v1.2.1: fix skorch problem, fix failing test for Mac OS

Author: dholzmueller

README.md

@@ -168,6 +168,8 @@ and https://docs.ray.io/en/latest/cluster/vms/user-guides/community/slurm.html
 
 ## Releases (see git tags)
 
+- v1.2.1:
+    - avoid error for older skorch versions
 - v1.2.0:
     - Included post-hoc calibration and more metrics through 
       [probmetrics](https://github.com/dholzmueller/probmetrics).

pytabkit/__about__.py

@@ -2,4 +2,4 @@
 #
 # SPDX-License-Identifier: Apache-2.0
 
-__version__ = "1.2.0"
+__version__ = "1.2.1"

pytabkit/models/nn_models/rtdl_resnet.py

@@ -974,20 +974,27 @@ def create_regressor_skorch(
         callbacks=callbacks,
         **kwargs)
 
-    try:
-        # try the torch_load_kwargs but it's only available in newer versions of skorch
-        model = nn_class(
-            model_class,
-            # Shuffle training data on each epoch
-            **new_kwargs,
-            torch_load_kwargs={'weights_only': False}, # quick-fix for pickling errors in torch>=2.6
-        )
-    except ValueError:
-        model = nn_class(
-            model_class,
-            # Shuffle training data on each epoch
-            **new_kwargs,
-        )
+    # cannot do the try/catch here because params are validated in fit()
+    # try:
+    #     # try the torch_load_kwargs but it's only available in newer versions of skorch
+    #     model = nn_class(
+    #         model_class,
+    #         # Shuffle training data on each epoch
+    #         **new_kwargs,
+    #         torch_load_kwargs={'weights_only': False}, # quick-fix for pickling errors in torch>=2.6
+    #     )
+    # except ValueError:
+    #     model = nn_class(
+    #         model_class,
+    #         # Shuffle training data on each epoch
+    #         **new_kwargs,
+    #     )
+
+    model = nn_class(
+        model_class,
+        # Shuffle training data on each epoch
+        **new_kwargs,
+    )
 
     return model
 

tests/test_ensemble.py

@@ -6,12 +6,13 @@
 from pytabkit.models.sklearn.sklearn_interfaces import Ensemble_HPO_Classifier, Ensemble_HPO_Regressor
 
 
-@pytest.mark.parametrize('model', [Ensemble_TD_Classifier(calibration_method='ts-mix', val_metric_name='ref-ll-ts'),
-                                    Ensemble_TD_Regressor(),
-                                    Ensemble_HPO_Classifier(calibration_method='ts-mix',
-                                                            val_metric_name='ref-ll-ts', n_hpo_steps=1),
-                                    Ensemble_HPO_Regressor(n_hpo_steps=1),
-                                    ])
+@pytest.mark.parametrize('model', [
+    Ensemble_TD_Classifier(calibration_method='ts-mix', val_metric_name='ref-ll-ts', device='cpu'),
+    Ensemble_TD_Regressor(device='cpu'),
+    Ensemble_HPO_Classifier(calibration_method='ts-mix',
+                            val_metric_name='ref-ll-ts', n_hpo_steps=1, device='cpu'),
+    Ensemble_HPO_Regressor(n_hpo_steps=1, device='cpu'),
+    ])
 def test_ensemble(model):
     np.random.seed(0)
     X = np.random.randn(100, 2)
@@ -20,4 +21,3 @@ def test_ensemble(model):
         y = y > 0.0
     model.fit(X, y)
     model.predict(X)
-

tests/test_sklearn_interfaces.py

@@ -17,6 +17,7 @@
     LGBM_TD_Classifier(min_data_in_leaf=2, subsample=1.0), LGBM_TD_Regressor(subsample=1.0),
     XGB_TD_Classifier(), XGB_TD_Regressor(),
     CatBoost_TD_Classifier(), CatBoost_TD_Regressor(),
+    # use CPU to avoid Mac OS errors with MPS backend
     RealMLP_TD_Classifier(n_epochs=8, device='cpu'), RealMLP_TD_Regressor(n_epochs=64, device='cpu'),
     TabM_D_Classifier(device='cpu', tabm_k=2, num_emb_type='pwl', arch_type='tabm-mini', num_emb_n_bins=2),
     TabM_D_Regressor(device='cpu', tabm_k=2, num_emb_type='pwl', arch_type='tabm-mini', num_emb_n_bins=2),