autogluon.tabular 1.4.1b20251116__tar.gz → 1.4.1b20251217__tar.gz

{autogluon_tabular-1.4.1b20251116/src/autogluon.tabular.egg-info → autogluon_tabular-1.4.1b20251217}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: autogluon.tabular
-Version: 1.4.1b20251116
+Version: 1.4.1b20251217
 Summary: Fast and Accurate ML in 3 Lines of Code
 Home-page: https://github.com/autogluon/autogluon
 Author: AutoGluon Community
@@ -23,15 +23,15 @@ Classifier: Operating System :: Microsoft :: Windows
 Classifier: Operating System :: POSIX
 Classifier: Operating System :: Unix
 Classifier: Programming Language :: Python :: 3
-Classifier: Programming Language :: Python :: 3.9
 Classifier: Programming Language :: Python :: 3.10
 Classifier: Programming Language :: Python :: 3.11
 Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
 Classifier: Topic :: Software Development
 Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
 Classifier: Topic :: Scientific/Engineering :: Information Analysis
 Classifier: Topic :: Scientific/Engineering :: Image Recognition
-Requires-Python: >=3.9, <3.13
+Requires-Python: >=3.10, <3.14
 Description-Content-Type: text/markdown
 License-File: LICENSE
 License-File: NOTICE
@@ -40,8 +40,8 @@ Requires-Dist: scipy<1.17,>=1.5.4
 Requires-Dist: pandas<2.4.0,>=2.0.0
 Requires-Dist: scikit-learn<1.8.0,>=1.4.0
 Requires-Dist: networkx<4,>=3.0
-Requires-Dist: autogluon.core==1.4.1b20251116
-Requires-Dist: autogluon.features==1.4.1b20251116
+Requires-Dist: autogluon.core==1.4.1b20251217
+Requires-Dist: autogluon.features==1.4.1b20251217
 Provides-Extra: lightgbm
 Requires-Dist: lightgbm<4.7,>=4.0; extra == "lightgbm"
 Provides-Extra: catboost
@@ -55,89 +55,86 @@ Provides-Extra: interpret
 Requires-Dist: interpret-core<0.8,>=0.7.2; extra == "interpret"
 Provides-Extra: fastai
 Requires-Dist: spacy<3.9; extra == "fastai"
-Requires-Dist: torch<2.8,>=2.6; extra == "fastai"
+Requires-Dist: torch<2.10,>=2.6; extra == "fastai"
 Requires-Dist: fastai<2.9,>=2.3.1; extra == "fastai"
-Requires-Dist: blis<1.2.1,>=0.7.0; (platform_system == "Windows" and python_version == "3.9") and extra == "fastai"
 Provides-Extra: tabm
-Requires-Dist: torch<2.8,>=2.6; extra == "tabm"
+Requires-Dist: torch<2.10,>=2.6; extra == "tabm"
 Provides-Extra: tabpfn
 Requires-Dist: tabpfn<2.2,>=2.0.9; extra == "tabpfn"
 Provides-Extra: tabpfnmix
-Requires-Dist: torch<2.8,>=2.6; extra == "tabpfnmix"
+Requires-Dist: torch<2.10,>=2.6; extra == "tabpfnmix"
 Requires-Dist: huggingface_hub[torch]<1.0; extra == "tabpfnmix"
 Requires-Dist: einops<0.9,>=0.7; extra == "tabpfnmix"
 Provides-Extra: mitra
 Requires-Dist: loguru; extra == "mitra"
 Requires-Dist: einx; extra == "mitra"
 Requires-Dist: omegaconf; extra == "mitra"
-Requires-Dist: torch<2.8,>=2.6; extra == "mitra"
+Requires-Dist: torch<2.10,>=2.6; extra == "mitra"
 Requires-Dist: transformers; extra == "mitra"
 Requires-Dist: huggingface_hub[torch]<1.0; extra == "mitra"
 Requires-Dist: einops<0.9,>=0.7; extra == "mitra"
 Provides-Extra: tabicl
 Requires-Dist: tabicl<0.2,>=0.1.3; extra == "tabicl"
 Provides-Extra: ray
-Requires-Dist: autogluon.core[all]==1.4.1b20251116; extra == "ray"
+Requires-Dist: autogluon.core[all]==1.4.1b20251217; extra == "ray"
 Provides-Extra: skex
 Requires-Dist: scikit-learn-intelex<2025.5,>=2024.0; extra == "skex"
 Provides-Extra: imodels
 Requires-Dist: imodels<2.1.0,>=1.3.10; extra == "imodels"
 Provides-Extra: skl2onnx
-Requires-Dist: onnx<1.16.2,>=1.13.0; platform_system == "Windows" and extra == "skl2onnx"
-Requires-Dist: onnx<1.18.0,>=1.13.0; platform_system != "Windows" and extra == "skl2onnx"
-Requires-Dist: skl2onnx<1.18.0,>=1.15.0; extra == "skl2onnx"
-Requires-Dist: onnxruntime<1.20.0,>=1.17.0; extra == "skl2onnx"
-Requires-Dist: onnxruntime-gpu<1.20.0,>=1.17.0; extra == "skl2onnx"
+Requires-Dist: skl2onnx<1.20.0,>=1.15.0; extra == "skl2onnx"
+Requires-Dist: onnx!=1.16.2,<1.21.0,>=1.13.0; platform_system == "Windows" and extra == "skl2onnx"
+Requires-Dist: onnx<1.21.0,>=1.13.0; platform_system != "Windows" and extra == "skl2onnx"
+Requires-Dist: onnxruntime<1.24.0,>=1.17.0; extra == "skl2onnx"
+Requires-Dist: onnxruntime-gpu<1.24.0,>=1.17.0; (platform_system != "Darwin" and platform_machine != "aarch64") and extra == "skl2onnx"
 Provides-Extra: all
-Requires-Dist: einops<0.9,>=0.7; extra == "all"
-Requires-Dist: numpy<2.3.0,>=1.25; extra == "all"
