autogluon.tabular 1.3.2b20250615__tar.gz → 1.3.2b20250616__tar.gz

{autogluon.tabular-1.3.2b20250615 → autogluon.tabular-1.3.2b20250616}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: autogluon.tabular
-Version: 1.3.2b20250615
+Version: 1.3.2b20250616
 Summary: Fast and Accurate ML in 3 Lines of Code
 Home-page: https://github.com/autogluon/autogluon
 Author: AutoGluon Community

{autogluon.tabular-1.3.2b20250615 → autogluon.tabular-1.3.2b20250616}/src/autogluon/tabular/models/_utils/rapids_utils.py

@@ -10,7 +10,7 @@ class RapidsModelMixin:
     @classmethod
     def _get_default_ag_args_ensemble(cls, **kwargs) -> dict:
         default_ag_args_ensemble = super()._get_default_ag_args_ensemble(**kwargs)
-        extra_ag_args_ensemble = {"use_child_oof": False}
+        extra_ag_args_ensemble = {"use_child_oof": False, "fold_fitting_strategy": "sequential_local"}
         default_ag_args_ensemble.update(extra_ag_args_ensemble)
         return default_ag_args_ensemble
 

autogluon.tabular-1.3.2b20250616/src/autogluon/tabular/models/lr/lr_rapids_model.py

@@ -0,0 +1,100 @@
+import logging
+
+from autogluon.common.utils.try_import import try_import_rapids_cuml
+from autogluon.core.constants import REGRESSION
+
+from .._utils.rapids_utils import RapidsModelMixin
+from .hyperparameters.parameters import get_param_baseline
+from .lr_model import LinearModel
+
+logger = logging.getLogger(__name__)
+
+
+# FIXME: If rapids is installed, normal CPU LinearModel crashes.
+class LinearRapidsModel(RapidsModelMixin, LinearModel):
+    """
+    RAPIDS Linear model : https://rapids.ai/start.html
+
+    NOTE: This code is experimental, it is recommend to not use this unless you are a developer.
+    This was tested on rapids-21.06 via:
+
+    conda create -n rapids-21.06 -c rapidsai -c nvidia -c conda-forge rapids=21.06 python=3.8 cudatoolkit=11.2
+    conda activate rapids-21.06
+    pip install --pre autogluon.tabular[all]
+    """
+
+    def _get_model_type(self):
+        penalty = self.params.get("penalty", "L2")
+        try_import_rapids_cuml()
+        from cuml.linear_model import Lasso, LogisticRegression, Ridge
+
+        if self.problem_type == REGRESSION:
+            if penalty == "L2":
+                model_type = Ridge
+            elif penalty == "L1":
+                model_type = Lasso
+            else:
+                raise AssertionError(f'Unknown value for penalty "{penalty}" - supported types are ["L1", "L2"]')
+        else:
+            model_type = LogisticRegression
+        return model_type
+
+    def _set_default_params(self):
+        default_params = {"fit_intercept": True, "max_iter": 10000}
+        if self.problem_type != REGRESSION:
+            default_params.update({"solver": "qn"})
+        default_params.update(get_param_baseline())
+        for param, val in default_params.items():
+            self._set_default_param_value(param, val)
+
+    def _preprocess(self, X, **kwargs):
+        X = super()._preprocess(X=X, **kwargs)
+        if hasattr(X, 'toarray'):  # Check if it's a sparse matrix
+            X = X.toarray()
+        return X
+
+    def _fit(self, X, y, **kwargs):
+        """
+        Custom fit method for RAPIDS cuML models that handles parameter compatibility
+        and bypasses sklearn-specific incremental training approach.
+        """
+        # Preprocess data
+        X = self.preprocess(X, is_train=True)
+        if self.problem_type == 'binary':
+            y = y.astype(int).values
+
+        # Create cuML model with filtered parameters
+        model_cls = self._get_model_type()
+
+        # Comprehensive parameter filtering for cuML compatibility
+        cuml_incompatible_params = {
+            # AutoGluon-specific preprocessing parameters
+            'vectorizer_dict_size', 'proc.ngram_range', 'proc.skew_threshold',
+            'proc.impute_strategy', 'handle_text',
+            # sklearn-specific parameters not supported by cuML
+            'n_jobs', 'warm_start', 'multi_class', 'dual', 'intercept_scaling',
+            'class_weight', 'random_state', 'verbose',
+            # Parameters that need conversion or special handling
+            'penalty', 'C'
+        }
+
+        # Filter out incompatible parameters
+        filtered_params = {k: v for k, v in self.params.items()
+                          if k not in cuml_incompatible_params}
+
+        # Handle parameter conversions for cuML
+        if self.problem_type == REGRESSION:
+            # Convert sklearn's C parameter to cuML's alpha
+            if 'C' in self.params:
+                filtered_params['alpha'] = 1.0 / self.params['C']
+        else:
+            # For classification, keep C parameter
+            if 'C' in self.params:
+                filtered_params['C'] = self.params['C']
+
+        # Create and fit cuML model - let cuML handle its own error messages
+        self.model = model_cls(**filtered_params)
+        self.model.fit(X, y)
+
+        # Add missing sklearn-compatible attributes for AutoGluon compatibility
+        self.model.n_iter_ = None  # cuML doesn't track iterations like sklearn

{autogluon.tabular-1.3.2b20250615 → autogluon.tabular-1.3.2b20250616}/src/autogluon/tabular/version.py

@@ -1,4 +1,4 @@
 """This is the autogluon version file."""
 
