Bug fix (#241)

* Fix fn _has_one_unfilled_arg in build_pipeline to correctly check the default value of the parameters
Restricts the maximum version of lightgbm to 4.5

* lint fix

* Update version and changelog

CHANGELOG.md

@@ -1,5 +1,10 @@
 # Changelog
 
+## [4.0.1] - 2025-02-26
+- **Bug Fix**
+  - Fix fn _has_one_unfilled_arg in build_pipeline to correctly check the default value of the parameters
+  - Restricts the maximum version of lightgbm to 4.5
+
 ## [4.0.0] - 2024-08-12
 - **Enhancement**
   - Add support for python 3.11 and 3.10

requirements_lgbm.txt

@@ -1 +1 @@
-lightgbm>=4,<5
+lightgbm>=4,<4.6

src/fklearn/resources/VERSION

@@ -1 +1 @@
-4.0.0
+4.0.1

src/fklearn/training/pipeline.py

@@ -44,7 +44,10 @@ def build_pipeline(*learners: LearnerFnType, has_repeated_learners: bool = False
     """
 
     def _has_one_unfilled_arg(learner: LearnerFnType) -> None:
-        no_default_list = [p for p, a in signature(learner).parameters.items() if a.default == '__no__default__']
+        no_default_list = [
+            p for p, a in signature(learner).parameters.items()
+            if isinstance(a.default, str) and a.default == '__no__default__'
+        ]
         if len(no_default_list) > 1:
             raise ValueError("Learner {0} has more than one unfilled argument: {1}\n"
                              "Make sure all learners are curried properly and only require one argument,"