAIN_METRIC_NAME,
score=0.7,
higher_score_is_better=False)
trial_split.add_metric_results_train(name=MAIN_METRIC_NAME,
score=0.4,
higher_score_is_better=False)
trial_split.add_metric_results_validation(name=MAIN_METRIC_NAME,
score=0.4,
higher_score_is_better=False)
assert trial_split.is_new_best_score()
def save_score_for_success_trial(self, hyperparams: HyperparameterSamples, score: float):
self.trials.append(Trial(hyperparams, score, TRIAL_STATUS.SUCCESS))
if self.print_success_trial:
self.print_func('score: {}'.format(score))
self.print_func(
'hyperparams:\n{}'.format(json.dumps(hyperparams.to_nested_dict(), sort_keys=True, indent=4)))
def setup(self):
self.hp = HyperparameterSamples({'a': 2})
self.repo = InMemoryHyperparamsRepository()
self.trial = Trial(trial_number=0,
save_trial_function=self.repo.save_trial,
hyperparams=self.hp,
main_metric_name=MAIN_METRIC_NAME)
def train(self, pipeline: BaseStep, data_inputs, expected_outputs=None) -> Trial:
"""
Train pipeline using the validation splitter.
Track training, and validation metrics for each epoch.
Note: the present method is just a shortcut to using the `execute_trial` method with less boilerplate code needed.
Refer to `execute_trial` for full flexibility
:param pipeline: pipeline to train on
:param data_inputs: data inputs
:param expected_outputs: expected ouptuts to fit on
:return: executed trial
"""
validation_splits: List[Tuple[DataContainer, DataContainer]] = self.validation_split_function.split_data_container(
DataContainer(data_inputs=data_inputs, expected_outputs=expected_outputs)
)
repo_trial: Trial = Trial(
pipeline=pipeline,
hyperparams=pipeline.get_hyperparams(),
main_metric_name=self.get_main_metric_name()
)
self.execute_trial(
pipeline=pipeline,
trial_number=1,
repo_trial=repo_trial,
context=ExecutionContext(),
validation_splits=validation_splits,
n_trial=1,
delete_pipeline_on_completion=False
)
return repo_trial
def test_success_trial_split_to_json():
hp = HyperparameterSamples({'a': 2})
trial = Trial(hyperparams=hp, main_metric_name=MAIN_METRIC_NAME)
with trial:
trial_split = given_success_trial_validation_split(trial)
trial_json = trial_split.to_json()
then_success_trial_split_json_is_valid(trial_json)
def test_failure_trial_split_to_json():
hp = HyperparameterSamples({'a': 2})
trial = Trial(hyperparams=hp, main_metric_name='mse')
with trial:
trial_split = given_failed_trial_split(trial)
trial_json = trial_split.to_json()
then_failed_validation_split_json_is_valid(trial_json, trial_split)
def test_trials_get_best_hyperparams_should_return_hyperparams_of_best_trial():
# Given
hp_trial_1 = HyperparameterSamples({'a': 2})
trial_1 = Trial(hyperparams=hp_trial_1, main_metric_name=MAIN_METRIC_NAME)
with trial_1:
given_success_trial_validation_split(trial_1, best_score=0.2)
hp_trial_2 = HyperparameterSamples({'b': 3})
trial_2 = Trial(hyperparams=hp_trial_2, main_metric_name=MAIN_METRIC_NAME)
with trial_2:
given_success_trial_validation_split(trial_2, best_score=0.1)
trials = Trials(trials=[trial_1, trial_2])
# When
best_hyperparams = trials.get_best_hyperparams()
# Then
assert best_hyperparams == hp_trial_2
def test_success_trial_get_validation_score():
hp = HyperparameterSamples({'a': 2})
trial = Trial(hyperparams=hp, main_metric_name='mse')
with trial:
given_success_trial_validation_split(trial, best_score=0.3)
validation_score = trial.get_validation_score()
assert validation_score == 0.3
def test_trial_should_create_new_split():
hp = HyperparameterSamples({'a': 2})
trial = Trial(hyperparams=hp, main_metric_name=MAIN_METRIC_NAME)
with trial.new_validation_split(Identity()) as trial_split:
trial_split.set_success()
assert isinstance(trial_split.start_time, datetime.datetime)
assert isinstance(trial_split.end_time, datetime.datetime)
assert trial_split.start_time < trial_split.end_time
assert trial.validation_splits[0] == trial_split
def test_success_trial_multiple_splits_should_average_the_scores():
hp = HyperparameterSamples({'a': 2})
trial = Trial(hyperparams=hp, main_metric_name='mse')
with trial:
given_success_trial_validation_split(trial, best_score=0.3)
given_success_trial_validation_split(trial, best_score=0.1)
validation_score = trial.get_validation_score()
assert validation_score == 0.2
def new_trial(self, auto_ml_container: 'AutoMLContainer') -> 'Trial':
"""
Create a new trial with the best next hyperparams.