-__version__ = "1.3.2b20250615"
+__version__ = "1.3.2b20250616"
 __lite__ = False

{autogluon.tabular-1.3.2b20250615 → autogluon.tabular-1.3.2b20250616}/src/autogluon.tabular.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: autogluon.tabular
-Version: 1.3.2b20250615
+Version: 1.3.2b20250616
 Summary: Fast and Accurate ML in 3 Lines of Code
 Home-page: https://github.com/autogluon/autogluon
 Author: AutoGluon Community

{autogluon.tabular-1.3.2b20250615 → autogluon.tabular-1.3.2b20250616}/src/autogluon.tabular.egg-info/requires.txt

@@ -3,19 +3,19 @@ scipy<1.16,>=1.5.4
 pandas<2.3.0,>=2.0.0
 scikit-learn<1.7.0,>=1.4.0
 networkx<4,>=3.0
-autogluon.core==1.3.2b20250615
-autogluon.features==1.3.2b20250615
+autogluon.core==1.3.2b20250616
+autogluon.features==1.3.2b20250616
 
 [all]
-xgboost<3.1,>=2.0
-huggingface_hub[torch]
-autogluon.core[all]==1.3.2b20250615
+fastai<2.9,>=2.3.1
 numpy<2.3.0,>=1.25
-catboost<1.3,>=1.2
 lightgbm<4.7,>=4.0
-fastai<2.9,>=2.3.1
 torch<2.7,>=2.2
+catboost<1.3,>=1.2
 einops<0.9,>=0.7
+autogluon.core[all]==1.3.2b20250616
+xgboost<3.1,>=2.0
+huggingface_hub[torch]
 spacy<3.9
 
 [catboost]
@@ -34,7 +34,7 @@ imodels<2.1.0,>=1.3.10
 lightgbm<4.7,>=4.0
 
 [ray]
-autogluon.core[all]==1.3.2b20250615
+autogluon.core[all]==1.3.2b20250616
 
 [skex]
 scikit-learn-intelex<2025.5,>=2024.0

autogluon.tabular-1.3.2b20250615/src/autogluon/tabular/models/lr/lr_rapids_model.py

@@ -1,60 +0,0 @@
-import logging
-
-import numpy as np
-
-from autogluon.common.utils.try_import import try_import_rapids_cuml
-from autogluon.core.constants import REGRESSION
-
-from .._utils.rapids_utils import RapidsModelMixin
-from .hyperparameters.parameters import get_param_baseline
-from .lr_model import LinearModel
-
-logger = logging.getLogger(__name__)
-
-
-# FIXME: If rapids is installed, normal CPU LinearModel crashes.
-class LinearRapidsModel(RapidsModelMixin, LinearModel):
-    """
-    RAPIDS Linear model : https://rapids.ai/start.html
-
-    NOTE: This code is experimental, it is recommend to not use this unless you are a developer.
-    This was tested on rapids-21.06 via:
-
-    conda create -n rapids-21.06 -c rapidsai -c nvidia -c conda-forge rapids=21.06 python=3.8 cudatoolkit=11.2
-    conda activate rapids-21.06
-    pip install --pre autogluon.tabular[all]
-    """
-
-    def _get_model_type(self):
-        penalty = self.params.get("penalty", "L2")
-        try_import_rapids_cuml()
-        from cuml.linear_model import Lasso, LogisticRegression, Ridge
-
-        if self.problem_type == REGRESSION:
-            if penalty == "L2":
-                model_type = Ridge
-            elif penalty == "L1":
-                model_type = Lasso
-            else:
-                raise AssertionError(f'Unknown value for penalty "{penalty}" - supported types are ["L1", "L2"]')
-        else:
-            model_type = LogisticRegression
-        return model_type
-
-    def _set_default_params(self):
-        default_params = {"fit_intercept": True, "max_iter": 10000}
-        if self.problem_type != REGRESSION:
-            default_params.update({"solver": "qn"})
-        default_params.update(get_param_baseline())
-        for param, val in default_params.items():
-            self._set_default_param_value(param, val)
-
-    def _preprocess(self, X, **kwargs):
-        X = super()._preprocess(X=X, **kwargs)
-        if not isinstance(X, np.ndarray):
-            X = X.toarray()
-        return X
-
-    def _fit(self, X, y, **kwargs):
-        kwargs.pop("sample_weight", None)  # sample_weight is not supported
-        super()._fit(X=X, y=y, **kwargs)