autogluon.tabular 1.4.0b20250724__tar.gz → 1.4.0b20250726__tar.gz

{autogluon.tabular-1.4.0b20250724 → autogluon.tabular-1.4.0b20250726}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: autogluon.tabular
-Version: 1.4.0b20250724
+Version: 1.4.0b20250726
 Summary: Fast and Accurate ML in 3 Lines of Code
 Home-page: https://github.com/autogluon/autogluon
 Author: AutoGluon Community
@@ -96,7 +96,7 @@ Build accurate end-to-end ML models in just 3 lines of code!
 
 ```python
 from autogluon.tabular import TabularPredictor
-predictor = TabularPredictor(label="class").fit("train.csv")
+predictor = TabularPredictor(label="class").fit("train.csv", presets="best")
 predictions = predictor.predict("test.csv")
 ```
 

{autogluon.tabular-1.4.0b20250724 → autogluon.tabular-1.4.0b20250726}/src/autogluon/tabular/configs/presets_configs.py

@@ -75,7 +75,7 @@ tabular_presets_dict = dict(
     # Experimental presets. Only use these presets if you are ok with unstable and potentially poor performing presets.
     #  Experimental presets can be removed or changed without warning.
 
-    # [EXPERIMENTAL PRESET] The `experimental_quality` preset may be changed or removed without warning.
+    # [EXPERIMENTAL PRESET] The `extreme` preset may be changed or removed without warning.
     # This preset acts as a testing ground for cutting edge features and models which could later be added to the `best_quality` preset in future releases.
     # Using this preset can lead to unexpected crashes, as it hasn't been as thoroughly tested as other presets.
     # Absolute best predictive accuracy with **zero** consideration to inference time or disk usage.
@@ -83,9 +83,9 @@ tabular_presets_dict = dict(
     # Significantly stronger than `best_quality`, but can be over 10x slower in inference.
     # Uses pre-trained tabular foundation models, which add a minimum of 1-2 GB to the predictor artifact's size.
     # For best results, use as large of an instance as possible with a GPU and as many CPU cores as possible (ideally 64+ cores)
-    # Aliases: experimental
+    # Aliases: extreme, experimental, experimental_quality
     # GPU STRONGLY RECOMMENDED
-    experimental_quality={
+    extreme_quality={
         "auto_stack": True,
         "dynamic_stacking": "auto",
         "num_bag_sets": 1,
@@ -128,7 +128,7 @@ tabular_presets_dict = dict(
 
 # Alias preset name alternatives
 tabular_presets_alias = dict(
-    experimental="experimental_quality",
+    extreme="extreme_quality",
     best="best_quality",
     high="high_quality",
     high_quality_fast_inference_only_refit="high_quality",
@@ -136,10 +136,12 @@ tabular_presets_alias = dict(
     good_quality_faster_inference_only_refit="good_quality",
     medium="medium_quality",
     medium_quality_faster_train="medium_quality",
-    eq="experimental_quality",
+    eq="extreme_quality",
     bq="best_quality",
     hq="high_quality",
     gq="good_quality",
     mq="medium_quality",
-    experimental_quality_v140="experimental_quality",
+    experimental="extreme_quality",
+    experimental_quality="extreme_quality",
+    experimental_quality_v140="extreme_quality",
 )

{autogluon.tabular-1.4.0b20250724 → autogluon.tabular-1.4.0b20250726}/src/autogluon/tabular/configs/zeroshot/zeroshot_portfolio_2025.py

@@ -303,6 +303,7 @@ hyperparameter_portfolio_zeroshot_2025_small = {
             "n_estimators": 1,
             "fine_tune": True,
             "fine_tune_steps": 50,
+            "ag.num_gpus": 1,
             "ag_args": {'priority': -21},
         },
     ],

{autogluon.tabular-1.4.0b20250724 → autogluon.tabular-1.4.0b20250726}/src/autogluon/tabular/models/automm/automm_model.py

@@ -65,6 +65,8 @@ class MultiModalPredictorModel(AbstractModel):
             Names of the features.
         feature_metadata
             The feature metadata.
+
+        .. versionadded:: 0.3.0
         """
         super().__init__(**kwargs)
         self._label_column_name = None

{autogluon.tabular-1.4.0b20250724 → autogluon.tabular-1.4.0b20250726}/src/autogluon/tabular/models/automm/ft_transformer.py

@@ -17,7 +17,8 @@ class FTTransformerModel(MultiModalPredictorModel):
     ag_name = "FTTransformer"
 
     def __init__(self, **kwargs):
-        """Wrapper of autogluon.multimodal.MultiModalPredictor.
+        """
+        FT-Transformer model.
 