:param auto_ml_container: auto ml data container
:return: trial
"""
hyperparams = self.hyperparameter_selection_strategy.find_next_best_hyperparams(auto_ml_container)
self.print_func('new trial:\n{}'.format(json.dumps(hyperparams.to_nested_dict(), sort_keys=True, indent=4)))
return Trial(hyperparams=hyperparams, main_metric_name=auto_ml_container.main_scoring_metric_name)
def test_success_trial_split_to_json():
hp = HyperparameterSamples({'a': 2})
repo = InMemoryHyperparamsRepository()
trial = Trial(save_trial_function=repo.save_trial,
hyperparams=hp,
main_metric_name=MAIN_METRIC_NAME)
with trial:
trial_split = given_success_trial_validation_split(trial)
trial_json = trial_split.to_json()
then_success_trial_split_json_is_valid(trial_json)
def test_trial_with_failed_split_should_only_average_successful_splits():
hp = HyperparameterSamples({'a': 2})
trial = Trial(hyperparams=hp, main_metric_name='mse')
with trial:
given_success_trial_validation_split(trial, best_score=0.3)
given_success_trial_validation_split(trial, best_score=0.1)
given_failed_trial_split(trial)
validation_score = trial.get_validation_score()
assert validation_score == 0.2
def test_failure_trial_split_to_json():
hp = HyperparameterSamples({'a': 2})
repo = InMemoryHyperparamsRepository()
trial = Trial(save_trial_function=repo.save_trial,
hyperparams=hp,
main_metric_name='mse')
with trial:
trial_split = given_failed_trial_split(trial)
trial_json = trial_split.to_json()
then_failed_validation_split_json_is_valid(trial_json, trial_split)
def new_trial(self, auto_ml_container: 'AutoMLContainer'):
"""
Create new hyperperams trial json file.
:param auto_ml_container: auto ml container
:return:
"""
hyperparams = self.hyperparameter_selection_strategy.find_next_best_hyperparams(auto_ml_container)
trial = Trial(hyperparams, cache_folder=self.cache_folder,
main_metric_name=auto_ml_container.main_scoring_metric_name)
self._create_trial_json(trial=trial)
return trial
def test_success_trial_get_validation_score():
hp = HyperparameterSamples({'a': 2})
repo = InMemoryHyperparamsRepository()
trial = Trial(save_trial_function=repo.save_trial,
hyperparams=hp,
main_metric_name='mse')
with trial:
given_success_trial_validation_split(trial, best_score=0.3)
validation_score = trial.get_validation_score()
assert validation_score == 0.3
def test_trial_split_is_new_best_score_should_return_true_with_one_score():
hp = HyperparameterSamples({'a': 2})
trial = Trial(hyperparams=hp, main_metric_name=MAIN_METRIC_NAME)
with trial.new_validation_split(Identity()) as trial_split:
trial_split.add_metric_results_train(name=MAIN_METRIC_NAME,
score=0.5,
higher_score_is_better=False)
trial_split.add_metric_results_validation(name=MAIN_METRIC_NAME,
score=0.5,
higher_score_is_better=False)
assert trial_split.is_new_best_score()
def new_trial(self, auto_ml_container: 'AutoMLContainer') -> 'Trial':
hyperparams = self.hyperparameter_selection_strategy.find_next_best_hyperparams(
auto_ml_container)
logger = self._create_logger_for_trial(auto_ml_container.trial_number)
logger.info('\nnew trial: {}'.format(
json.dumps(hyperparams.to_nested_dict(), sort_keys=True,
indent=4)))
return Trial(
cache_folder=self.cache_folder,
save_trial_function=self.save_trial,
logger=logger,
hyperparams=hyperparams,
main_metric_name=auto_ml_container.main_scoring_metric_name)
def new_trial(self, auto_ml_container: 'AutoMLContainer') -> Trial:
"""
Create a new trial with the best next hyperparams.
:param context:
:param auto_ml_container: auto ml data container
:return: trial
"""
hyperparams = self.hyperparameter_selection_strategy.find_next_best_hyperparams(
auto_ml_container)
trial = Trial(
trial_number=auto_ml_container.trial_number,
hyperparams=hyperparams,
save_trial_function=self.save_trial,
cache_folder=self.cache_folder,
main_metric_name=auto_ml_container.main_scoring_metric_name)
return trial
def train(self,
pipeline: BaseStep,
data_inputs,
expected_outputs=None,
context: ExecutionContext = None,
trial_number=0) -> Trial:
"""
Train pipeline using the validation splitter.
Track training, and validation metrics for each epoch.
Note: the present method is just a shortcut to using the `execute_trial` method with less boilerplate code needed. Refer to `execute_trial` for full flexibility
:param pipeline: pipeline to train on
:param data_inputs: data inputs
:param expected_outputs: expected ouptuts to fit on
:return: executed trial
"""
assert not (
context is None
) # TODO: change order of arguments so that context isn't an optional argument
validation_splits: List[Tuple[
DataContainer,
DataContainer]] = self.validation_split_function.split_data_container(
DataContainer(data_inputs=data_inputs,
expected_outputs=expected_outputs),
context=context)
repo_trial: Trial = Trial(
pipeline=pipeline,
logger=context.logger,
hyperparams=pipeline.get_hyperparams(),
main_metric_name=self.get_main_metric_name(),
save_trial_function=self.hyperparams_repository.save_trial,
trial_number=trial_number)
self.execute_trial(pipeline=pipeline,
repo_trial=repo_trial,
context=context,
validation_splits=validation_splits,
n_trial=1,
delete_pipeline_on_completion=False)
return repo_trial
def test_trials_get_best_hyperparams_should_return_hyperparams_of_best_trial(
self):
# Given
trial_1 = self.trial
with trial_1:
self._given_success_trial_validation_split(trial_1, best_score=0.2)
hp_trial_2 = HyperparameterSamples({'b': 3})
trial_2 = Trial(trial_number=1,
save_trial_function=self.repo.save_trial,
hyperparams=hp_trial_2,
main_metric_name=MAIN_METRIC_NAME)
with trial_2:
self._given_success_trial_validation_split(trial_2, best_score=0.1)
trials = Trials(trials=[trial_1, trial_2])
# When
best_hyperparams = trials.get_best_hyperparams()
# Then
assert best_hyperparams == hp_trial_2
def from_json(trial_json) -> 'Trial':
return Trial(
hyperparams=trial_json['hyperparams'],
score=trial_json['score'],
status=trial_json['status']
)