def run_pipeline_from_automl(train_file_path: str,
test_file_path: str,
max_run_time: timedelta = timedelta(minutes=10)):
""" Function run pipeline with Auto ML models in nodes
:param train_file_path: path to the csv file with data for train
:param test_file_path: path to the csv file with data for validation
:param max_run_time: maximum running time for customization of the "tpot" model
:return roc_auc_value: ROC AUC metric for pipeline
"""
train_data = InputData.from_csv(train_file_path)
test_data = InputData.from_csv(test_file_path)
testing_target = test_data.target
node_scaling = PrimaryNode('scaling')
node_tpot = PrimaryNode('tpot')
node_tpot.operation.params = {'max_run_time_sec': max_run_time.seconds}
node_lda = SecondaryNode('lda', nodes_from=[node_scaling])
node_rf = SecondaryNode('rf', nodes_from=[node_tpot, node_lda])
OperationTypesRepository.assign_repo('model', 'automl_repository.json')
pipeline = Pipeline(node_rf)
pipeline.fit(train_data)
results = pipeline.predict(test_data)
roc_auc_value = roc_auc(y_true=testing_target, y_score=results.predict)
print(roc_auc_value)
return roc_auc_value
def test_forecast_with_exog():
train_source_ts, predict_source_ts, train_exog_ts, predict_exog_ts, ts_test = synthetic_with_exogenous_ts(
)
# Source data for lagged node
node_lagged = PrimaryNode('lagged')
# Set window size for lagged transformation
node_lagged.custom_params = {'window_size': window_size}
# Exogenous variable for exog node
node_exog = PrimaryNode('exog_ts_data_source')
node_final = SecondaryNode('linear', nodes_from=[node_lagged, node_exog])
pipeline = Pipeline(node_final)
pipeline.fit(input_data=MultiModalData({
'exog_ts_data_source': train_exog_ts,
'lagged': train_source_ts
}))
forecast = pipeline.predict(
input_data=MultiModalData({
'exog_ts_data_source': predict_exog_ts,
'lagged': predict_source_ts
}))
prediction = np.ravel(np.array(forecast.predict))
assert tuple(prediction) == tuple(ts_test)
def test_save_load_fitted_atomized_pipeline_correctly():
pipeline = create_pipeline_with_several_nested_atomized_model()
train_data, test_data = create_data_for_train()
pipeline.fit(train_data)
json_actual = pipeline.save(
'test_save_load_fitted_atomized_pipeline_correctly')
json_path_load = create_correct_path(
'test_save_load_fitted_atomized_pipeline_correctly')
pipeline_loaded = Pipeline()
pipeline_loaded.load(json_path_load)
json_expected = pipeline_loaded.save(
'test_save_load_fitted_atomized_pipeline_correctly_loaded')
assert pipeline.length == pipeline_loaded.length
assert json_actual == json_expected
before_save_predicted = pipeline.predict(test_data)
pipeline_loaded.fit(train_data)
after_save_predicted = pipeline_loaded.predict(test_data)
bfr_tun_mse = mean_squared_error(y_true=test_data.target,
y_pred=before_save_predicted.predict)
aft_tun_mse = mean_squared_error(y_true=test_data.target,
y_pred=after_save_predicted.predict)
assert aft_tun_mse <= bfr_tun_mse
def run_tpot_vs_fedot_example(train_file_path: str, test_file_path: str):
train_data = InputData.from_csv(train_file_path)
test_data = InputData.from_csv(test_file_path)
training_features = train_data.features
testing_features = test_data.features
training_target = train_data.target
testing_target = test_data.target
# Average CV score on the training set was: 0.93755
exported_pipeline = make_pipeline(
StackingEstimator(estimator=BernoulliNB()), RandomForestClassifier())
# Fix random state for all the steps in exported pipeline
set_param_recursive(exported_pipeline.steps, 'random_state', 1)
exported_pipeline.fit(training_features, training_target)
results = exported_pipeline.predict_proba(testing_features)[:, 1]
roc_auc_value = roc_auc(y_true=testing_target, y_score=results)
print(f'ROC AUC for TPOT: {roc_auc_value}')
node_scaling = PrimaryNode('scaling')
