def setup_method(self, method):
# super().setup_method(method)
self.model = XGBoost(config={
'n_estimators': 5,
'max_depth': 2,
'tree_method': 'hist'
})
feature_cols = ["f", "f2"]
target_col = "t"
train_df = pd.DataFrame({
"f": np.random.randn(20),
"f2": np.random.randn(20),
"t": np.random.randint(20)
})
val_df = pd.DataFrame({
"f": np.random.randn(5),
"f2": np.random.randn(5),
"t": np.random.randint(5)
})
ft = IdentityTransformer(feature_cols=feature_cols,
target_col=target_col)
self.x, self.y = ft.transform(train_df)
self.val_x, self.val_y = ft.transform(val_df)
def build(self, config):
from zoo.orca.automl.xgboost.XGBoost import XGBoost
model = XGBoost(model_type=self.model_type, config=config)
if self.n_cpus is not None:
model.set_params(n_jobs=self.n_cpus)
return model
class TestXgbregressor(ZooTestCase):
def setup_method(self, method):
# super().setup_method(method)
self.model = XGBoost(config={
'n_estimators': 5,
'max_depth': 2,
'tree_method': 'hist'
})
feature_cols = ["f", "f2"]
target_col = "t"
train_df = pd.DataFrame({
"f": np.random.randn(20),
"f2": np.random.randn(20),
"t": np.random.randint(20)
})
val_df = pd.DataFrame({
"f": np.random.randn(5),
"f2": np.random.randn(5),
"t": np.random.randint(5)
})
ft = IdentityTransformer(feature_cols=feature_cols,
target_col=target_col)
self.x, self.y = ft.transform(train_df)
self.val_x, self.val_y = ft.transform(val_df)
def teardown_method(self, method):
pass
def test_fit_predict_evaluate(self):
self.model.fit_eval((self.x, self.y), [(self.val_x, self.val_y)])
# test predict
result = self.model.predict(self.val_x)
# test evaluate
evaluate_result = self.model.evaluate(self.val_x, self.val_y)
def test_save_restore(self):
self.model.fit_eval((self.x, self.y), [(self.val_x, self.val_y)])
result_save = self.model.predict(self.val_x)
model_file = "tmp.pkl"
self.model.save(model_file)
assert os.path.isfile(model_file)
new_model = XGBoost()
new_model.restore(model_file)
assert new_model.model
result_restore = new_model.predict(self.val_x)
assert_array_almost_equal(result_save, result_restore, decimal=2), \
"Prediction values are not the same after restore: " \
"predict before is {}, and predict after is {}".format(result_save, result_restore)
os.remove(model_file)
def test_save_restore(self):
self.model.fit_eval((self.x, self.y), [(self.val_x, self.val_y)])
result_save = self.model.predict(self.val_x)
model_file = "tmp.pkl"
self.model.save(model_file)
assert os.path.isfile(model_file)
new_model = XGBoost()
new_model.restore(model_file)
assert new_model.model
result_restore = new_model.predict(self.val_x)
assert_array_almost_equal(result_save, result_restore, decimal=2), \
"Prediction values are not the same after restore: " \
"predict before is {}, and predict after is {}".format(result_save, result_restore)
os.remove(model_file)
def load_xgboost_pipeline(file, model_type="regressor"):
from zoo.zouwu.feature.identity_transformer import IdentityTransformer
from zoo.orca.automl.xgboost.XGBoost import XGBoost
feature_transformers = IdentityTransformer()
model = XGBoost(model_type=model_type)
all_config = restore_zip(file, feature_transformers, model)
ts_pipeline = TimeSequencePipeline(feature_transformers=feature_transformers,
model=model,
config=all_config)
print("Restore pipeline from", file)
return ts_pipeline
class TestXgbregressor(ZooTestCase):
def setup_method(self, method):
# super().setup_method(method)
self.model = XGBoost(config={