+Requires-Dist: autogluon.core[all]==1.4.1b20251217; extra == "all"
 Requires-Dist: loguru; extra == "all"
-Requires-Dist: omegaconf; extra == "all"
 Requires-Dist: spacy<3.9; extra == "all"
+Requires-Dist: huggingface_hub[torch]<1.0; extra == "all"
 Requires-Dist: einx; extra == "all"
-Requires-Dist: blis<1.2.1,>=0.7.0; (platform_system == "Windows" and python_version == "3.9") and extra == "all"
-Requires-Dist: catboost<1.3,>=1.2; extra == "all"
-Requires-Dist: xgboost<3.1,>=2.0; extra == "all"
 Requires-Dist: fastai<2.9,>=2.3.1; extra == "all"
-Requires-Dist: autogluon.core[all]==1.4.1b20251116; extra == "all"
-Requires-Dist: huggingface_hub[torch]<1.0; extra == "all"
-Requires-Dist: torch<2.8,>=2.6; extra == "all"
+Requires-Dist: xgboost<3.1,>=2.0; extra == "all"
+Requires-Dist: torch<2.10,>=2.6; extra == "all"
+Requires-Dist: numpy<2.3.0,>=1.25; extra == "all"
+Requires-Dist: einops<0.9,>=0.7; extra == "all"
 Requires-Dist: lightgbm<4.7,>=4.0; extra == "all"
+Requires-Dist: omegaconf; extra == "all"
+Requires-Dist: catboost<1.3,>=1.2; extra == "all"
 Requires-Dist: transformers; extra == "all"
 Provides-Extra: tabarena
-Requires-Dist: tabicl<0.2,>=0.1.3; extra == "tabarena"
-Requires-Dist: omegaconf; extra == "tabarena"
-Requires-Dist: spacy<3.9; extra == "tabarena"
 Requires-Dist: catboost<1.3,>=1.2; extra == "tabarena"
-Requires-Dist: torch<2.8,>=2.6; extra == "tabarena"
-Requires-Dist: lightgbm<4.7,>=4.0; extra == "tabarena"
+Requires-Dist: autogluon.core[all]==1.4.1b20251217; extra == "tabarena"
+Requires-Dist: tabicl<0.2,>=0.1.3; extra == "tabarena"
 Requires-Dist: loguru; extra == "tabarena"
-Requires-Dist: interpret-core<0.8,>=0.7.2; extra == "tabarena"
-Requires-Dist: tabpfn<2.2,>=2.0.9; extra == "tabarena"
-Requires-Dist: pytabkit<1.7,>=1.6; extra == "tabarena"
-Requires-Dist: transformers; extra == "tabarena"
-Requires-Dist: einops<0.9,>=0.7; extra == "tabarena"
-Requires-Dist: numpy<2.3.0,>=1.25; extra == "tabarena"
-Requires-Dist: xgboost<3.1,>=2.0; extra == "tabarena"
-Requires-Dist: autogluon.core[all]==1.4.1b20251116; extra == "tabarena"
+Requires-Dist: huggingface_hub[torch]<1.0; extra == "tabarena"
 Requires-Dist: einx; extra == "tabarena"
-Requires-Dist: blis<1.2.1,>=0.7.0; (platform_system == "Windows" and python_version == "3.9") and extra == "tabarena"
+Requires-Dist: pytabkit<1.7,>=1.6; extra == "tabarena"
 Requires-Dist: fastai<2.9,>=2.3.1; extra == "tabarena"
-Requires-Dist: huggingface_hub[torch]<1.0; extra == "tabarena"
+Requires-Dist: xgboost<3.1,>=2.0; extra == "tabarena"
+Requires-Dist: torch<2.10,>=2.6; extra == "tabarena"
+Requires-Dist: numpy<2.3.0,>=1.25; extra == "tabarena"
+Requires-Dist: einops<0.9,>=0.7; extra == "tabarena"
+Requires-Dist: lightgbm<4.7,>=4.0; extra == "tabarena"
+Requires-Dist: omegaconf; extra == "tabarena"
+Requires-Dist: spacy<3.9; extra == "tabarena"
+Requires-Dist: transformers; extra == "tabarena"
+Requires-Dist: tabpfn<2.2,>=2.0.9; extra == "tabarena"
+Requires-Dist: interpret-core<0.8,>=0.7.2; extra == "tabarena"
 Provides-Extra: tests
 Requires-Dist: interpret-core<0.8,>=0.7.2; extra == "tests"
 Requires-Dist: tabicl<0.2,>=0.1.3; extra == "tests"
 Requires-Dist: tabpfn<2.2,>=2.0.9; extra == "tests"
 Requires-Dist: pytabkit<1.7,>=1.6; extra == "tests"
-Requires-Dist: torch<2.8,>=2.6; extra == "tests"
+Requires-Dist: torch<2.10,>=2.6; extra == "tests"
 Requires-Dist: huggingface_hub[torch]<1.0; extra == "tests"
 Requires-Dist: einops<0.9,>=0.7; extra == "tests"
 Requires-Dist: imodels<2.1.0,>=1.3.10; extra == "tests"
-Requires-Dist: onnx<1.16.2,>=1.13.0; platform_system == "Windows" and extra == "tests"
-Requires-Dist: onnx<1.18.0,>=1.13.0; platform_system != "Windows" and extra == "tests"
-Requires-Dist: skl2onnx<1.18.0,>=1.15.0; extra == "tests"
-Requires-Dist: onnxruntime<1.20.0,>=1.17.0; extra == "tests"
-Requires-Dist: onnxruntime-gpu<1.20.0,>=1.17.0; extra == "tests"
+Requires-Dist: skl2onnx<1.20.0,>=1.15.0; extra == "tests"
+Requires-Dist: onnx!=1.16.2,<1.21.0,>=1.13.0; platform_system == "Windows" and extra == "tests"
+Requires-Dist: onnx<1.21.0,>=1.13.0; platform_system != "Windows" and extra == "tests"
+Requires-Dist: onnxruntime<1.24.0,>=1.17.0; extra == "tests"
+Requires-Dist: onnxruntime-gpu<1.24.0,>=1.17.0; (platform_system != "Darwin" and platform_machine != "aarch64") and extra == "tests"
 Dynamic: author
 Dynamic: classifier
 Dynamic: description
@@ -160,7 +157,7 @@ Dynamic: summary
 