node_bernb = SecondaryNode('bernb', nodes_from=[node_scaling])
node_rf = SecondaryNode('rf', nodes_from=[node_bernb, node_scaling])
pipeline = Pipeline(node_rf)
pipeline.fit(train_data)
results = pipeline.predict(test_data)
roc_auc_value = roc_auc(y_true=testing_target, y_score=results.predict)
print(f'ROC AUC for FEDOT: {roc_auc_value}')
return roc_auc_value
def _get_metric_value(self, pipeline: Pipeline,
metric: MetricByTask) -> float:
pipeline.fit(self._train_data, use_fitted=False)
predicted = pipeline.predict(self._test_data)
metric_value = metric.get_value(true=self._test_data,
predicted=predicted)
return metric_value
def test_data_preparation_for_multi_target_correct(multi_target_data_setup):
train, test = multi_target_data_setup
simple_pipeline = Pipeline(PrimaryNode('linear'))
simple_pipeline.fit(input_data=train)
source_shape = test.target.shape
# Get converted data
results, new_test = QualityMetric()._simple_prediction(
simple_pipeline, test)
number_elements = len(new_test.target)
assert source_shape[0] * source_shape[1] == number_elements
def test_pipeline_with_wrong_data():
pipeline = Pipeline(PrimaryNode('linear'))
data_seq = np.arange(0, 10)
task = Task(TaskTypesEnum.ts_forecasting,
TsForecastingParams(forecast_length=10))
data = InputData(idx=data_seq, features=data_seq, target=data_seq,
data_type=DataTypesEnum.ts, task=task)
with pytest.raises(ValueError):
pipeline.fit(data)
def test_pipeline_unfit(data_fixture, request):
data = request.getfixturevalue(data_fixture)
pipeline = Pipeline(PrimaryNode('logit'))
pipeline.fit(data)
assert pipeline.is_fitted
pipeline.unfit()
assert not pipeline.is_fitted
assert not pipeline.root_node.fitted_operation
with pytest.raises(ValueError) as exc:
assert pipeline.predict(data)
def execute_pipeline_for_text_problem(train_data, test_data):
node_text_clean = PrimaryNode('text_clean')
node_tfidf = SecondaryNode('tfidf', nodes_from=[node_text_clean])
model_node = SecondaryNode('multinb', nodes_from=[node_tfidf])
pipeline = Pipeline(model_node)
pipeline.fit(train_data)
predicted = pipeline.predict(test_data)
roc_auc_metric = roc_auc(y_true=test_data.target,
y_score=predicted.predict)
return roc_auc_metric
def test_import_custom_json_object_to_pipeline_and_fit_correctly_no_exception():
test_file_path = str(os.path.dirname(__file__))
file = '../../data/test_custom_json_template.json'
json_path_load = os.path.join(test_file_path, file)
train_file_path, test_file_path = get_scoring_case_data_paths()
train_data = InputData.from_csv(train_file_path)
pipeline = Pipeline()
pipeline.load(json_path_load)
pipeline.fit(train_data)
pipeline.save('test_import_custom_json_object_to_pipeline_and_fit_correctly_no_exception')
def test_log_clustering_fit_correct(data_fixture, request):
data = request.getfixturevalue(data_fixture)
train_data, test_data = train_test_data_setup(data=data)
# Scaling pipeline. Fit predict it
scaling_pipeline = Pipeline(PrimaryNode('normalization'))
scaling_pipeline.fit(train_data)
scaled_data = scaling_pipeline.predict(train_data)
kmeans = Model(operation_type='kmeans')
_, train_predicted = kmeans.fit(data=scaled_data)
assert all(np.unique(train_predicted.predict) == [0, 1])
def test_secondary_nodes_is_invariant_to_inputs_order(data_setup):
data = data_setup
train, test = train_test_data_setup(data)
first = PrimaryNode(operation_type='logit')
second = PrimaryNode(operation_type='lda')
third = PrimaryNode(operation_type='knn')
final = SecondaryNode(operation_type='xgboost',
nodes_from=[first, second, third])
pipeline = Pipeline()
for node in [first, second, third, final]:
pipeline.add_node(node)
first = deepcopy(first)
second = deepcopy(second)
third = deepcopy(third)
final_shuffled = SecondaryNode(operation_type='xgboost',
nodes_from=[third, first, second])