'n_estimators': 5,
'max_depth': 2,
'tree_method': 'hist'
})
feature_cols = ["f", "f2"]
target_col = "t"
train_df = pd.DataFrame({
"f": np.random.randn(20),
"f2": np.random.randn(20),
"t": np.random.randint(20)
})
val_df = pd.DataFrame({
"f": np.random.randn(5),
"f2": np.random.randn(5),
"t": np.random.randint(5)
})
ft = IdentityTransformer(feature_cols=feature_cols,
target_col=target_col)
self.x, self.y = ft.transform(train_df)
self.val_x, self.val_y = ft.transform(val_df)
def teardown_method(self, method):
pass
def test_fit_predict_evaluate(self):
self.model.fit_eval((self.x, self.y), [(self.val_x, self.val_y)])
# test predict
result = self.model.predict(self.val_x)
# test evaluate
evaluate_result = self.model.evaluate(self.val_x, self.val_y)
def test_save_restore(self):
self.model.fit_eval((self.x, self.y), [(self.val_x, self.val_y)])
result_save = self.model.predict(self.val_x)
model_file = "tmp.pkl"
self.model.save(model_file)
assert os.path.isfile(model_file)
new_model = XGBoost()
new_model.restore(model_file)
assert new_model.model
result_restore = new_model.predict(self.val_x)
assert_array_almost_equal(result_save, result_restore, decimal=2), \
"Prediction values are not the same after restore: " \
"predict before is {}, and predict after is {}".format(result_save, result_restore)
os.remove(model_file)
def test_metric(self):
# metric not in XGB_METRIC_NAME but in Evaluator.metrics_func.keys()
self.model.fit_eval(data=(self.x, self.y),
validation_data=[(self.val_x, self.val_y)],
metric="mse")
# metric in XGB_METRIC_NAME
self.model.fit_eval(data=(self.x, self.y),
validation_data=[(self.val_x, self.val_y)],
metric="rmsle")
with pytest.raises(ValueError):
self.model.fit_eval(data=(self.x, self.y),
validation_data=[(self.val_x, self.val_y)],
metric="wrong_metric")
# metric func
def pyrmsle(y_true, y_pred):
y_pred[y_pred < -1] = -1 + 1e-6
elements = np.power(np.log1p(y_true) - np.log1p(y_pred), 2)
return float(np.sqrt(np.sum(elements) / len(y_true)))
result = self.model.fit_eval(data=(self.x, self.y),
validation_data=[(self.val_x, self.val_y)
],
metric_func=pyrmsle)
assert "pyrmsle" in result
def test_data_creator(self):
def get_x_y(size, config):
values = np.random.randn(size, 4)
df = pd.DataFrame(values, columns=["f1", "f2", "f3", "t"])
selected_features = config["features"]
x = df[selected_features].to_numpy()
y = df["t"].to_numpy()
return x, y
from functools import partial
train_data_creator = partial(get_x_y, 20)
val_data_creator = partial(get_x_y, 5)
config = {'n_estimators': 5, 'max_depth': 2, 'tree_method': 'hist'}
model_builder = XGBoostModelBuilder(model_type="regressor",
cpus_per_trial=1,
**config)
model = model_builder.build(config={"features": ["f1", "f2"]})
model.fit_eval(train_data_creator,
validation_data=val_data_creator,
metric="mae")
def create_model():
_model = XGBoost(model_type=model_type, config=config)
if "cpu" in resources_per_trial:
_model.set_params(n_jobs=resources_per_trial.get("cpu"))
return _model
def build_from_ckpt(self, checkpoint_filename):
from zoo.orca.automl.xgboost.XGBoost import XGBoost
model = XGBoost(model_type=self.model_type, config=self.model_config)
model.restore(checkpoint_filename)
return model
def build(self, config):
from zoo.orca.automl.xgboost.XGBoost import XGBoost
model = XGBoost(model_type=self.model_type, config=self.model_config)
model._build(**config)
return model