 [![Latest Release](https://img.shields.io/github/v/release/autogluon/autogluon)](https://github.com/autogluon/autogluon/releases)
 [![Conda Forge](https://img.shields.io/conda/vn/conda-forge/autogluon.svg)](https://anaconda.org/conda-forge/autogluon)
-[![Python Versions](https://img.shields.io/badge/python-3.9%20%7C%203.10%20%7C%203.11%20%7C%203.12-blue)](https://pypi.org/project/autogluon/)
+[![Python Versions](https://img.shields.io/badge/python-3.10%20%7C%203.11%20%7C%203.12%20%7C%203.13-blue)](https://pypi.org/project/autogluon/)
 [![Downloads](https://pepy.tech/badge/autogluon/month)](https://pepy.tech/project/autogluon)
 [![GitHub license](https://img.shields.io/badge/License-Apache_2.0-blue.svg)](./LICENSE)
 [![Discord](https://img.shields.io/discord/1043248669505368144?color=7289da&label=Discord&logo=discord&logoColor=ffffff)](https://discord.gg/wjUmjqAc2N)
@@ -177,7 +174,7 @@ AutoGluon, developed by AWS AI, automates machine learning tasks enabling you to
 
 ## 💾 Installation
 
-AutoGluon is supported on Python 3.9 - 3.12 and is available on Linux, MacOS, and Windows.
+AutoGluon is supported on Python 3.10 - 3.13 and is available on Linux, MacOS, and Windows.
 
 You can install AutoGluon with:
 
