autogluon.tabular 1.2.1b20250304__tar.gz → 1.2.1b20250306__tar.gz

{autogluon.tabular-1.2.1b20250304 → autogluon.tabular-1.2.1b20250306}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: autogluon.tabular
-Version: 1.2.1b20250304
+Version: 1.2.1b20250306
 Summary: Fast and Accurate ML in 3 Lines of Code
 Home-page: https://github.com/autogluon/autogluon
 Author: AutoGluon Community

{autogluon.tabular-1.2.1b20250304 → autogluon.tabular-1.2.1b20250306}/setup.py

@@ -40,11 +40,8 @@ extras_require = {
         "numpy>=1.25,<2.0.0",  # TODO support numpy>=2.0.0 once issue resolved https://github.com/catboost/catboost/issues/2671
         "catboost>=1.2,<1.3",
     ],
-    # FIXME: Debug why xgboost 1.6 has 4x+ slower inference on multiclass datasets compared to 1.4
-    #  It is possibly only present on MacOS, haven't tested linux.
-    # XGBoost made API breaking changes in 1.6 with custom metric and callback support, so we don't support older versions.
     "xgboost": [
-        "xgboost>=1.6,<2.2",  # <{N+1} upper cap, where N is the latest released minor version
+        "xgboost>=2.0,<2.2",  # <{N+1} upper cap, where N is the latest released minor version
     ],
     "fastai": [
         "spacy<3.8",  # cap for issue https://github.com/explosion/spaCy/issues/13653

{autogluon.tabular-1.2.1b20250304 → autogluon.tabular-1.2.1b20250306}/src/autogluon/tabular/models/automm/automm_model.py

@@ -1,4 +1,5 @@
 """Wrapper of the MultiModalPredictor."""
+from __future__ import annotations
 
 import logging
 import os
@@ -83,11 +84,14 @@ class MultiModalPredictorModel(AbstractModel):
         default_ag_args = super()._get_default_ag_args()
         extra_ag_args = {
             "valid_stacker": False,
-            "problem_types": [BINARY, MULTICLASS, REGRESSION],
         }
         default_ag_args.update(extra_ag_args)
         return default_ag_args
 
+    @classmethod
+    def supported_problem_types(cls) -> list[str] | None:
+        return ["binary", "multiclass", "regression"]
+
     # FIXME: Enable parallel bagging once AutoMM supports being run within Ray without hanging
     @classmethod
     def _get_default_ag_args_ensemble(cls, **kwargs) -> dict:

{autogluon.tabular-1.2.1b20250304 → autogluon.tabular-1.2.1b20250306}/src/autogluon/tabular/models/automm/ft_transformer.py

@@ -1,4 +1,5 @@
 """Wrapper of the MultiModalPredictor."""
+from __future__ import annotations
 
 import logging
 from typing import Dict

{autogluon.tabular-1.2.1b20250304 → autogluon.tabular-1.2.1b20250306}/src/autogluon/tabular/models/catboost/catboost_model.py

@@ -1,3 +1,5 @@
+from __future__ import annotations
+
 import logging
 import math
 import os

{autogluon.tabular-1.2.1b20250304 → autogluon.tabular-1.2.1b20250306}/src/autogluon/tabular/models/fasttext/fasttext_model.py

@@ -1,3 +1,5 @@
+from __future__ import annotations
+
 __all__ = ["FastTextModel"]
 
 import contextlib
@@ -49,10 +51,14 @@ class FastTextModel(AbstractModel):
     @classmethod
     def _get_default_ag_args(cls) -> dict:
         default_ag_args = super()._get_default_ag_args()
-        extra_ag_args = {"valid_stacker": False, "problem_types": [BINARY, MULTICLASS]}
+        extra_ag_args = {"valid_stacker": False}
         default_ag_args.update(extra_ag_args)
         return default_ag_args
 
+    @classmethod
+    def supported_problem_types(cls) -> list[str] | None:
+        return ["binary", "multiclass"]
+
     def _fit(self, X, y, sample_weight=None, **kwargs):
         if self.problem_type not in (BINARY, MULTICLASS):
             raise ValueError("FastText model only supports binary or multiclass classification")

{autogluon.tabular-1.2.1b20250304 → autogluon.tabular-1.2.1b20250306}/src/autogluon/tabular/models/image_prediction/image_predictor.py

