def test_handling_alias_metrics():
# type: () -> None
for alias in [
"lambdarank",
"rank_xendcg",
"xendcg",
"xe_ndcg",
"xe_ndcg_mart",
"xendcg_mart",
"ndcg",
]:
lgbm_params = {"metric": alias}
_handling_alias_metrics(lgbm_params)
assert lgbm_params["metric"] == "ndcg"
for alias in ["mean_average_precision", "map"]:
lgbm_params = {"metric": alias}
_handling_alias_metrics(lgbm_params)
assert lgbm_params["metric"] == "map"
lgbm_params = {}
_handling_alias_metrics(lgbm_params)
assert lgbm_params == {}
lgbm_params = {"metric": "auc"}
_handling_alias_metrics(lgbm_params)
assert lgbm_params["metric"] == "auc"
lgbm_params = {"metric": "rmse"}
_handling_alias_metrics(lgbm_params)
assert lgbm_params["metric"] == "rmse"
def test_handling_alias_metrics():
# type: () -> None
for alias in [
'lambdarank', 'rank_xendcg', 'xendcg', 'xe_ndcg', 'xe_ndcg_mart',
'xendcg_mart', 'ndcg'
]:
lgbm_params = {'metric': alias}
_handling_alias_metrics(lgbm_params)
assert lgbm_params['metric'] == 'ndcg'
for alias in ['mean_average_precision', 'map']:
lgbm_params = {'metric': alias}
_handling_alias_metrics(lgbm_params)
assert lgbm_params['metric'] == 'map'
lgbm_params = {}
_handling_alias_metrics(lgbm_params)
assert lgbm_params == {}
lgbm_params = {'metric': 'auc'}
_handling_alias_metrics(lgbm_params)
assert lgbm_params['metric'] == 'auc'
lgbm_params = {'metric': 'rmse'}
_handling_alias_metrics(lgbm_params)
assert lgbm_params['metric'] == 'rmse'
def __init__(
self,
params, # type: Dict[str, Any]
train_set, # type: lgb.Dataset
num_boost_round=1000, # type: int
valid_sets=None, # type: Optional[VALID_SET_TYPE]
valid_names=None, # type: Optional[Any]
fobj=None, # type: Optional[Callable[..., Any]]
feval=None, # type: Optional[Callable[..., Any]]
feature_name="auto", # type: str
categorical_feature="auto", # type: str
early_stopping_rounds=None, # type: Optional[int]
evals_result=None, # type: Optional[Dict[Any, Any]]
verbose_eval=True, # type: Optional[bool]
learning_rates=None, # type: Optional[List[float]]
keep_training_booster=False, # type: Optional[bool]
callbacks=None, # type: Optional[List[Callable[..., Any]]]
time_budget=None, # type: Optional[int]
sample_size=None, # type: Optional[int]
best_params=None, # type: Optional[Dict[str, Any]]
tuning_history=None, # type: Optional[List[Dict[str, Any]]]
verbosity=1, # type: Optional[int]
):
params = copy.deepcopy(params)
# Handling alias metrics.
_handling_alias_metrics(params)
args = [params, train_set]
kwargs = dict(
num_boost_round=num_boost_round,
valid_sets=valid_sets,
valid_names=valid_names,
fobj=fobj,
feval=feval,
feature_name=feature_name,
categorical_feature=categorical_feature,
early_stopping_rounds=early_stopping_rounds,
evals_result=evals_result,
verbose_eval=verbose_eval,
learning_rates=learning_rates,
keep_training_booster=keep_training_booster,
callbacks=callbacks,
time_budget=time_budget,
verbosity=verbosity,
sample_size=sample_size,
) # type: Dict[str, Any]
self._parse_args(*args, **kwargs)
self.best_booster = None
self.best_score = -np.inf if self.higher_is_better() else np.inf
self.best_params = {} if best_params is None else best_params
self.tuning_history = [] if tuning_history is None else tuning_history
# Set default parameters as best.
self.best_params.update(DEFAULT_LIGHTGBM_PARAMETERS)
if valid_sets is None:
raise ValueError("`valid_sets` is required.")
def __init__(self, lgbm_params=None, lgbm_kwargs=None):
# type: (Dict[str, Any], Dict[str,Any]) -> None
# Handling alias metrics.
if lgbm_params is not None:
_handling_alias_metrics(lgbm_params)
self.lgbm_params = lgbm_params or {}
self.lgbm_kwargs = lgbm_kwargs or {}
def __init__(
self,
params, # type: Dict[str, Any]
train_set, # type: lgb.Dataset
num_boost_round=1000, # type: int
valid_sets=None, # type: Optional[VALID_SET_TYPE]
valid_names=None, # type: Optional[Any]
fobj=None, # type: Optional[Callable[..., Any]]
feval=None, # type: Optional[Callable[..., Any]]
feature_name="auto", # type: str
categorical_feature="auto", # type: str
early_stopping_rounds=None, # type: Optional[int]
evals_result=None, # type: Optional[Dict[Any, Any]]
verbose_eval=True, # type: Optional[bool]
learning_rates=None, # type: Optional[List[float]]
keep_training_booster=False, # type: Optional[bool]
callbacks=None, # type: Optional[List[Callable[..., Any]]]
time_budget=None, # type: Optional[int]
sample_size=None, # type: Optional[int]
best_params=None, # type: Optional[Dict[str, Any]]
tuning_history=None, # type: Optional[List[Dict[str, Any]]]
study=None, # type: Optional[Study]
model_dir=None, # type: Optional[str]
verbosity=1, # type: Optional[int]
):
# type: (...) -> None
params = copy.deepcopy(params)
