Define leaves_cutoff_threshold as constant (#148)

m2cgen/assemblers/boosting.py

@@ -5,12 +5,15 @@
 from m2cgen.assemblers.base import ModelAssembler
 
 
+LEAVES_CUTOFF_THRESHOLD = 3000
+
+
 class BaseBoostingAssembler(ModelAssembler):
 
     classifier_name = None
 
     def __init__(self, model, trees, base_score=0, tree_limit=None,
-                 leaves_cutoff_threshold=3000):
+                 leaves_cutoff_threshold=LEAVES_CUTOFF_THRESHOLD):
         super().__init__(model)
         self.all_trees = trees
         self._base_score = base_score
@@ -130,7 +133,8 @@ class XGBoostModelAssembler(BaseBoostingAssembler):
 
     classifier_name = "XGBClassifier"
 
-    def __init__(self, model, leaves_cutoff_threshold=3000):
+    def __init__(self, model,
+                 leaves_cutoff_threshold=LEAVES_CUTOFF_THRESHOLD):
         feature_names = model.get_booster().feature_names
         self._feature_name_to_idx = {
             name: idx for idx, name in enumerate(feature_names or [])
@@ -198,7 +202,8 @@ class LightGBMModelAssembler(BaseBoostingAssembler):
 
     classifier_name = "LGBMClassifier"
 
-    def __init__(self, model, leaves_cutoff_threshold=3000):
+    def __init__(self, model,
+                 leaves_cutoff_threshold=LEAVES_CUTOFF_THRESHOLD):
         model_dump = model.booster_.dump_model()
         trees = [m["tree_structure"] for m in model_dump["tree_info"]]
         self.n_iter = len(trees) // model_dump["num_tree_per_iteration"]