@@ -1,3 +1,5 @@
+from __future__ import annotations
+
 import logging
 
 import numpy as np
@@ -123,3 +125,7 @@ class ImagePredictorModel(MultiModalPredictorModel):
         else:
             raise NotImplementedError(f"Computing dummy pred_proba is not implemented for {self.problem_type}.")
         return pred_proba_mean
+
+    @classmethod
+    def supported_problem_types(cls) -> list[str] | None:
+        return ["binary", "multiclass", "regression"]

{autogluon.tabular-1.2.1b20250304 → autogluon.tabular-1.2.1b20250306}/src/autogluon/tabular/models/imodels/imodels_models.py

@@ -1,3 +1,5 @@
+from __future__ import annotations
+
 from abc import abstractmethod
 
 import numpy as np
@@ -63,6 +65,10 @@ class _IModelsModel(AbstractModel):
         for param, val in default_params.items():
             self._set_default_param_value(param, val)
 
+    @classmethod
+    def supported_problem_types(cls) -> list[str] | None:
+        return ["binary", "multiclass", "regression"]
+
     def _get_default_auxiliary_params(self) -> dict:
         default_auxiliary_params = super()._get_default_auxiliary_params()
         extra_auxiliary_params = dict(

{autogluon.tabular-1.2.1b20250304 → autogluon.tabular-1.2.1b20250306}/src/autogluon/tabular/models/knn/knn_model.py

@@ -1,3 +1,5 @@
+from __future__ import annotations
+
 import logging
 import math
 import time

{autogluon.tabular-1.2.1b20250304 → autogluon.tabular-1.2.1b20250306}/src/autogluon/tabular/models/lr/lr_model.py

@@ -1,3 +1,5 @@
+from __future__ import annotations
+
 import logging
 import re
 import time

{autogluon.tabular-1.2.1b20250304 → autogluon.tabular-1.2.1b20250306}/src/autogluon/tabular/models/rf/rf_model.py

@@ -399,8 +399,10 @@ class RFModel(AbstractModel):
             tags["valid_oof"] = True
         return tags
 
-    def _valid_compilers(self):
+    @classmethod
+    def _valid_compilers(cls):
         return [RFNativeCompiler, RFOnnxCompiler]
 
-    def _default_compiler(self):
+    @classmethod
+    def _default_compiler(cls):
         return RFNativeCompiler

{autogluon.tabular-1.2.1b20250304 → autogluon.tabular-1.2.1b20250306}/src/autogluon/tabular/models/tab_transformer/tab_transformer_model.py

@@ -1,4 +1,5 @@
 """TabTransformer model"""
+from __future__ import annotations
 
 import logging
 import os
@@ -486,6 +487,10 @@ class TabTransformerModel(AbstractNeuralNetworkModel):
     def _get_default_searchspace(self):
         return get_default_searchspace()
 
+    @classmethod
+    def supported_problem_types(cls) -> list[str] | None:
+        return ["binary", "multiclass", "regression"]
+
     def save(self, path: str = None, verbose=True) -> str:
         import torch
 

{autogluon.tabular-1.2.1b20250304 → autogluon.tabular-1.2.1b20250306}/src/autogluon/tabular/models/tabpfn/tabpfn_model.py

@@ -1,3 +1,5 @@
+from __future__ import annotations
+
 import numpy as np
 import pandas as pd
 
@@ -97,13 +99,8 @@ class TabPFNModel(AbstractModel):
             self._set_default_param_value(param, val)
 
     @classmethod
-    def _get_default_ag_args(cls) -> dict:
-        default_ag_args = super()._get_default_ag_args()
-        extra_ag_args = {
-            "problem_types": [BINARY, MULTICLASS],
-        }
-        default_ag_args.update(extra_ag_args)
-        return default_ag_args
+    def supported_problem_types(cls) -> list[str] | None:
+        return ["binary", "multiclass"]
 
     def _get_default_auxiliary_params(self) -> dict:
         """

{autogluon.tabular-1.2.1b20250304 → autogluon.tabular-1.2.1b20250306}/src/autogluon/tabular/models/tabpfnmix/tabpfnmix_model.py

@@ -297,13 +297,8 @@ class TabPFNMixModel(AbstractModel):
         return os.path.join(self.path, self.weights_file_name)
 
     @classmethod
-    def _get_default_ag_args(cls) -> dict:
-        default_ag_args = super()._get_default_ag_args()
-        extra_ag_args = {
-            "problem_types": [BINARY, MULTICLASS, REGRESSION],
-        }
-        default_ag_args.update(extra_ag_args)
-        return default_ag_args
+    def supported_problem_types(cls) -> list[str] | None:
+        return ["binary", "multiclass", "regression"]
 
     def _get_default_auxiliary_params(self) -> dict:
         default_auxiliary_params = super()._get_default_auxiliary_params()
@@ -314,15 +309,6 @@ class TabPFNMixModel(AbstractModel):
         )
         return default_auxiliary_params
 
