def test_train_test_provided():
"""Assert that input train, test works."""
dataset = pd.concat([bin_train, bin_test]).reset_index(drop=True)
trainer = DirectClassifier("LR", random_state=1)
trainer.run(bin_train, bin_test)
assert trainer.dataset.equals(dataset)
def test_input_data_in_training():
"""Assert that the data does not change once in a training pipeline."""
train = bin_train.copy()
trainer = DirectClassifier("LR", random_state=1)
trainer.run(train, bin_test)
train.iloc[3, 2] = 99 # Change an item of the original variable
assert 99 not in trainer.dataset # Is unchanged in the pipeline
def test_load():
"""Assert that a trainer is loaded correctly."""
trainer = DirectClassifier("LR", random_state=1)
trainer.save(FILE_DIR + "trainer")
trainer2 = ATOMLoader(FILE_DIR + "trainer")
assert trainer2.__class__.__name__ == "DirectClassifier"
def test_plot():
"""Assert that plotting the BO runs without errors."""
trainer = DirectClassifier(
models=["lSVM", "kSVM", "MLP"],
n_calls=45,
n_initial_points=20,
bo_params={"plot": True},
random_state=1,
)
trainer.run(bin_train, bin_test)
def test_est_params_per_model():
"""Assert that est_params passes the parameters per model."""
trainer = DirectClassifier(
models=["XGB", "LGB"],
est_params={"xgb": {"n_estimators": 100}, "lgb": {"n_estimators": 200}},
random_state=1,
)
trainer.run(bin_train, bin_test)
assert trainer.xgb.estimator.get_params()["n_estimators"] == 100
assert trainer.lgb.estimator.get_params()["n_estimators"] == 200
def test_all_callbacks():
"""Assert that all callbacks predefined callbacks work as intended."""
trainer = DirectClassifier(
models="LR",
n_calls=2,
n_initial_points=2,
bo_params={"max_time": 50, "delta_x": 5, "delta_y": 5},
random_state=1,
)
trainer.run(bin_train, bin_test)
def test_callbacks(callbacks):
"""Assert that custom callbacks are accepted."""
trainer = DirectClassifier(
models="LR",
n_calls=2,
n_initial_points=2,
bo_params={"callbacks": callbacks},
random_state=1,
)
trainer.run(bin_train, bin_test)
def test_close_plot_after_error():
"""Assert that the BO plot is closed after an error."""
trainer = DirectClassifier(
models=["LR", "LDA"],
n_calls=4,
n_initial_points=[2, 5],
bo_params={"plot": True},
random_state=1,
)
trainer.run(bin_train, bin_test)
assert PlotCallback.c == 1 # First model is 0, after error passes to 1
def test_est_params_all_models():
"""Assert that est_params passes the parameters to all models."""
trainer = DirectClassifier(
models=["XGB", "LGB"],
n_calls=5,
est_params={"n_estimators": 220},
random_state=1,
)
trainer.run(bin_train, bin_test)
assert trainer.lgb.estimator.get_params()["n_estimators"] == 220
assert trainer.xgb.estimator.get_params()["n_estimators"] == 220
def test_optimizer_kwargs():
"""Assert that the kwargs provided are passed to the optimizer."""
trainer = DirectClassifier(
models="LR",
n_calls=2,
n_initial_points=2,
bo_params={"acq_func": "EI"},
random_state=1,
)
trainer.run(bin_train, bin_test)
assert trainer._bo_kwargs.get("acq_func") == "EI"
def test_4_data_provided():
"""Assert that input X_train, X_test, y_train, y_test works."""
dataset = pd.concat([bin_train, bin_test]).reset_index(drop=True)
X_train = bin_train.iloc[:, :-1]
X_test = bin_test.iloc[:, :-1]
y_train = bin_train.iloc[:, -1]
y_test = bin_test.iloc[:, -1]
trainer = DirectClassifier("LR", random_state=1)
trainer.run(X_train, X_test, y_train, y_test)
assert trainer.dataset.equals(dataset)
def test_error_handling():
"""Assert that models with errors are removed from the pipeline."""
trainer = DirectClassifier(
models=["LR", "LDA"],
n_calls=4,
n_initial_points=[2, 5],
random_state=1,
)
trainer.run(bin_train, bin_test)
assert trainer.errors.get("LDA")
assert "LDA" not in trainer.models
assert "LDA" not in trainer.results.index
def test_custom_dimensions_all_models():
"""Assert that the custom dimensions are for all models if not dict."""
trainer = DirectClassifier(
models=["LR1", "LR2"],
n_calls=2,
n_initial_points=2,
bo_params={"dimensions": [Integer(100, 1000, name="max_iter")]},
random_state=1,
)
trainer.run(bin_train, bin_test)
assert list(trainer.lr1.best_params.keys()) == ["max_iter"]
assert list(trainer.lr2.best_params.keys()) == ["max_iter"]
def test_sequence_parameters():
"""Assert that every model get his corresponding parameters."""