# Handling alias metrics.
_handling_alias_metrics(params)
args = [params, train_set]
kwargs = dict(
num_boost_round=num_boost_round,
valid_sets=valid_sets,
valid_names=valid_names,
fobj=fobj,
feval=feval,
feature_name=feature_name,
categorical_feature=categorical_feature,
early_stopping_rounds=early_stopping_rounds,
evals_result=evals_result,
verbose_eval=verbose_eval,
learning_rates=learning_rates,
keep_training_booster=keep_training_booster,
callbacks=callbacks,
time_budget=time_budget,
verbosity=verbosity,
sample_size=sample_size,
) # type: Dict[str, Any]
self._parse_args(*args, **kwargs)
self._best_booster_with_trial_number = None # type: Optional[Tuple[lgb.Booster, int]]
self._start_time = None # type: Optional[float]
self._model_dir = model_dir
if self._model_dir is not None and not os.path.exists(self._model_dir):
os.mkdir(self._model_dir)
if best_params is not None:
warnings.warn(
"The `best_params` argument is deprecated. "
"Please get the parameter values via `lightgbm.basic.Booster.params`.",
DeprecationWarning,
)
if tuning_history is not None:
warnings.warn(
"The `tuning_history` argument is deprecated. "
"Please use the `study` argument to access optimization history.",
DeprecationWarning,
)
self._best_params = {} if best_params is None else best_params
self.tuning_history = [] if tuning_history is None else tuning_history
# Set default parameters as best.
self._best_params.update(DEFAULT_LIGHTGBM_PARAMETERS)
if study is None:
self.study = optuna.create_study(
direction="maximize" if self.higher_is_better() else "minimize"
)
else:
self.study = study
if self.higher_is_better():
if self.study.direction != optuna.study.StudyDirection.MAXIMIZE:
metric_name = self.lgbm_params.get("metric", "binary_logloss")
raise ValueError(
"Study direction is inconsistent with the metric {}. "
"Please set 'maximize' as the direction.".format(metric_name)
)
else:
if self.study.direction != optuna.study.StudyDirection.MINIMIZE:
metric_name = self.lgbm_params.get("metric", "binary_logloss")
raise ValueError(
"Study direction is inconsistent with the metric {}. "
"Please set 'minimize' as the direction.".format(metric_name)
)
if valid_sets is None:
raise ValueError("`valid_sets` is required.")
def __init__(
self,
params: Dict[str, Any],
train_set: "lgb.Dataset",
num_boost_round: int = 1000,
fobj: Optional[Callable[..., Any]] = None,
feval: Optional[Callable[..., Any]] = None,
feature_name: str = "auto",
categorical_feature: str = "auto",
early_stopping_rounds: Optional[int] = None,
verbose_eval: Optional[Union[bool, int]] = True,
callbacks: Optional[List[Callable[..., Any]]] = None,
time_budget: Optional[int] = None,
sample_size: Optional[int] = None,
study: Optional[optuna.study.Study] = None,
optuna_callbacks: Optional[List[Callable[[Study, FrozenTrial], None]]] = None,
verbosity: Optional[int] = 1,
) -> None:
_imports.check()
params = copy.deepcopy(params)
# Handling alias metrics.
_handling_alias_metrics(params)
args = [params, train_set]
kwargs = dict(
num_boost_round=num_boost_round,
fobj=fobj,
feval=feval,
feature_name=feature_name,
categorical_feature=categorical_feature,
early_stopping_rounds=early_stopping_rounds,
verbose_eval=verbose_eval,
callbacks=callbacks,
time_budget=time_budget,
sample_size=sample_size,
verbosity=verbosity,
) # type: Dict[str, Any]
self._parse_args(*args, **kwargs)
self._start_time = None # type: Optional[float]
self._optuna_callbacks = optuna_callbacks
self._best_params = {}
# Set default parameters as best.
self._best_params.update(DEFAULT_LIGHTGBM_PARAMETERS)
if study is None:
self.study = optuna.create_study(
direction="maximize" if self.higher_is_better() else "minimize"
)
else:
self.study = study
if self.higher_is_better():
if self.study.direction != optuna.study.StudyDirection.MAXIMIZE:
metric_name = self.lgbm_params.get("metric", "binary_logloss")
raise ValueError(
"Study direction is inconsistent with the metric {}. "
"Please set 'maximize' as the direction.".format(metric_name)
)
else:
if self.study.direction != optuna.study.StudyDirection.MINIMIZE:
metric_name = self.lgbm_params.get("metric", "binary_logloss")
raise ValueError(
"Study direction is inconsistent with the metric {}. "
"Please set 'minimize' as the direction.".format(metric_name)
)