@@ -200,8 +197,8 @@ predictions = predictor.predict("test.csv")
 | AutoGluon Task      |                                                                                Quickstart                                                                                |                                                                                API                                                                                |
 |:--------------------|:------------------------------------------------------------------------------------------------------------------------------------------------------------------------:|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------:|
 | TabularPredictor    | [![Quick Start](https://img.shields.io/static/v1?label=&message=tutorial&color=grey)](https://auto.gluon.ai/stable/tutorials/tabular/tabular-quick-start.html) |                 [![API](https://img.shields.io/badge/api-reference-blue.svg)](https://auto.gluon.ai/stable/api/autogluon.tabular.TabularPredictor.html)                 |
-| MultiModalPredictor | [![Quick Start](https://img.shields.io/static/v1?label=&message=tutorial&color=grey)](https://auto.gluon.ai/stable/tutorials/multimodal/multimodal_prediction/multimodal-quick-start.html)            | [![API](https://img.shields.io/badge/api-reference-blue.svg)](https://auto.gluon.ai/stable/api/autogluon.multimodal.MultiModalPredictor.html) |
 | TimeSeriesPredictor | [![Quick Start](https://img.shields.io/static/v1?label=&message=tutorial&color=grey)](https://auto.gluon.ai/stable/tutorials/timeseries/forecasting-quick-start.html)            | [![API](https://img.shields.io/badge/api-reference-blue.svg)](https://auto.gluon.ai/stable/api/autogluon.timeseries.TimeSeriesPredictor.html) |
+| MultiModalPredictor | [![Quick Start](https://img.shields.io/static/v1?label=&message=tutorial&color=grey)](https://auto.gluon.ai/stable/tutorials/multimodal/multimodal_prediction/multimodal-quick-start.html)            | [![API](https://img.shields.io/badge/api-reference-blue.svg)](https://auto.gluon.ai/stable/api/autogluon.multimodal.MultiModalPredictor.html) |
 
 ## :mag: Resources
 
@@ -224,7 +221,10 @@ Below is a curated list of recent tutorials and talks on AutoGluon. A comprehens
 - [Benchmarking Multimodal AutoML for Tabular Data with Text Fields](https://datasets-benchmarks-proceedings.neurips.cc/paper/2021/file/9bf31c7ff062936a96d3c8bd1f8f2ff3-Paper-round2.pdf) (*NeurIPS*, 2021) ([BibTeX](CITING.md#autogluonmultimodal))
 - [XTab: Cross-table Pretraining for Tabular Transformers](https://proceedings.mlr.press/v202/zhu23k/zhu23k.pdf) (*ICML*, 2023)
 - [AutoGluon-TimeSeries: AutoML for Probabilistic Time Series Forecasting](https://arxiv.org/abs/2308.05566) (*AutoML Conf*, 2023) ([BibTeX](CITING.md#autogluontimeseries))
-- [TabRepo: A Large Scale Repository of Tabular Model Evaluations and its AutoML Applications](https://arxiv.org/pdf/2311.02971.pdf) (*Under Review*, 2024)
+- [TabRepo: A Large Scale Repository of Tabular Model Evaluations and its AutoML Applications](https://arxiv.org/pdf/2311.02971.pdf) (*AutoML Conf*, 2024)
+- [AutoGluon-Multimodal (AutoMM): Supercharging Multimodal AutoML with Foundation Models](https://arxiv.org/pdf/2404.16233) (*AutoML Conf*, 2024) ([BibTeX](CITING.md#autogluonmultimodal))
+- [Multi-layer Stack Ensembles for Time Series Forecasting](https://arxiv.org/abs/2511.15350) (*AutoML Conf*, 2025) ([BibTeX](CITING.md#autogluontimeseries))
+- [Chronos-2: From Univariate to Universal Forecasting](https://arxiv.org/abs/2510.15821) (*Arxiv*, 2025) ([BibTeX](CITING.md#autogluontimeseries))
 
 ### Articles
 - [AutoGluon-TimeSeries: Every Time Series Forecasting Model In One Library](https://towardsdatascience.com/autogluon-timeseries-every-time-series-forecasting-model-in-one-library-29a3bf6879db) (*Towards Data Science*, Jan 2024)

{autogluon_tabular-1.4.1b20251116 → autogluon_tabular-1.4.1b20251217}/README.md

@@ -7,7 +7,7 @@
 
 [![Latest Release](https://img.shields.io/github/v/release/autogluon/autogluon)](https://github.com/autogluon/autogluon/releases)
 [![Conda Forge](https://img.shields.io/conda/vn/conda-forge/autogluon.svg)](https://anaconda.org/conda-forge/autogluon)
-[![Python Versions](https://img.shields.io/badge/python-3.9%20%7C%203.10%20%7C%203.11%20%7C%203.12-blue)](https://pypi.org/project/autogluon/)
+[![Python Versions](https://img.shields.io/badge/python-3.10%20%7C%203.11%20%7C%203.12%20%7C%203.13-blue)](https://pypi.org/project/autogluon/)
 [![Downloads](https://pepy.tech/badge/autogluon/month)](https://pepy.tech/project/autogluon)
 [![GitHub license](https://img.shields.io/badge/License-Apache_2.0-blue.svg)](./LICENSE)
 [![Discord](https://img.shields.io/discord/1043248669505368144?color=7289da&label=Discord&logo=discord&logoColor=ffffff)](https://discord.gg/wjUmjqAc2N)
@@ -24,7 +24,7 @@ AutoGluon, developed by AWS AI, automates machine learning tasks enabling you to
 
 ## 💾 Installation
 
-AutoGluon is supported on Python 3.9 - 3.12 and is available on Linux, MacOS, and Windows.
+AutoGluon is supported on Python 3.10 - 3.13 and is available on Linux, MacOS, and Windows.
 
