From 4e19e88ad6ea58089a296f5a33fd71b1ed8145f2 Mon Sep 17 00:00:00 2001
From: RektPunk <110188257+RektPunk@users.noreply.github.com>
Date: Tue, 17 Sep 2024 18:29:33 +0900
Subject: [PATCH] [Feature] remove feval (#5)

---
 imlightgbm/engine.py    |  6 ++---
 imlightgbm/objective.py | 57 ++++++++++++++++++++++-------------------
 2 files changed, 33 insertions(+), 30 deletions(-)

diff --git a/imlightgbm/engine.py b/imlightgbm/engine.py
index df5a3b7..8b2c147 100644
--- a/imlightgbm/engine.py
+++ b/imlightgbm/engine.py
@@ -22,13 +22,12 @@ def train(
     keep_training_booster: bool = False,
     callbacks: list[Callable] | None = None,
 ) -> lgb.Booster:
-    _params, feval = set_params(params=params, train_set=train_set)
+    _params = set_params(params=params, train_set=train_set)
     return lgb.train(
         params=_params,
         train_set=train_set,
         valid_sets=valid_sets,
         valid_names=valid_names,
-        feval=feval,
         num_boost_round=num_boost_round,
         init_model=init_model,
         feature_name=feature_name,
@@ -61,7 +60,7 @@ def cv(
     eval_train_metric: bool = False,
     return_cvbooster: bool = False,
 ) -> dict[str, list[float] | lgb.CVBooster]:
-    _params, feval = set_params(params=params, train_set=train_set)
+    _params = set_params(params=params, train_set=train_set)
     return lgb.cv(
         params=_params,
         train_set=train_set,
@@ -71,7 +70,6 @@ def cv(
         stratified=stratified,
         shuffle=shuffle,
         metrics=metrics,
-        feval=feval,
         init_model=init_model,
         feature_name=feature_name,
         categorical_feature=categorical_feature,
diff --git a/imlightgbm/objective.py b/imlightgbm/objective.py
index 230c106..e13c9bc 100644
--- a/imlightgbm/objective.py
+++ b/imlightgbm/objective.py
@@ -13,38 +13,30 @@
 ALPHA_DEFAULT: float = 0.25
 GAMMA_DEFAULT: float = 2.0
 OBJECTIVE_STR: str = "objective"
-IS_HIGHER_BETTER = False
+METRIC_STR: str = "metric"
+IS_HIGHER_BETTER: bool = False
 
 
 def _power(num_base: np.ndarray, num_pow: float):
+    """Safe power."""
     return np.sign(num_base) * (np.abs(num_base)) ** (num_pow)
 
 
 def _log(array: np.ndarray, is_prob: bool = False) -> np.ndarray:
+    """Safe log."""
     _upper = 1 if is_prob else None
     return np.log(np.clip(array, 1e-6, _upper))
 
 
 def _sigmoid(x: np.ndarray) -> np.ndarray:
-    """Convert raw predictions to probabilities in binary task"""
+    """Convert raw predictions to probabilities in binary task."""
     return 1 / (1 + np.exp(-x))
 
 
-def binary_focal_eval(
-    pred: np.ndarray, train_data: Dataset, alpha: float, gamma: float
-) -> tuple[str, float, bool]:
-    label = train_data.get_label()
-    pred_prob = _sigmoid(pred)
-    p_t = np.where(label == 1, pred_prob, 1 - pred_prob)
-    loss = -alpha * ((1 - p_t) ** gamma) * _log(p_t, True)
-
-    focal_loss = np.mean(loss)
-    return "binary_focal", focal_loss, IS_HIGHER_BETTER
-
-
 def binary_focal_objective(
     pred: np.ndarray, train_data: Dataset, gamma: float
 ) -> tuple[np.ndarray, np.ndarray]:
+    """Return binary focal objective."""
     label = train_data.get_label()
     pred_prob = _sigmoid(pred)
 
@@ -65,11 +57,17 @@ def binary_focal_objective(
     return grad, hess
 
 
-def multiclass_focal_eval(
+def binary_focal_eval(
     pred: np.ndarray, train_data: Dataset, alpha: float, gamma: float
 ) -> tuple[str, float, bool]:
-    # TODO
-    return
+    """Return binary focal eval."""
+    label = train_data.get_label()
+    pred_prob = _sigmoid(pred)
+    p_t = np.where(label == 1, pred_prob, 1 - pred_prob)
+    loss = -alpha * ((1 - p_t) ** gamma) * _log(p_t, True)
+
+    focal_loss = np.mean(loss)
+    return "focal", focal_loss, IS_HIGHER_BETTER
 
 
 def multiclass_focal_objective(
@@ -79,9 +77,17 @@ def multiclass_focal_objective(
     return
 
 
-def set_fobj_feval(
+def multiclass_focal_eval(
+    pred: np.ndarray, train_data: Dataset, alpha: float, gamma: float
+) -> tuple[str, float, bool]:
+    # TODO
+    return
+
+
+def _set_fobj_feval(
     train_set: Dataset, alpha: float, gamma: float
 ) -> tuple[ObjLike, EvalLike]:
+    """Return obj and eval with respect to task type."""
     inferred_task = type_of_target(train_set.get_label())
     if inferred_task not in {"binary", "multiclass"}:
         raise ValueError(
@@ -92,8 +98,8 @@ def set_fobj_feval(
         "multiclass": partial(multiclass_focal_objective, alpha=alpha, gamma=gamma),
     }
     eval_mapper: dict[str, EvalLike] = {
-        "binary": partial(binary_focal_eval, alpha=alpha, gamma=gamma),
-        "multiclass": partial(multiclass_focal_eval, alpha=alpha, gamma=gamma),
+        "binary": "binary_logloss",
+        "multiclass": "multi_logloss",
     }
     fobj = objective_mapper[inferred_task]
     feval = eval_mapper[inferred_task]
@@ -101,9 +107,8 @@ def set_fobj_feval(
     return fobj, feval
 
 
-def set_params(
-    params: dict[str, Any], train_set: Dataset
-) -> tuple[dict[str, Any], EvalLike]:
+def set_params(params: dict[str, Any], train_set: Dataset) -> dict[str, Any]:
+    """Set params and eval finction, objective in params."""
     _params = deepcopy(params)
     if OBJECTIVE_STR in _params:
         logger.warning(f"'{OBJECTIVE_STR}' exists in params will not used.")
@@ -112,6 +117,6 @@ def set_params(
     _alpha = _params.pop("alpha", ALPHA_DEFAULT)
     _gamma = _params.pop("gamma", GAMMA_DEFAULT)
 
-    fobj, feval = set_fobj_feval(train_set=train_set, alpha=_alpha, gamma=_gamma)
-    _params.update({OBJECTIVE_STR: fobj})
-    return _params, feval
+    fobj, feval = _set_fobj_feval(train_set=train_set, alpha=_alpha, gamma=_gamma)
+    _params.update({OBJECTIVE_STR: fobj, METRIC_STR: feval})
+    return _params