-    @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 = {
-           #  "fold_fitting_strategy": "sequential_local",  # FIXME: Comment out after debugging for large speedup
-        }
-        default_ag_args_ensemble.update(extra_ag_args_ensemble)
-        return default_ag_args_ensemble
-
     def _get_maximum_resources(self) -> dict[str, int | float]:
         # torch model trains slower when utilizing virtual cores and this issue scale up when the number of cpu cores increases
         return {"num_cpus": ResourceManager.get_cpu_count_psutil(logical=False)}

{autogluon.tabular-1.2.1b20250304 → autogluon.tabular-1.2.1b20250306}/src/autogluon/tabular/models/tabular_nn/torch/tabular_nn_torch.py

@@ -897,10 +897,12 @@ class TabularNeuralNetTorchModel(AbstractNeuralNetworkModel):
             minimum_resources["num_gpus"] = 1
         return minimum_resources
 
-    def _valid_compilers(self):
+    @classmethod
+    def _valid_compilers(cls):
         return [TabularNeuralNetTorchNativeCompiler, TabularNeuralNetTorchOnnxCompiler]
 
-    def _default_compiler(self):
+    @classmethod
+    def _default_compiler(cls):
         return TabularNeuralNetTorchNativeCompiler
 
     def _ag_params(self) -> set:

{autogluon.tabular-1.2.1b20250304 → autogluon.tabular-1.2.1b20250306}/src/autogluon/tabular/models/text_prediction/text_prediction_v1_model.py

@@ -1,3 +1,5 @@
+from __future__ import annotations
+
 import logging
 
 from autogluon.common.features.types import (
@@ -30,3 +32,7 @@ class TextPredictorModel(MultiModalPredictorModel):
         )
         default_auxiliary_params.update(extra_auxiliary_params)
         return default_auxiliary_params
+
+    @classmethod
+    def supported_problem_types(cls) -> list[str] | None:
+        return ["binary", "multiclass", "regression"]

{autogluon.tabular-1.2.1b20250304 → autogluon.tabular-1.2.1b20250306}/src/autogluon/tabular/models/vowpalwabbit/vowpalwabbit_model.py

@@ -1,3 +1,5 @@
+from __future__ import annotations
+
 import logging
 import os
 import time
@@ -267,11 +269,14 @@ class VowpalWabbitModel(AbstractModel):
         default_ag_args = super()._get_default_ag_args()
         extra_ag_args = {
             "valid_stacker": False,
-            "problem_types": [BINARY, MULTICLASS, REGRESSION],
         }
         default_ag_args.update(extra_ag_args)
         return default_ag_args
 
+    @classmethod
+    def supported_problem_types(cls) -> list[str] | None:
+        return ["binary", "multiclass", "regression"]
+
     def _more_tags(self):
         # `can_refit_full=True` because best epoch is communicated at end of `_fit`: `self.params_trained['passes'] = epoch`
         return {"can_refit_full": True}

{autogluon.tabular-1.2.1b20250304 → autogluon.tabular-1.2.1b20250306}/src/autogluon/tabular/models/xgboost/xgboost_model.py

@@ -1,3 +1,5 @@
+from __future__ import annotations
+
 import logging
 import math
 import os
@@ -141,10 +143,11 @@ class XGBoostModel(AbstractModel):
                 eval_set["test"] = (X_test, y_test)
 
         if num_gpus != 0:
-            params["tree_method"] = "gpu_hist"
-            if "gpu_id" not in params:
-                params["gpu_id"] = 0
-        elif "tree_method" not in params:
+            if "device" not in params:
+                # FIXME: figure out which GPUs are available to this model instead of hardcoding GPU 0.
+                #  Need to update BaggedEnsembleModel
+                params["device"] = "cuda:0"
+        if "tree_method" not in params:
             params["tree_method"] = "hist"
 
         try_import_xgboost()

{autogluon.tabular-1.2.1b20250304 → autogluon.tabular-1.2.1b20250306}/src/autogluon/tabular/models/xt/xt_model.py

@@ -1,3 +1,5 @@
+from __future__ import annotations
+
 from autogluon.core.constants import QUANTILE, REGRESSION
 
 from ..rf.rf_model import RFModel

autogluon.tabular-1.2.1b20250306/src/autogluon/tabular/testing/__init__.py

@@ -0,0 +1,2 @@
+from .fit_helper import FitHelper
+from .model_fit_helper import ModelFitHelper