 You can install AutoGluon with:
 
@@ -47,8 +47,8 @@ predictions = predictor.predict("test.csv")
 | AutoGluon Task      |                                                                                Quickstart                                                                                |                                                                                API                                                                                |
 |:--------------------|:------------------------------------------------------------------------------------------------------------------------------------------------------------------------:|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------:|
 | TabularPredictor    | [![Quick Start](https://img.shields.io/static/v1?label=&message=tutorial&color=grey)](https://auto.gluon.ai/stable/tutorials/tabular/tabular-quick-start.html) |                 [![API](https://img.shields.io/badge/api-reference-blue.svg)](https://auto.gluon.ai/stable/api/autogluon.tabular.TabularPredictor.html)                 |
-| MultiModalPredictor | [![Quick Start](https://img.shields.io/static/v1?label=&message=tutorial&color=grey)](https://auto.gluon.ai/stable/tutorials/multimodal/multimodal_prediction/multimodal-quick-start.html)            | [![API](https://img.shields.io/badge/api-reference-blue.svg)](https://auto.gluon.ai/stable/api/autogluon.multimodal.MultiModalPredictor.html) |
 | TimeSeriesPredictor | [![Quick Start](https://img.shields.io/static/v1?label=&message=tutorial&color=grey)](https://auto.gluon.ai/stable/tutorials/timeseries/forecasting-quick-start.html)            | [![API](https://img.shields.io/badge/api-reference-blue.svg)](https://auto.gluon.ai/stable/api/autogluon.timeseries.TimeSeriesPredictor.html) |
+| MultiModalPredictor | [![Quick Start](https://img.shields.io/static/v1?label=&message=tutorial&color=grey)](https://auto.gluon.ai/stable/tutorials/multimodal/multimodal_prediction/multimodal-quick-start.html)            | [![API](https://img.shields.io/badge/api-reference-blue.svg)](https://auto.gluon.ai/stable/api/autogluon.multimodal.MultiModalPredictor.html) |
 
 ## :mag: Resources
 
@@ -71,7 +71,10 @@ Below is a curated list of recent tutorials and talks on AutoGluon. A comprehens
 - [Benchmarking Multimodal AutoML for Tabular Data with Text Fields](https://datasets-benchmarks-proceedings.neurips.cc/paper/2021/file/9bf31c7ff062936a96d3c8bd1f8f2ff3-Paper-round2.pdf) (*NeurIPS*, 2021) ([BibTeX](CITING.md#autogluonmultimodal))
 - [XTab: Cross-table Pretraining for Tabular Transformers](https://proceedings.mlr.press/v202/zhu23k/zhu23k.pdf) (*ICML*, 2023)
 - [AutoGluon-TimeSeries: AutoML for Probabilistic Time Series Forecasting](https://arxiv.org/abs/2308.05566) (*AutoML Conf*, 2023) ([BibTeX](CITING.md#autogluontimeseries))
-- [TabRepo: A Large Scale Repository of Tabular Model Evaluations and its AutoML Applications](https://arxiv.org/pdf/2311.02971.pdf) (*Under Review*, 2024)
+- [TabRepo: A Large Scale Repository of Tabular Model Evaluations and its AutoML Applications](https://arxiv.org/pdf/2311.02971.pdf) (*AutoML Conf*, 2024)
+- [AutoGluon-Multimodal (AutoMM): Supercharging Multimodal AutoML with Foundation Models](https://arxiv.org/pdf/2404.16233) (*AutoML Conf*, 2024) ([BibTeX](CITING.md#autogluonmultimodal))
+- [Multi-layer Stack Ensembles for Time Series Forecasting](https://arxiv.org/abs/2511.15350) (*AutoML Conf*, 2025) ([BibTeX](CITING.md#autogluontimeseries))
+- [Chronos-2: From Univariate to Universal Forecasting](https://arxiv.org/abs/2510.15821) (*Arxiv*, 2025) ([BibTeX](CITING.md#autogluontimeseries))
 
 ### Articles
 - [AutoGluon-TimeSeries: Every Time Series Forecasting Model In One Library](https://towardsdatascience.com/autogluon-timeseries-every-time-series-forecasting-model-in-one-library-29a3bf6879db) (*Towards Data Science*, Jan 2024)

{autogluon_tabular-1.4.1b20251116 → autogluon_tabular-1.4.1b20251217}/setup.py

@@ -57,7 +57,6 @@ extras_require = {
         "spacy<3.9",
         "torch",  # version range defined in `core/_setup_utils.py`
         "fastai>=2.3.1,<2.9",  # <{N+1} upper cap, where N is the latest released minor version
-        "blis>=0.7.0,<1.2.1;platform_system=='Windows' and python_version=='3.9'",  # blis not publishing Python 3.9 wheels for Windows, TODO: remove this after dropping Python 3.9 support
     ],
     "tabm": [
         "torch",  # version range defined in `core/_setup_utils.py`
@@ -93,28 +92,16 @@ extras_require = {
     ],
 }
 