         The features can be a mix of
         - categorical column
@@ -48,6 +49,8 @@ class FTTransformerModel(MultiModalPredictorModel):
             Names of the features.
         feature_metadata
             The feature metadata.
+
+        .. versionadded:: 0.6.0
         """
         super().__init__(**kwargs)
 

{autogluon.tabular-1.4.0b20250724 → autogluon.tabular-1.4.0b20250726}/src/autogluon/tabular/models/mitra/mitra_model.py

@@ -15,6 +15,19 @@ logger = logging.getLogger(__name__)
 
 
 class MitraModel(AbstractModel):
+    """
+    Mitra is a tabular foundation model pre-trained purely on synthetic data with the goal
+    of optimizing fine-tuning performance over in-context learning performance.
+    Mitra was developed by the AutoGluon team @ AWS AI.
+
+    Mitra's default hyperparameters outperforms all methods for small datasets on TabArena-v0.1 (excluding ensembling): https://tabarena.ai
+
+    Authors: Xiyuan Zhang, Danielle C. Maddix, Junming Yin, Nick Erickson, Abdul Fatir Ansari, Boran Han, Shuai Zhang, Leman Akoglu, Christos Faloutsos, Michael W. Mahoney, Cuixiong Hu, Huzefa Rangwala, George Karypis, Bernie Wang
+    Blog Post: https://www.amazon.science/blog/mitra-mixed-synthetic-priors-for-enhancing-tabular-foundation-models
+    License: Apache-2.0
+
+    .. versionadded:: 1.4.0
+    """
     ag_key = "MITRA"
     ag_name = "Mitra"
     weights_file_name = "model.pt"
@@ -106,6 +119,13 @@ class MitraModel(AbstractModel):
             else:
                 hyp["device"] = self._get_default_device()
 
+        if hyp["device"] == "cpu" and hyp.get("fine_tune", True):
+            logger.log(
+                30,
+                f"\tWarning: Attempting to fine-tune Mitra on CPU. This will be very slow. "
+                f"We strongly recommend using a GPU instance to fine-tune Mitra."
+            )
+
         if "state_dict_classification" in hyp:
             state_dict_classification = hyp.pop("state_dict_classification")
             if self.problem_type in ["binary", "multiclass"]:
@@ -216,24 +236,6 @@ class MitraModel(AbstractModel):
 
         return num_cpus, num_gpus
 
-    def get_minimum_resources(self, is_gpu_available: bool = False) -> dict[str, int | float]:
-        """
-        Parameters
-        ----------
-        is_gpu_available : bool, default = False
-            Whether gpu is available in the system.
-            Model that can be trained both on cpu and gpu can decide the minimum resources based on this.
-
-        Returns a dictionary of minimum resource requirements to fit the model.
-        Subclass should consider overriding this method if it requires more resources to train.
-        If a resource is not part of the output dictionary, it is considered unnecessary.
-        Valid keys: 'num_cpus', 'num_gpus'.
-        """
-        return {
-            "num_cpus": 1,
-            "num_gpus": 0.5,
-        }
-
     def _estimate_memory_usage(self, X: pd.DataFrame, **kwargs) -> int:
         return self.estimate_memory_usage_static(
             X=X, problem_type=self.problem_type, num_classes=self.num_classes, **kwargs

{autogluon.tabular-1.4.0b20250724 → autogluon.tabular-1.4.0b20250726}/src/autogluon/tabular/models/realmlp/realmlp_model.py

@@ -1,11 +1,5 @@
 """
 Code Adapted from TabArena: https://github.com/autogluon/tabrepo/blob/main/tabrepo/benchmark/models/ag/realmlp/realmlp_model.py
-
-Model: RealMLP
-Paper: Better by Default: Strong Pre-Tuned MLPs and Boosted Trees on Tabular Data
-Authors: David Holzmüller, Léo Grinsztajn, Ingo Steinwart
-Codebase: https://github.com/dholzmueller/pytabkit
-License: Apache-2.0
 """
 
 from __future__ import annotations
@@ -41,6 +35,19 @@ def set_logger_level(logger_name: str, level: int):
 