trainer = DirectClassifier(
models=["LR", "Tree", "LGB"],
n_calls=(2, 3, 4),
n_initial_points=(1, 2, 3),
bagging=[2, 5, 7],
random_state=1,
)
trainer.run(bin_train, bin_test)
assert len(trainer.LR.bo) == 2
assert sum(trainer.tree.bo.index.str.startswith("Initial")) == 2
assert len(trainer.lgb.metric_bagging) == 7
def test_goals_trainers():
"""Assert that the goal of every Trainer class is set correctly."""
trainer = DirectClassifier("LR")
assert trainer.goal == "classification"
trainer = DirectRegressor("OLS")
assert trainer.goal == "regression"
def test_only_task_models():
"""Assert that an error is raised for models at invalid task."""
trainer = DirectClassifier("OLS", random_state=1) # Only regression
pytest.raises(ValueError, trainer.run, bin_train, bin_test)
trainer = DirectRegressor("LDA", random_state=1) # Only classification
pytest.raises(ValueError, trainer.run, reg_train, reg_test)
def test_custom_dimensions_per_model():
"""Assert that the custom dimensions are distributed over the models."""
trainer = DirectClassifier(
models=["LR1", "LR2"],
n_calls=2,
n_initial_points=2,
bo_params={
"dimensions": {
"lr1": [Integer(100, 200, name="max_iter")],
"lr2": [Integer(300, 400, name="max_iter")],
},
},
random_state=1,
)
trainer.run(bin_train, bin_test)
assert 100 <= trainer.lr1.best_params["max_iter"] <= 200
assert 300 <= trainer.lr2.best_params["max_iter"] <= 400
def test_invalid_sequence_parameter():
"""Assert that an error is raised for parameters with the wrong length."""
trainer = DirectClassifier(
models="LR",
metric=f1_score,
needs_proba=[True, False],
random_state=1,
)
pytest.raises(ValueError, trainer.run, bin_train, bin_test)
def test_data_already_set():
"""Assert that if there already is data, the call to run can be empty."""
dataset = pd.concat([bin_train, bin_test]).reset_index(drop=True)
trainer = DirectClassifier("LR", random_state=1)
trainer.run(bin_train, bin_test)
trainer.run()
assert trainer.dataset.equals(dataset)
assert trainer.branch.idx == [len(bin_train), len(bin_test)]
def test_default_metric():
"""Assert that a default metric is assigned depending on the task."""
trainer = DirectClassifier("LR", random_state=1)
trainer.run(bin_train, bin_test)
assert trainer.metric == "f1"
trainer = DirectClassifier("LR", random_state=1)
trainer.run(class_train, class_test)
assert trainer.metric == "f1_weighted"
trainer = DirectRegressor("LGB", random_state=1)
trainer.run(reg_train, reg_test)
assert trainer.metric == "r2"
def test_duplicate_models():
"""Assert that duplicate inputs are ignored."""
trainer = DirectClassifier(["lr", "LR", "lgb"], random_state=1)
trainer.run(bin_train, bin_test)
assert len(trainer.models) == 2
def test_invalid_model_name():
"""Assert that an error is raised when the model is unknown."""
trainer = DirectClassifier(models="invalid", random_state=1)
pytest.raises(ValueError, trainer.run, bin_train, bin_test)
def test_multiple_same_models():
"""Assert that the same model can used with different names."""
trainer = DirectClassifier(["lr", "lr2", "lr_3"], random_state=1)
trainer.run(bin_train, bin_test)
assert trainer.models == ["LR", "LR2", "LR_3"]
def test_models_get_right_name():
"""Assert that the model names are transformed to the correct acronyms."""
trainer = DirectClassifier(["lR", "tReE"], random_state=1)
trainer.run(bin_train, bin_test)
assert trainer.models == ["LR", "Tree"]
def test_all_models_failed():
"""Assert that an error is raised when all models failed."""
trainer = DirectClassifier("LR", n_calls=4, n_initial_points=5, random_state=1)
pytest.raises(RuntimeError, trainer.run, bin_train, bin_test)
def test_default_mapping_assignment():
"""Assert that a default mapping is assigned."""
trainer = DirectClassifier("LR", random_state=1)
trainer.run(bin_train, bin_test)
assert trainer.mapping == {"0": 0, "1": 1}
def test_scorer_metric_parameter():
"""Assert that a scorer metric works."""
trainer = DirectClassifier("LR", metric=get_scorer("f1"), random_state=1)
trainer.run(bin_train, bin_test)
assert trainer.metric == "f1"
def test_model_is_custom():
"""Assert that custom models are accepted."""
trainer = DirectClassifier(RandomForestClassifier, random_state=1)
trainer.run(bin_train, bin_test)
assert trainer.models == "RFC"
def test_sequence_parameters_invalid_length():
"""Assert that an error is raised when the length is invalid."""
trainer = DirectClassifier("LR", n_calls=(2, 2), random_state=1)
pytest.raises(ValueError, trainer.run, bin_train, bin_test)
def test_function_metric_parameter():
"""Assert that a function metric works."""
trainer = DirectClassifier("LR", metric=f1_score, random_state=1)
trainer.run(bin_train, bin_test)
assert trainer.metric == "f1_score"