-is_aarch64 = platform.machine() == "aarch64"
-is_darwin = sys.platform == "darwin"
-
-if is_darwin or is_aarch64:
-    # For macOS or aarch64, only use CPU version
-    extras_require["skl2onnx"] = [
-        "onnx>=1.13.0,<1.16.2;platform_system=='Windows'",  # cap at 1.16.1 for issue https://github.com/onnx/onnx/issues/6267
-        "onnx>=1.13.0,<1.18.0;platform_system!='Windows'",
-        "skl2onnx>=1.15.0,<1.18.0",
-        # For macOS, there isn't a onnxruntime-gpu package installed with skl2onnx.
-        # Therefore, we install onnxruntime explicitly here just for macOS.
-        "onnxruntime>=1.17.0,<1.20.0",
-    ]
-else:
-    # For other platforms, include both CPU and GPU versions
-    extras_require["skl2onnx"] = [
-        "onnx>=1.13.0,<1.16.2;platform_system=='Windows'",  # cap at 1.16.1 for issue https://github.com/onnx/onnx/issues/6267
-        "onnx>=1.13.0,<1.18.0;platform_system!='Windows'",
-        "skl2onnx>=1.15.0,<1.18.0",
-        "onnxruntime>=1.17.0,<1.20.0",  # install for gpu system due to https://github.com/autogluon/autogluon/issues/3804
-        "onnxruntime-gpu>=1.17.0,<1.20.0",
-    ]
+extras_require["skl2onnx"] = [
+    "skl2onnx>=1.15.0,<1.20.0",
+    # Sync ONNX requirements with multimodal/setup.py
+    "onnx>=1.13.0,!=1.16.2,<1.21.0;platform_system=='Windows'",  # exclude 1.16.2 for issue https://github.com/onnx/onnx/issues/6267
+    "onnx>=1.13.0,<1.21.0;platform_system!='Windows'",
+    # For macOS, there isn't a onnxruntime-gpu package installed with skl2onnx.
+    # Therefore, we install onnxruntime explicitly here just for macOS.
+    "onnxruntime>=1.17.0,<1.24.0",
+    "onnxruntime-gpu>=1.17.0,<1.24.0; platform_system != 'Darwin' and platform_machine != 'aarch64'",
+]
 
 # TODO: v1.0: Rename `all` to `core`, make `all` contain everything.
 all_requires = []

{autogluon_tabular-1.4.1b20251116 → autogluon_tabular-1.4.1b20251217}/src/autogluon/tabular/learner/default_learner.py

@@ -143,6 +143,7 @@ class DefaultLearner(AbstractTabularLearner):
             infer_limit=infer_limit,
             infer_limit_batch_size=infer_limit_batch_size,
             groups=groups,
+            label_cleaner=copy.deepcopy(self.label_cleaner),
             **trainer_fit_kwargs,
         )
         self.save_trainer(trainer=trainer)

{autogluon_tabular-1.4.1b20251116 → autogluon_tabular-1.4.1b20251217}/src/autogluon/tabular/models/__init__.py

@@ -1,5 +1,6 @@
 from autogluon.core.models.abstract.abstract_model import AbstractModel
 
+from .tabprep.prep_lgb_model import PrepLGBModel
 from .automm.automm_model import MultiModalPredictorModel
 from .automm.ft_transformer import FTTransformerModel
 from .catboost.catboost_model import CatBoostModel

{autogluon_tabular-1.4.1b20251116 → autogluon_tabular-1.4.1b20251217}/src/autogluon/tabular/models/catboost/catboost_model.py

@@ -146,7 +146,7 @@ class CatBoostModel(AbstractModel):
         num_cols_train = len(X.columns)
         num_classes = self.num_classes if self.num_classes else 1  # self.num_classes could be None after initialization if it's a regression problem
 
-        X = self.preprocess(X)
+        X = self.preprocess(X, y=y, is_train=True)
         cat_features = list(X.select_dtypes(include="category").columns)
         X = Pool(data=X, label=y, cat_features=cat_features, weight=sample_weight)
 

{autogluon_tabular-1.4.1b20251116 → autogluon_tabular-1.4.1b20251217}/src/autogluon/tabular/models/lgb/lgb_model.py