pipeline_shuffled = Pipeline()
# change order of nodes in list
for node in [final_shuffled, third, first, second]:
pipeline_shuffled.add_node(node)
train_predicted = pipeline.fit(input_data=train)
train_predicted_shuffled = pipeline_shuffled.fit(input_data=train)
# train results should be invariant
assert pipeline.root_node.descriptive_id == pipeline_shuffled.root_node.descriptive_id
assert np.equal(train_predicted.predict, train_predicted_shuffled.predict).all()
test_predicted = pipeline.predict(input_data=test)
test_predicted_shuffled = pipeline_shuffled.predict(input_data=test)
# predict results should be invariant
assert np.equal(test_predicted.predict, test_predicted_shuffled.predict).all()
# change parents order for the nodes fitted pipeline
nodes_for_change = pipeline.nodes[3].nodes_from
pipeline.nodes[3].nodes_from = [nodes_for_change[2], nodes_for_change[0], nodes_for_change[1]]
pipeline.nodes[3].unfit()
pipeline.fit(train)
test_predicted_re_shuffled = pipeline.predict(input_data=test)
# predict results should be invariant
assert np.equal(test_predicted.predict, test_predicted_re_shuffled.predict).all()
def test_multi_modal_pipeline():
task = Task(TaskTypesEnum.classification)
images_size = (128, 128)
files_path = os.path.join('test', 'data', 'multi_modal')
path = os.path.join(str(fedot_project_root()), files_path)
train_num, _, train_img, _, train_text, _ = \
prepare_multi_modal_data(path, task, images_size, with_split=False)
# image
image_node = PrimaryNode('cnn')
image_node.custom_params = {'image_shape': (images_size[0], images_size[1], 1),
'architecture': 'simplified',
'num_classes': 2,
'epochs': 1,
'batch_size': 128}
# image
ds_image = PrimaryNode('data_source_img')
image_node = SecondaryNode('cnn', nodes_from=[ds_image])
image_node.custom_params = {'image_shape': (images_size[0], images_size[1], 1),
'architecture': 'simplified',
'num_classes': 2,
'epochs': 15,
'batch_size': 128}
# table
ds_table = PrimaryNode('data_source_table')
scaling_node = SecondaryNode('scaling', nodes_from=[ds_table])
numeric_node = SecondaryNode('rf', nodes_from=[scaling_node])
# text
ds_text = PrimaryNode('data_source_text')
node_text_clean = SecondaryNode('text_clean', nodes_from=[ds_text])
text_node = SecondaryNode('tfidf', nodes_from=[node_text_clean])
pipeline = Pipeline(SecondaryNode('logit', nodes_from=[numeric_node, image_node, text_node]))
fit_data = MultiModalData({
'data_source_img': train_img,
'data_source_table': train_num,
'data_source_text': train_text
})
pipeline.fit(fit_data)
prediction = pipeline.predict(fit_data)
assert prediction is not None
def composer_metric(self, metrics, train_data: Union[InputData,
MultiModalData],
test_data: Union[InputData, MultiModalData],
pipeline: Pipeline) -> Optional[Tuple[Any]]:
try:
validate(pipeline)
pipeline.log = self.log
if type(metrics) is not list:
metrics = [metrics]
if self.cache is not None:
# TODO improve cache
pipeline.fit_from_cache(self.cache)
if not pipeline.is_fitted:
self.log.debug(
f'Pipeline {pipeline.root_node.descriptive_id} fit started'
)
pipeline.fit(input_data=train_data,
time_constraint=self.composer_requirements.
max_pipeline_fit_time)
try:
self.cache.save_pipeline(pipeline)
except Exception as ex:
self.log.info(f'Cache can not be saved: {ex}. Continue.')
evaluated_metrics = ()
for metric in metrics:
if callable(metric):
metric_func = metric
else:
metric_func = MetricsRepository().metric_by_id(metric)
evaluated_metrics = evaluated_metrics + (metric_func(
pipeline, reference_data=test_data), )
self.log.debug(
f'Pipeline {pipeline.root_node.descriptive_id} with metrics: {list(evaluated_metrics)}'
)
# enforce memory cleaning
pipeline.unfit()
gc.collect()
except Exception as ex:
self.log.info(f'Pipeline assessment warning: {ex}. Continue.')