 # pip install pytabkit
 class RealMLPModel(AbstractModel):
+    """
+    RealMLP is an improved multilayer perception (MLP) model
+    through a bag of tricks and better default hyperparameters.
+
+    RealMLP is the top performing method overall on TabArena-v0.1: https://tabarena.ai
+
+    Paper: Better by Default: Strong Pre-Tuned MLPs and Boosted Trees on Tabular Data
+    Authors: David Holzmüller, Léo Grinsztajn, Ingo Steinwart
+    Codebase: https://github.com/dholzmueller/pytabkit
+    License: Apache-2.0
+
+    .. versionadded:: 1.4.0
+    """
     ag_key = "REALMLP"
     ag_name = "RealMLP"
     ag_priority = 75

{autogluon.tabular-1.4.0b20250724 → autogluon.tabular-1.4.0b20250726}/src/autogluon/tabular/models/tabicl/tabicl_model.py

@@ -1,10 +1,5 @@
 """
 Code Adapted from TabArena: https://github.com/autogluon/tabrepo/blob/main/tabrepo/benchmark/models/ag/tabicl/tabicl_model.py
-Model: TabICL
-Paper: TabICL: A Tabular Foundation Model for In-Context Learning on Large Data
-Authors: Jingang Qu, David Holzmüller, Gaël Varoquaux, Marine Le Morvan
-Codebase: https://github.com/soda-inria/tabicl
-License: BSD-3-Clause
 """
 
 from __future__ import annotations
@@ -23,6 +18,20 @@ logger = logging.getLogger(__name__)
 
 # TODO: Verify if crashes when weights are not yet downloaded and fit in parallel
 class TabICLModel(AbstractModel):
+    """
+    TabICL is a foundation model for tabular data using in-context learning
+    that is scalable to larger datasets than TabPFNv2. It is pretrained purely on synthetic data.
+    TabICL currently only supports classification tasks.
+
+    TabICL is one of the top performing methods overall on TabArena-v0.1: https://tabarena.ai
+
+    Paper: TabICL: A Tabular Foundation Model for In-Context Learning on Large Data
+    Authors: Jingang Qu, David Holzmüller, Gaël Varoquaux, Marine Le Morvan
+    Codebase: https://github.com/soda-inria/tabicl
+    License: BSD-3-Clause
+
+    .. versionadded:: 1.4.0
+    """
     ag_key = "TABICL"
     ag_name = "TabICL"
     ag_priority = 65

{autogluon.tabular-1.4.0b20250724 → autogluon.tabular-1.4.0b20250726}/src/autogluon/tabular/models/tabm/tabm_model.py

@@ -4,12 +4,6 @@ Note: This is a custom implementation of TabM based on TabArena. Because the Aut
 the same time as TabM became available on PyPi, we chose to use TabArena's implementation
 for the AutoGluon 1.4 release as it has already been benchmarked.
 
-Model: TabM
-Paper: TabM: Advancing Tabular Deep Learning with Parameter-Efficient Ensembling
-Authors: Yury Gorishniy, Akim Kotelnikov, Artem Babenko
-Codebase: https://github.com/yandex-research/tabm
-License: Apache-2.0
-
 Partially adapted from pytabkit's TabM implementation.
 """
 
@@ -28,6 +22,20 @@ logger = logging.getLogger(__name__)
 
 
 class TabMModel(AbstractModel):
+    """
+    TabM is an efficient ensemble of MLPs that is trained simultaneously with mostly shared parameters.
+
+    TabM is one of the top performing methods overall on TabArena-v0.1: https://tabarena.ai
+
+    Paper: TabM: Advancing Tabular Deep Learning with Parameter-Efficient Ensembling
+    Authors: Yury Gorishniy, Akim Kotelnikov, Artem Babenko
+    Codebase: https://github.com/yandex-research/tabm
+    License: Apache-2.0
+
+    Partially adapted from pytabkit's TabM implementation.
+
+    .. versionadded:: 1.4.0
+    """
     ag_key = "TABM"
     ag_name = "TabM"
     ag_priority = 85

{autogluon.tabular-1.4.0b20250724 → autogluon.tabular-1.4.0b20250726}/src/autogluon/tabular/models/tabpfnmix/tabpfnmix_model.py

@@ -26,6 +26,8 @@ class TabPFNMixModel(AbstractModel):
 
     TabPFNMix is based off of the TabPFN and TabForestPFN models.
 
+    We recommend using Mitra instead, as it is an improved version of TabPFNMix.
+
     It is a tabular transformer model pre-trained on purely synthetic data.
 
