autogluon.tabular 1.2.1b20250226__tar.gz → 1.2.1b20250228__tar.gz

{autogluon.tabular-1.2.1b20250226 → autogluon.tabular-1.2.1b20250228}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: autogluon.tabular
-Version: 1.2.1b20250226
+Version: 1.2.1b20250228
 Summary: Fast and Accurate ML in 3 Lines of Code
 Home-page: https://github.com/autogluon/autogluon
 Author: AutoGluon Community

{autogluon.tabular-1.2.1b20250226 → autogluon.tabular-1.2.1b20250228}/src/autogluon/tabular/learner/abstract_learner.py

@@ -39,19 +39,19 @@ class AbstractTabularLearner(AbstractLearner):
         self,
         path_context: str,
         label: str,
-        feature_generator: PipelineFeatureGenerator,
+        feature_generator: PipelineFeatureGenerator | None = None,
         ignored_columns: list = None,
-        label_count_threshold=10,
-        problem_type=None,
-        quantile_levels=None,
-        eval_metric=None,
-        positive_class=None,
-        cache_data=True,
-        is_trainer_present=False,
-        random_state=0,
-        sample_weight=None,
-        weight_evaluation=False,
-        groups=None,
+        label_count_threshold: int = 10,
+        problem_type: str | None = None,
+        quantile_levels: list[float] | None = None,
+        eval_metric: Scorer | None = None,
+        positive_class: str | None = None,
+        cache_data: bool = True,
+        is_trainer_present: bool = False,
+        random_state: int = 0,
+        sample_weight: str | None = None,
+        weight_evaluation: bool = False,
+        groups: str | None = None,
     ):
         super().__init__(path_context=path_context, random_state=random_state)
         self.label = label

{autogluon.tabular-1.2.1b20250226 → autogluon.tabular-1.2.1b20250228}/src/autogluon/tabular/learner/default_learner.py

@@ -43,16 +43,17 @@ class DefaultLearner(AbstractTabularLearner):
     def _fit(
         self,
         X: DataFrame,
-        X_val: DataFrame = None,
-        X_test: DataFrame = None,
-        X_unlabeled: DataFrame = None,
-        holdout_frac=0.1,
-        num_bag_folds=0,
-        num_bag_sets=1,
-        time_limit=None,
-        infer_limit=None,
-        infer_limit_batch_size=None,
-        verbosity=2,
+        X_val: DataFrame | None = None,
+        X_test: DataFrame | None = None,
+        X_unlabeled: DataFrame | None = None,
+        holdout_frac: float = 0.1,
+        num_bag_folds: int = 0,
+        num_bag_sets: int = 1,
+        time_limit: float | None = None,
+        infer_limit: float | None = None,
+        infer_limit_batch_size: int | None = None,
+        verbosity: int = 2,
+        raise_on_model_failure: bool = False,
         **trainer_fit_kwargs,
     ):
         """Arguments:
@@ -121,6 +122,7 @@ class DefaultLearner(AbstractTabularLearner):
             save_data=self.cache_data,
             random_state=self.random_state,
             verbosity=verbosity,
+            raise_on_model_failure=raise_on_model_failure,
         )
 
         self.trainer_path = trainer.path

{autogluon.tabular-1.2.1b20250226 → autogluon.tabular-1.2.1b20250228}/src/autogluon/tabular/predictor/predictor.py

@@ -990,6 +990,14 @@ class TabularPredictor:
                 to any amount of labeled data.
             verbosity : int
                 If specified, overrides the existing `predictor.verbosity` value.
+            raise_on_model_failure: bool, default = False
+                If True, will raise on any exception during model training.
+                    This is useful when using a debugger during development to identify the cause of model failures.
+                    This should only be used for debugging.
+                If False, will try to skip to the next model if an exception occurred during model training.
+                    This is the default logic and is a core principle of AutoGluon's design.
+
+                .. versionadded:: 1.3.0
             raise_on_no_models_fitted: bool, default = True
                 If True, will raise a RuntimeError if no models were successfully fit during `fit()`.
             calibrate: bool or str, default = 'auto'
@@ -1109,6 +1117,7 @@ class TabularPredictor:
         delay_bag_sets: bool = kwargs["delay_bag_sets"]
         test_data = kwargs["test_data"]
         learning_curves = kwargs["learning_curves"]
+        raise_on_model_failure = kwargs["raise_on_model_failure"]
 
         if ag_args is None:
             ag_args = {}
@@ -1256,6 +1265,7 @@ class TabularPredictor:
             verbosity=verbosity,
             use_bag_holdout=use_bag_holdout,
             callbacks=callbacks,
+            raise_on_model_failure=raise_on_model_failure,
         )
         ag_post_fit_kwargs = dict(
             keep_only_best=kwargs["keep_only_best"],
@@ -5036,6 +5046,7 @@ class TabularPredictor:
             # learning curves and test data (for logging purposes only)
             learning_curves=False,
             test_data=None,
+            raise_on_model_failure=False,
         )
         kwargs, ds_valid_keys = self._sanitize_dynamic_stacking_kwargs(kwargs)
         kwargs = self._validate_fit_extra_kwargs(kwargs, extra_valid_keys=list(fit_kwargs_default.keys()) + ds_valid_keys)