evaluated_metrics = None
return evaluated_metrics
def test_svc_fit_correct(data_fixture, request):
data = request.getfixturevalue(data_fixture)
train_data, test_data = train_test_data_setup(data=data)
# Scaling pipeline. Fit predict it
scaling_pipeline = Pipeline(PrimaryNode('normalization'))
scaling_pipeline.fit(train_data)
scaled_data = scaling_pipeline.predict(train_data)
svc = Model(operation_type='svc')
_, train_predicted = svc.fit(data=scaled_data)
roc_on_train = get_roc_auc(valid_data=train_data,
predicted_data=train_predicted)
roc_threshold = 0.95
assert roc_on_train >= roc_threshold
def test_pipeline_hierarchy_fit_correct(data_setup):
data = data_setup
train, _ = train_test_data_setup(data)
first = PrimaryNode(operation_type='logit')
second = SecondaryNode(operation_type='logit', nodes_from=[first])
third = SecondaryNode(operation_type='logit', nodes_from=[first])
final = SecondaryNode(operation_type='logit', nodes_from=[second, third])
pipeline = Pipeline()
for node in [first, second, third, final]:
pipeline.add_node(node)
pipeline.unfit()
train_predicted = pipeline.fit(input_data=train)
assert pipeline.root_node.descriptive_id == (
'((/n_logit_default_params;)/'
'n_logit_default_params;;(/'
'n_logit_default_params;)/'
'n_logit_default_params;)/'
'n_logit_default_params')
assert pipeline.length == 4
assert pipeline.depth == 3
assert train_predicted.predict.shape[0] == train.target.shape[0]
assert final.fitted_operation is not None
def test_pca_model_removes_redunant_features_correct():
n_informative = 5
data = classification_dataset_with_redunant_features(
n_samples=1000, n_features=100, n_informative=n_informative)
train_data, test_data = train_test_data_setup(data=data)
# Scaling pipeline. Fit predict it
scaling_pipeline = Pipeline(PrimaryNode('normalization'))
scaling_pipeline.fit(train_data)
scaled_data = scaling_pipeline.predict(train_data)
pca = DataOperation(operation_type='pca')
_, train_predicted = pca.fit(data=scaled_data)
transformed_features = train_predicted.predict
assert transformed_features.shape[1] < data.features.shape[1]
def test_forecast_with_sparse_lagged():
train_source_ts, predict_source_ts, train_exog_ts, predict_exog_ts, ts_test = synthetic_with_exogenous_ts(
)
# Source data for lagged node
node_lagged = PrimaryNode('sparse_lagged')
# Set window size for lagged transformation
node_lagged.custom_params = {'window_size': window_size}
node_final = SecondaryNode('linear', nodes_from=[node_lagged])
pipeline = Pipeline(node_final)
pipeline.fit(input_data=MultiModalData({'sparse_lagged': train_source_ts}))
forecast = pipeline.predict(
input_data=MultiModalData({'sparse_lagged': predict_source_ts}))
is_forecasted = True
assert is_forecasted
def test_pipeline_with_datamodel_fit_correct(data_setup):
data = data_setup
train_data, test_data = train_test_data_setup(data)
pipeline = Pipeline()
node_data = PrimaryNode('logit')
node_first = PrimaryNode('bernb')
node_second = SecondaryNode('rf')
node_second.nodes_from = [node_first, node_data]
pipeline.add_node(node_data)
pipeline.add_node(node_first)
pipeline.add_node(node_second)
pipeline.fit(train_data)
results = np.asarray(probs_to_labels(pipeline.predict(test_data).predict))
assert results.shape == test_data.target.shape
def test_clean_text_preprocessing():
test_text = [
'This is the first document.',
'This document is the second document.',
'And this is the third one.',
'Is this the first document?',
]
input_data = InputData(features=test_text,
target=[0, 1, 1, 0],
idx=np.arange(0, len(test_text)),
task=Task(TaskTypesEnum.classification),
data_type=DataTypesEnum.text)
preprocessing_pipeline = Pipeline(PrimaryNode('text_clean'))
preprocessing_pipeline.fit(input_data)
predicted_output = preprocessing_pipeline.predict(input_data)
cleaned_text = predicted_output.predict
assert len(test_text) == len(cleaned_text)
def test_regression_pipeline_with_data_operation_fit_correct():
data = get_synthetic_regression_data()
train_data, test_data = train_test_data_setup(data)
# linear
# / \
# ridge |
# | |
# ransac_lin_reg lasso
# \ /
# scaling
node_scaling = PrimaryNode('scaling')
node_ransac = SecondaryNode('ransac_lin_reg', nodes_from=[node_scaling])
node_lasso = SecondaryNode('lasso', nodes_from=[node_scaling])
node_ridge = SecondaryNode('ridge', nodes_from=[node_ransac])
node_root = SecondaryNode('linear', nodes_from=[node_lasso, node_ridge])
pipeline = Pipeline(node_root)
pipeline.fit(train_data)
results = pipeline.predict(test_data)
assert results.predict.shape == test_data.target.shape
def test_pipeline_with_custom_params_for_model(data_setup):
data = data_setup
custom_params = dict(n_neighbors=1,
weights='uniform',
p=1)
first = PrimaryNode(operation_type='logit')
second = PrimaryNode(operation_type='lda')
final = SecondaryNode(operation_type='knn', nodes_from=[first, second])
pipeline = Pipeline()
pipeline.add_node(final)
pipeline_default_params = deepcopy(pipeline)
pipeline.root_node.custom_params = custom_params
pipeline_default_params.fit(data)
pipeline.fit(data)
custom_params_prediction = pipeline.predict(data).predict
default_params_prediction = pipeline_default_params.predict(data).predict
assert not np.array_equal(custom_params_prediction, default_params_prediction)