     It currently has several limitations:
@@ -34,6 +36,8 @@ class TabPFNMixModel(AbstractModel):
     3. Does not support GPU
 
     For more information, refer to the `./_internals/README.md` file.
+
+    .. versionadded:: 1.2.0
     """
     ag_key = "TABPFNMIX"
     ag_name = "TabPFNMix"

{autogluon.tabular-1.4.0b20250724 → autogluon.tabular-1.4.0b20250726}/src/autogluon/tabular/models/tabpfnv2/tabpfnv2_model.py

@@ -1,11 +1,5 @@
 """
 Code Adapted from TabArena: https://github.com/autogluon/tabrepo/blob/main/tabrepo/benchmark/models/ag/tabpfnv2/tabpfnv2_model.py
-
-Model: TabPFNv2
-Paper: Accurate predictions on small data with a tabular foundation model
-Authors: Noah Hollmann, Samuel Müller, Lennart Purucker, Arjun Krishnakumar, Max Körfer, Shi Bin Hoo, Robin Tibor Schirrmeister & Frank Hutter
-Codebase: https://github.com/PriorLabs/TabPFN
-License: https://github.com/PriorLabs/TabPFN/blob/main/LICENSE
 """
 
 from __future__ import annotations
@@ -111,6 +105,20 @@ class FixedSafePowerTransformer(PowerTransformer):
 
 
 class TabPFNV2Model(AbstractModel):
+    """
+    TabPFNv2 is a tabular foundation model pre-trained purely on synthetic data that achieves
+    state-of-the-art results with in-context learning on small datasets with <=10000 samples and <=500 features.
+    TabPFNv2 is developed and maintained by PriorLabs: https://priorlabs.ai/
+
+    TabPFNv2 is the top performing method for small datasets on TabArena-v0.1: https://tabarena.ai
+
+    Paper: Accurate predictions on small data with a tabular foundation model
+    Authors: Noah Hollmann, Samuel Müller, Lennart Purucker, Arjun Krishnakumar, Max Körfer, Shi Bin Hoo, Robin Tibor Schirrmeister & Frank Hutter
+    Codebase: https://github.com/PriorLabs/TabPFN
+    License: https://github.com/PriorLabs/TabPFN/blob/main/LICENSE
+
+    .. versionadded:: 1.4.0
+    """
     ag_key = "TABPFNV2"
     ag_name = "TabPFNv2"
     ag_priority = 105

{autogluon.tabular-1.4.0b20250724 → autogluon.tabular-1.4.0b20250726}/src/autogluon/tabular/predictor/predictor.py

@@ -1068,11 +1068,11 @@ class TabularPredictor:
                 20,
                 "No presets specified! To achieve strong results with AutoGluon, it is recommended to use the available presets. Defaulting to `'medium'`...\n"
                 "\tRecommended Presets (For more details refer to https://auto.gluon.ai/stable/tutorials/tabular/tabular-essentials.html#presets):\n"
-                "\tpresets='experimental' : New in v1.4: Massively better than 'best' on datasets <10000 samples by using new models: TabPFNv2, TabICL, Mitra, and TabM. Absolute best accuracy. Requires a GPU. Recommended 64 GB CPU memory and 32+ GB GPU memory.\n"
-                "\tpresets='best'         : Maximize accuracy. Recommended for most users. Use in competitions and benchmarks.\n"
-                "\tpresets='high'         : Strong accuracy with fast inference speed.\n"
-                "\tpresets='good'         : Good accuracy with very fast inference speed.\n"
-                "\tpresets='medium'       : Fast training time, ideal for initial prototyping.",
+                "\tpresets='extreme' : New in v1.4: Massively better than 'best' on datasets <30000 samples by using new models meta-learned on https://tabarena.ai: TabPFNv2, TabICL, Mitra, and TabM. Absolute best accuracy. Requires a GPU. Recommended 64 GB CPU memory and 32+ GB GPU memory.\n"
+                "\tpresets='best'    : Maximize accuracy. Recommended for most users. Use in competitions and benchmarks.\n"
+                "\tpresets='high'    : Strong accuracy with fast inference speed.\n"
+                "\tpresets='good'    : Good accuracy with very fast inference speed.\n"
+                "\tpresets='medium'  : Fast training time, ideal for initial prototyping.",
             )
 
         kwargs_orig = kwargs.copy()
@@ -1130,7 +1130,7 @@ class TabularPredictor:
         # TODO: Temporary for v1.4. Make this more extensible for v1.5 by letting users make their own dynamic hyperparameters.
         dynamic_hyperparameters = kwargs["_experimental_dynamic_hyperparameters"]
         if dynamic_hyperparameters:
-            logger.log(20, f"Experimental preset uses a dynamic portfolio based on dataset size...")
+            logger.log(20, f"`extreme` preset uses a dynamic portfolio based on dataset size...")
             assert hyperparameters is None, f"hyperparameters must be unspecified when `_experimental_dynamic_hyperparameters=True`."
             n_samples = len(train_data)
             if n_samples > 30000:

{autogluon.tabular-1.4.0b20250724 → autogluon.tabular-1.4.0b20250726}/src/autogluon/tabular/version.py

@@ -1,4 +1,4 @@
 """This is the autogluon version file."""
 
