def test_random_forest_regressor(self, compress_model_definition):
# Train model
training_data = datasets.make_regression(n_features=5)
regressor = RandomForestRegressor()
regressor.fit(training_data[0], training_data[1])
# Serialise the models to Elasticsearch
feature_names = ["f0", "f1", "f2", "f3", "f4"]
model_id = "test_random_forest_regressor"
es_model = MLModel.import_model(
ES_TEST_CLIENT,
model_id,
regressor,
feature_names,
es_if_exists="replace",
es_compress_model_definition=compress_model_definition,
)
# Get some test results
check_prediction_equality(es_model, regressor,
random_rows(training_data[0], 20))
match = f"Trained machine learning model {model_id} already exists"
with pytest.raises(ValueError, match=match):
MLModel.import_model(
ES_TEST_CLIENT,
model_id,
regressor,
feature_names,
es_if_exists="fail",
es_compress_model_definition=compress_model_definition,
)
# Clean up
es_model.delete_model()
def test_predict_single_feature_vector(self):
# Train model
training_data = datasets.make_regression(n_features=1)
regressor = XGBRegressor()
regressor.fit(training_data[0], training_data[1])
# Get some test results
test_data = [[0.1]]
test_results = regressor.predict(np.asarray(test_data))
# Serialise the models to Elasticsearch
feature_names = ["f0"]
model_id = "test_xgb_regressor"
es_model = MLModel.import_model(ES_TEST_CLIENT,
model_id,
regressor,
feature_names,
es_if_exists="replace")
# Single feature
es_results = es_model.predict(test_data[0])
np.testing.assert_almost_equal(test_results, es_results, decimal=2)
# Clean up
es_model.delete_model()
def test_xgb_regressor(self, compress_model_definition, objective,
booster):
# Train model
training_data = datasets.make_regression(n_features=5)
regressor = XGBRegressor(objective=objective, booster=booster)
regressor.fit(
training_data[0],
np.exp(training_data[1] - np.max(training_data[1])) /
sum(np.exp(training_data[1])),
)
# Serialise the models to Elasticsearch
feature_names = ["f0", "f1", "f2", "f3", "f4"]
model_id = "test_xgb_regressor"
es_model = MLModel.import_model(
ES_TEST_CLIENT,
model_id,
regressor,
feature_names,
es_if_exists="replace",
es_compress_model_definition=compress_model_definition,
)
# Get some test results
check_prediction_equality(es_model, regressor,
random_rows(training_data[0], 20))
# Clean up
es_model.delete_model()
def test_xgb_classifier_objectives_and_booster(self, objective, booster):
# test both multiple and binary classification
if objective.startswith("multi"):
skip_if_multiclass_classifition()
training_data = datasets.make_classification(n_features=5,
n_classes=3,
n_informative=3)
classifier = XGBClassifier(booster=booster, objective=objective)
else:
training_data = datasets.make_classification(n_features=5)
classifier = XGBClassifier(booster=booster, objective=objective)
# Train model
classifier.fit(training_data[0], training_data[1])
# Serialise the models to Elasticsearch
feature_names = [
"feature0", "feature1", "feature2", "feature3", "feature4"
]
model_id = "test_xgb_classifier"
es_model = MLModel.import_model(ES_TEST_CLIENT,
model_id,
classifier,
feature_names,
es_if_exists="replace")
# Get some test results
check_prediction_equality(es_model, classifier,
random_rows(training_data[0], 20))
# Clean up
es_model.delete_model()
def test_decision_tree_classifier(self, compress_model_definition):
# Train model
training_data = datasets.make_classification(n_features=5)
classifier = DecisionTreeClassifier()
classifier.fit(training_data[0], training_data[1])
# Serialise the models to Elasticsearch
feature_names = ["f0", "f1", "f2", "f3", "f4"]
model_id = "test_decision_tree_classifier"
es_model = MLModel.import_model(
ES_TEST_CLIENT,
model_id,
classifier,
feature_names,
es_if_exists="replace",
es_compress_model_definition=compress_model_definition,
)
# Get some test results
check_prediction_equality(es_model, classifier,
random_rows(training_data[0], 20))
# Clean up
es_model.delete_model()
def test_xgb_classifier(self, compress_model_definition, multi_class):
# test both multiple and binary classification
if multi_class:
skip_if_multiclass_classifition()
training_data = datasets.make_classification(n_features=5,
n_classes=3,
n_informative=3)
classifier = XGBClassifier(booster="gbtree",
objective="multi:softmax")
else:
training_data = datasets.make_classification(n_features=5)
classifier = XGBClassifier(booster="gbtree")
# Train model
classifier.fit(training_data[0], training_data[1])
# Serialise the models to Elasticsearch
feature_names = ["f0", "f1", "f2", "f3", "f4"]
model_id = "test_xgb_classifier"
es_model = MLModel.import_model(
ES_TEST_CLIENT,
model_id,
classifier,
feature_names,
es_if_exists="replace",
es_compress_model_definition=compress_model_definition,
)
# Get some test results
check_prediction_equality(es_model, classifier,
random_rows(training_data[0], 20))
# Clean up
es_model.delete_model()
def test_es_if_exists_fail(self, compress_model_definition):
# Train model
training_data = datasets.make_regression(n_features=5)
regressor = RandomForestRegressor()
regressor.fit(training_data[0], training_data[1])
feature_names = ["f0", "f1", "f2", "f3", "f4"]
model_id = "test_random_forest_regressor"