@@ -103,10 +103,46 @@ class LGBModel(AbstractModel):
             Scales linearly with the number of estimators, number of classes, and number of leaves.
             Memory usage peaks during model saving, with the peak consuming approximately 2-4x the size of the model in memory.
         """
+        data_mem_usage = get_approximate_df_mem_usage(X).sum()
+        return cls._estimate_memory_usage_common(
+            num_features=X.shape[1],
+            data_mem_usage=data_mem_usage,
+            hyperparameters=hyperparameters,
+            num_classes=num_classes,
+        )
+
+    @classmethod
+    def _estimate_memory_usage_static_lite(
+        cls,
+        num_samples: int,
+        num_features: int,
+        num_bytes_per_cell: float = 4,
+        hyperparameters: dict = None,
+        num_classes: int = 1,
+        **kwargs,
+    ) -> int:
+        data_mem_usage = num_samples * num_features * num_bytes_per_cell
+        return cls._estimate_memory_usage_common(
+            num_features=num_features,
+            data_mem_usage=data_mem_usage,
+            hyperparameters=hyperparameters,
+            num_classes=num_classes,
+        )
+
+    @classmethod
+    def _estimate_memory_usage_common(
+        cls,
+        num_features: int,
+        data_mem_usage: int | float,
+        hyperparameters: dict | None = None,
+        num_classes: int = 1,
+    ) -> int:
+        """
+        Utility method to avoid code duplication
+        """
         if hyperparameters is None:
             hyperparameters = {}
         num_classes = num_classes if num_classes else 1  # num_classes could be None after initialization if it's a regression problem
-        data_mem_usage = get_approximate_df_mem_usage(X).sum()
         data_mem_usage_bytes = data_mem_usage * 5 + data_mem_usage / 4 * num_classes  # TODO: Extremely crude approximation, can be vastly improved
 
         n_trees_per_estimator = num_classes if num_classes > 2 else 1
@@ -114,7 +150,7 @@ class LGBModel(AbstractModel):
         max_bins = hyperparameters.get("max_bins", 255)
         num_leaves = hyperparameters.get("num_leaves", 31)
         # Memory usage of histogram based on https://github.com/microsoft/LightGBM/issues/562#issuecomment-304524592
-        histogram_mem_usage_bytes = 20 * max_bins * len(X.columns) * num_leaves
+        histogram_mem_usage_bytes = 20 * max_bins * num_features * num_leaves
         histogram_mem_usage_bytes_max = hyperparameters.get("histogram_pool_size", None)
         if histogram_mem_usage_bytes_max is not None:
             histogram_mem_usage_bytes_max *= 1e6  # Convert megabytes to bytes, `histogram_pool_size` is in MB.
@@ -124,11 +160,11 @@ class LGBModel(AbstractModel):
 
         mem_size_per_estimator = n_trees_per_estimator * num_leaves * 100  # very rough estimate
         n_estimators = hyperparameters.get("num_boost_round", DEFAULT_NUM_BOOST_ROUND)
-        n_estimators_min = min(n_estimators, 1000)
-        mem_size_estimators = n_estimators_min * mem_size_per_estimator  # memory estimate after fitting up to 1000 estimators
+        n_estimators_min = min(n_estimators, 5000)
+        mem_size_estimators = n_estimators_min * mem_size_per_estimator  # memory estimate after fitting up to 5000 estimators
 
         approx_mem_size_req = data_mem_usage_bytes + histogram_mem_usage_bytes + mem_size_estimators
-        return approx_mem_size_req
+        return int(approx_mem_size_req)
 
     def _fit(self, X, y, X_val=None, y_val=None, time_limit=None, num_gpus=0, num_cpus=0, sample_weight=None, sample_weight_val=None, verbosity=2, **kwargs):
         try_import_lightgbm()  # raise helpful error message if LightGBM isn't installed
@@ -371,6 +407,9 @@ class LGBModel(AbstractModel):
         X = self.preprocess(X, **kwargs)
 
         y_pred_proba = self.model.predict(X, num_threads=num_cpus)
+        return self._post_process_predictions(y_pred_proba=y_pred_proba)
+
+    def _post_process_predictions(self, y_pred_proba) -> np.ndarray:
         if self.problem_type == QUANTILE:
             # y_pred_proba is a pd.DataFrame, need to convert
             y_pred_proba = y_pred_proba.to_numpy()
@@ -423,7 +462,7 @@ class LGBModel(AbstractModel):
         self,
         X: DataFrame,
         y: Series,
-        params,
+        params: dict,
         X_val=None,
         y_val=None,
         X_test=None,
@@ -432,11 +471,14 @@ class LGBModel(AbstractModel):
         sample_weight_val=None,
         sample_weight_test=None,
         save=False,
+        init_train=None,
+        init_val=None,
+        init_test=None,
     ):
         lgb_dataset_params_keys = ["two_round"]  # Keys that are specific to lightGBM Dataset object construction.
         data_params = {key: params[key] for key in lgb_dataset_params_keys if key in params}.copy()
 
-        X = self.preprocess(X, is_train=True)
+        X = self.preprocess(X, y=y, is_train=True)
         if X_val is not None:
             X_val = self.preprocess(X_val)
         if X_test is not None:
@@ -458,7 +500,13 @@ class LGBModel(AbstractModel):
 