-__version__ = "1.4.0b20250724"
+__version__ = "1.4.0b20250726"
 __lite__ = False

{autogluon.tabular-1.4.0b20250724 → autogluon.tabular-1.4.0b20250726}/src/autogluon.tabular.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: autogluon.tabular
-Version: 1.4.0b20250724
+Version: 1.4.0b20250726
 Summary: Fast and Accurate ML in 3 Lines of Code
 Home-page: https://github.com/autogluon/autogluon
 Author: AutoGluon Community
@@ -96,7 +96,7 @@ Build accurate end-to-end ML models in just 3 lines of code!
 
 ```python
 from autogluon.tabular import TabularPredictor
-predictor = TabularPredictor(label="class").fit("train.csv")
+predictor = TabularPredictor(label="class").fit("train.csv", presets="best")
 predictions = predictor.predict("test.csv")
 ```
 

{autogluon.tabular-1.4.0b20250724 → autogluon.tabular-1.4.0b20250726}/src/autogluon.tabular.egg-info/requires.txt

@@ -3,23 +3,23 @@ scipy<1.17,>=1.5.4
 pandas<2.4.0,>=2.0.0
 scikit-learn<1.8.0,>=1.4.0
 networkx<4,>=3.0
-autogluon.core==1.4.0b20250724
-autogluon.features==1.4.0b20250724
+autogluon.core==1.4.0b20250726
+autogluon.features==1.4.0b20250726
 
 [all]
-transformers
-huggingface_hub[torch]
-numpy<2.3.0,>=1.25
-autogluon.core[all]==1.4.0b20250724
-spacy<3.9
-fastai<2.9,>=2.3.1
-torch<2.8,>=2.2
-catboost<1.3,>=1.2
+lightgbm<4.7,>=4.0
 xgboost<3.1,>=2.0
+torch<2.8,>=2.2
 loguru
+numpy<2.3.0,>=1.25
+transformers
+huggingface_hub[torch]
 einx
-lightgbm<4.7,>=4.0
+fastai<2.9,>=2.3.1
+spacy<3.9
 omegaconf
+autogluon.core[all]==1.4.0b20250726
+catboost<1.3,>=1.2
 
 [all:platform_system == "Windows" and python_version == "3.9"]
 blis<1.2.1,>=0.7.0
@@ -51,7 +51,7 @@ transformers
 huggingface_hub[torch]
 
 [ray]
-autogluon.core[all]==1.4.0b20250724
+autogluon.core[all]==1.4.0b20250726
 
 [realmlp]
 pytabkit<1.6,>=1.5
@@ -71,22 +71,22 @@ onnx<1.18.0,>=1.13.0
 onnx<1.16.2,>=1.13.0
 
 [tabarena]
-tabpfn<2.2,>=2.0.9
-transformers
-huggingface_hub[torch]
-pytabkit<1.6,>=1.5
-tabicl<0.2,>=0.1.3
-numpy<2.3.0,>=1.25
-autogluon.core[all]==1.4.0b20250724
-spacy<3.9
-fastai<2.9,>=2.3.1
-torch<2.8,>=2.2
-catboost<1.3,>=1.2
+lightgbm<4.7,>=4.0
 xgboost<3.1,>=2.0
+torch<2.8,>=2.2
 loguru
+numpy<2.3.0,>=1.25
+transformers
+huggingface_hub[torch]
+tabpfn<2.2,>=2.0.9
 einx
-lightgbm<4.7,>=4.0
+fastai<2.9,>=2.3.1
+spacy<3.9
 omegaconf
+tabicl<0.2,>=0.1.3
+pytabkit<1.6,>=1.5
+autogluon.core[all]==1.4.0b20250726
+catboost<1.3,>=1.2
 
 [tabarena:platform_system == "Windows" and python_version == "3.9"]
 blis<1.2.1,>=0.7.0