# If both overwrite and es_if_exists is given.
match = f"Trained machine learning model {model_id} already exists"
with pytest.raises(ValueError, match=match):
MLModel.import_model(
ES_TEST_CLIENT,
model_id,
regressor,
feature_names,
es_if_exists="fail",
es_compress_model_definition=compress_model_definition,
)
def test_imported_mlmodel_overwrite_true(self, compress_model_definition,
overwrite):
# Train model
training_data = datasets.make_regression(n_features=5)
regressor = RandomForestRegressor()
regressor.fit(training_data[0], training_data[1])
feature_names = ["f0", "f1", "f2", "f3", "f4"]
model_id = "test_random_forest_regressor"
match = "'overwrite' parameter is deprecated, use 'es_if_exists' instead"
with pytest.warns(DeprecationWarning, match=match):
MLModel.import_model(
ES_TEST_CLIENT,
model_id,
regressor,
feature_names,
overwrite=overwrite,
es_compress_model_definition=compress_model_definition,
)
def test_imported_mlmodel_bothparams(self, compress_model_definition,
es_if_exists, overwrite):
# Train model
training_data = datasets.make_regression(n_features=5)
regressor = RandomForestRegressor()
regressor.fit(training_data[0], training_data[1])
feature_names = ["f0", "f1", "f2", "f3", "f4"]
model_id = "test_random_forest_regressor"
match = "Using 'overwrite' and 'es_if_exists' together is invalid, use only 'es_if_exists'"
with pytest.raises(ValueError, match=match):
MLModel.import_model(
ES_TEST_CLIENT,
model_id,
regressor,
feature_names,
es_if_exists=es_if_exists,
overwrite=overwrite,
es_compress_model_definition=compress_model_definition,
)
def test_imported_mlmodel_overwrite_false(self, compress_model_definition,
overwrite):
# Train model
training_data = datasets.make_regression(n_features=5)
regressor = RandomForestRegressor()
regressor.fit(training_data[0], training_data[1])
feature_names = ["f0", "f1", "f2", "f3", "f4"]
model_id = "test_random_forest_regressor"
match_error = f"Trained machine learning model {model_id} already exists"
match_warning = (
"'overwrite' parameter is deprecated, use 'es_if_exists' instead")
with pytest.raises(ValueError, match=match_error):
with pytest.warns(DeprecationWarning, match=match_warning):
MLModel.import_model(
ES_TEST_CLIENT,
model_id,
regressor,
feature_names,
overwrite=overwrite,
es_compress_model_definition=compress_model_definition,
)
def test_unpack_and_raise_errors_in_ingest_simulate(self, mocker):
# Train model
training_data = datasets.make_classification(n_features=5)
classifier = DecisionTreeClassifier()
classifier.fit(training_data[0], training_data[1])
# Serialise the models to Elasticsearch
feature_names = ["f0", "f1", "f2", "f3", "f4"]
model_id = "test_decision_tree_classifier"
test_data = [[0.1, 0.2, 0.3, -0.5, 1.0], [1.6, 2.1, -10, 50, -1.0]]
es_model = MLModel.import_model(
ES_TEST_CLIENT,
model_id,
classifier,
feature_names,
es_if_exists="replace",
es_compress_model_definition=True,
)
# Mock the ingest.simulate API to return an error within {'docs': [...]}
mock = mocker.patch.object(ES_TEST_CLIENT.ingest, "simulate")
mock.return_value = {
"docs": [{
"error": {
"type":
"x_content_parse_exception",
"reason":
"[1:1052] [inference_model_definition] failed to parse field [trained_model]",
}
}]
}
with pytest.raises(RuntimeError) as err:
es_model.predict(test_data)
assert repr(err.value) == (
'RuntimeError("Failed to run prediction for model ID '
"'test_decision_tree_classifier'\", {'type': 'x_content_parse_exception', "
"'reason': '[1:1052] [inference_model_definition] failed to parse "
"field [trained_model]'})")
def test_lgbm_classifier_objectives_and_booster(self,
compress_model_definition,
objective, booster):
# test both multiple and binary classification
if objective.startswith("multi"):
skip_if_multiclass_classifition()
training_data = datasets.make_classification(n_features=5,
n_classes=3,
n_informative=3)
classifier = LGBMClassifier(boosting_type=booster,
objective=objective)
else:
training_data = datasets.make_classification(n_features=5)
classifier = LGBMClassifier(boosting_type=booster,
objective=objective)
# Train model
classifier.fit(training_data[0], training_data[1])
# Serialise the models to Elasticsearch
feature_names = [
"Column_0", "Column_1", "Column_2", "Column_3", "Column_4"
]
model_id = "test_lgbm_classifier"
es_model = MLModel.import_model(
ES_TEST_CLIENT,
model_id,
classifier,
feature_names,
es_if_exists="replace",
es_compress_model_definition=compress_model_definition,
)
check_prediction_equality(es_model, classifier,
random_rows(training_data[0], 20))
# Clean up
es_model.delete_model()
def test_lgbm_regressor(self, compress_model_definition, objective,
booster):
# Train model
training_data = datasets.make_regression(n_features=5)
if booster == "rf":
regressor = LGBMRegressor(
boosting_type=booster,
objective=objective,
bagging_fraction=0.5,
bagging_freq=3,
)
else:
regressor = LGBMRegressor(boosting_type=booster,
objective=objective)
regressor.fit(training_data[0], training_data[1])
# Serialise the models to Elasticsearch
feature_names = [
"Column_0", "Column_1", "Column_2", "Column_3", "Column_4"
]
model_id = "test_lgbm_regressor"
es_model = MLModel.import_model(
ES_TEST_CLIENT,
model_id,
regressor,
feature_names,
es_if_exists="replace",
es_compress_model_definition=compress_model_definition,
)
# Get some test results
check_prediction_equality(es_model, regressor,
random_rows(training_data[0], 20))
# Clean up
es_model.delete_model()