fix tests and epsilon mismatch

mypy.ini

@@ -10,7 +10,7 @@ warn_return_any = True
 warn_unreachable = True
 warn_unused_configs = True
 warn_unused_ignores = True
-plugins = pydantic.mypy, numpy.typing.mypy_plugin
+plugins = pydantic.mypy
 
 [mypy-anytree]
 ignore_missing_imports = True

ocean/mip/_builders/model.py

@@ -140,7 +140,7 @@ def _cset(
         #   :: mu[j] <= 1 - flow[node.left],
         #   :: mu[j] >= epsilon * flow[node.right].
 
-        epsilon = self._find_best_epsilon(model, var)
+        epsilon = self._find_best_epsilon(model, var, self._epsilon)
         threshold = node.threshold
         j = int(np.searchsorted(var.levels, threshold))
 
@@ -227,9 +227,14 @@ def _eset(
         model.addConstr(x >= tree[node.right.node_id])
 
     @staticmethod
-    def _find_best_epsilon(model: BaseModel, var: FeatureVar) -> float:
+    def _find_best_epsilon(
+        model: BaseModel,
+        var: FeatureVar,
+        epsilon: float,
+    ) -> float:
         # Find the best epsilon value for the given feature variable.
-        # This
+        # This is done by finding the minimum difference between
+        # the split levels and the tolerance of the solver.
         tol: float = model.getParamInfo("FeasibilityTol")[2]
         min_tol: float = 1e-9
         delta: float = min(*np.diff(var.levels))
@@ -242,6 +247,8 @@ def _find_best_epsilon(model: BaseModel, var: FeatureVar) -> float:
             msg += " Consider not scaling the data or using bigger intervals."
             warnings.warn(msg, category=UserWarning, stacklevel=2)
             return eps
+        if delta * epsilon > tol:
+            return epsilon
         while 2 * tol / delta >= 1.0:
             tol /= 2
         feas_tol = model.getParamInfo("FeasibilityTol")[2]

tests/mip/model/test_objective.py

@@ -130,7 +130,6 @@ def test_set_majority_class(
             assert model.Status == gp.GRB.OPTIMAL
 
             explanation = model.explanation
-
             validate_solution(explanation)
             validate_paths(*model.trees, explanation=explanation)
             validate_sklearn_paths(clf, explanation, model.trees)