{autogluon.tabular-1.2.1b20250226 → autogluon.tabular-1.2.1b20250228}/src/autogluon/tabular/trainer/abstract_trainer.py

@@ -128,6 +128,12 @@ class AbstractTabularTrainer(AbstractTrainer[AbstractModel]):
         Higher levels correspond to more detailed print statements (you can set verbosity = 0 to suppress warnings).
         If using logging, you can alternatively control amount of information printed via `logger.setLevel(L)`,
         where `L` ranges from 0 to 50 (Note: higher values of `L` correspond to fewer print statements, opposite of verbosity levels).
+    raise_on_model_failure : bool, default = False
+        If True, Trainer will raise on any exception during model training.
+            This is ideal when using a debugger during development.
+        If False, Trainer will try to skip to the next model if an exception occurred during model training.
+
+        .. versionadded:: 1.3.0
     """
 
     distill_stackname = "distill"  # name of stack-level for distilled student models
@@ -149,6 +155,7 @@ class AbstractTabularTrainer(AbstractTrainer[AbstractModel]):
         save_data: bool = False,
         random_state: int = 0,
         verbosity: int = 2,
+        raise_on_model_failure: bool = False,
     ):
         super().__init__(
             path=path,
@@ -163,6 +170,7 @@ class AbstractTabularTrainer(AbstractTrainer[AbstractModel]):
         #: Integer value added to the stack level to get the random_state for kfold splits or the train/val split if bagging is disabled
         self.random_state = random_state
         self.verbosity = verbosity
+        self.raise_on_model_failure = raise_on_model_failure
 
         # TODO: consider redesign where Trainer doesn't need sample_weight column name and weights are separate from X
         self.sample_weight = sample_weight
@@ -2188,6 +2196,10 @@ class AbstractTabularTrainer(AbstractTrainer[AbstractModel]):
             # TODO: Add recursive=True to avoid repeatedly loading models each time this is called for bagged ensembles (especially during repeated bagging)
             self.save_model(model=model)
         except Exception as exc:
+            if self.raise_on_model_failure:
+                # immediately raise instead of skipping to next model, useful for debugging during development
+                logger.warning("Model failure occurred... Raising exception instead of continuing to next model. (raise_on_model_failure=True)")
+                raise exc
             exception = exc  # required to reference exc outside of `except` statement
             del_model = True
             if isinstance(exception, TimeLimitExceeded):

{autogluon.tabular-1.2.1b20250226 → autogluon.tabular-1.2.1b20250228}/src/autogluon/tabular/version.py

@@ -1,4 +1,4 @@
 """This is the autogluon version file."""
 
-__version__ = "1.2.1b20250226"
+__version__ = "1.2.1b20250228"
 __lite__ = False

{autogluon.tabular-1.2.1b20250226 → autogluon.tabular-1.2.1b20250228}/src/autogluon.tabular.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: autogluon.tabular
-Version: 1.2.1b20250226
+Version: 1.2.1b20250228
 Summary: Fast and Accurate ML in 3 Lines of Code
 Home-page: https://github.com/autogluon/autogluon
 Author: AutoGluon Community

{autogluon.tabular-1.2.1b20250226 → autogluon.tabular-1.2.1b20250228}/src/autogluon.tabular.egg-info/requires.txt

@@ -3,20 +3,20 @@ scipy<1.16,>=1.5.4
 pandas<2.3.0,>=2.0.0
 scikit-learn<1.5.3,>=1.4.0
 networkx<4,>=3.0
-autogluon.core==1.2.1b20250226
-autogluon.features==1.2.1b20250226
+autogluon.core==1.2.1b20250228
+autogluon.features==1.2.1b20250228
 
 [all]
-numpy<2.0.0,>=1.25
-lightgbm<4.6,>=4.0
-autogluon.core[all]==1.2.1b20250226
+huggingface_hub[torch]
+torch<2.6,>=2.2
 fastai<2.8,>=2.3.1
+einops<0.9,>=0.7
+autogluon.core[all]==1.2.1b20250228
+numpy<2.0.0,>=1.25
 xgboost<2.2,>=1.6
-spacy<3.8
-torch<2.6,>=2.2
 catboost<1.3,>=1.2
-einops<0.9,>=0.7
-huggingface_hub[torch]
+spacy<3.8
+lightgbm<4.6,>=4.0
 
 [catboost]
 numpy<2.0.0,>=1.25
@@ -34,7 +34,7 @@ imodels<1.4.0,>=1.3.10
 lightgbm<4.6,>=4.0
 
 [ray]
-autogluon.core[all]==1.2.1b20250226
+autogluon.core[all]==1.2.1b20250228
 
 [skex]
 scikit-learn-intelex<2025.1,>=2024.0