         # X, W_train = self.convert_to_weight(X=X)
         dataset_train = construct_dataset(
-            x=X, y=y, location=os.path.join("self.path", "datasets", "train"), params=data_params, save=save, weight=sample_weight
+            x=X,
+            y=y,
+            location=os.path.join("self.path", "datasets", "train"),
+            params=data_params,
+            save=save,
+            weight=sample_weight,
+            init_score=init_train,
         )
         # dataset_train = construct_dataset_lowest_memory(X=X, y=y, location=self.path + 'datasets/train', params=data_params)
         if X_val is not None:
@@ -471,6 +519,7 @@ class LGBModel(AbstractModel):
                 params=data_params,
                 save=save,
                 weight=sample_weight_val,
+                init_score=init_val,
             )
             # dataset_val = construct_dataset_lowest_memory(X=X_val, y=y_val, location=self.path + 'datasets/val', reference=dataset_train, params=data_params)
         else:
@@ -485,6 +534,7 @@ class LGBModel(AbstractModel):
                 params=data_params,
                 save=save,
                 weight=sample_weight_test,
+                init_score=init_test,
             )
         else:
             dataset_test = None

{autogluon_tabular-1.4.1b20251116 → autogluon_tabular-1.4.1b20251217}/src/autogluon/tabular/models/lgb/lgb_utils.py

@@ -104,11 +104,11 @@ def softclass_lgbobj(preds, train_data):
     return grad.flatten("F"), hess.flatten("F")
 
 
-def construct_dataset(x: DataFrame, y: Series, location=None, reference=None, params=None, save=False, weight=None):
+def construct_dataset(x: DataFrame, y: Series, location=None, reference=None, params=None, save=False, weight=None, init_score=None):
     try_import_lightgbm()
     import lightgbm as lgb
 
-    dataset = lgb.Dataset(data=x, label=y, reference=reference, free_raw_data=True, params=params, weight=weight)
+    dataset = lgb.Dataset(data=x, label=y, reference=reference, free_raw_data=True, params=params, weight=weight, init_score=init_score)
 
     if save:
         assert location is not None

{autogluon_tabular-1.4.1b20251116 → autogluon_tabular-1.4.1b20251217}/src/autogluon/tabular/models/rf/rf_model.py

@@ -151,13 +151,13 @@ class RFModel(AbstractModel):
             hyperparameters = {}
         n_estimators_final = hyperparameters.get("n_estimators", 300)
         if isinstance(n_estimators_final, int):
-            n_estimators_minimum = min(40, n_estimators_final)
+            n_estimators = n_estimators_final
         else:  # if search space
-            n_estimators_minimum = 40
+            n_estimators = 40
         num_trees_per_estimator = cls._get_num_trees_per_estimator_static(problem_type=problem_type, num_classes=num_classes)
         bytes_per_estimator = num_trees_per_estimator * len(X) / 60000 * 1e6  # Underestimates by 3x on ExtraTrees
-        expected_min_memory_usage = int(bytes_per_estimator * n_estimators_minimum)
-        return expected_min_memory_usage
+        expected_memory_usage = int(bytes_per_estimator * n_estimators)
+        return expected_memory_usage
 
     def _validate_fit_memory_usage(self, mem_error_threshold: float = 0.5, mem_warning_threshold: float = 0.4, mem_size_threshold: int = 1e7, **kwargs):
         return super()._validate_fit_memory_usage(

{autogluon_tabular-1.4.1b20251116 → autogluon_tabular-1.4.1b20251217}/src/autogluon/tabular/models/tabm/tabm_model.py

@@ -49,6 +49,7 @@ class TabMModel(AbstractModel):
         self._indicator_columns = None
         self._features_bool = None
         self._bool_to_cat = None
+        self.device = None
 
     def _fit(
         self,
@@ -87,7 +88,7 @@ class TabMModel(AbstractModel):
         if X_val is None:
             from autogluon.core.utils import generate_train_test_split
 
-            X_train, X_val, y_train, y_val = generate_train_test_split(
+            X, X_val, y, y_val = generate_train_test_split(
                 X=X,
                 y=y,
                 problem_type=self.problem_type,
@@ -98,7 +99,7 @@ class TabMModel(AbstractModel):
         hyp = self._get_model_params()
         bool_to_cat = hyp.pop("bool_to_cat", True)
 
-        X = self.preprocess(X, is_train=True, bool_to_cat=bool_to_cat)
+        X = self.preprocess(X, y=y, is_train=True, bool_to_cat=bool_to_cat)
         if X_val is not None:
             X_val = self.preprocess(X_val)
 
@@ -142,6 +143,81 @@ class TabMModel(AbstractModel):
 
         return X
 
+    def save(self, path: str = None, verbose=True) -> str:
+        """
+        Need to set device to CPU to be able to load on a non-GPU environment
+        """
+        import torch
+
+        # Save on CPU to ensure the model can be loaded without GPU
+        if self.model is not None:
+            self.device = self.model.device_
+            device_cpu = torch.device("cpu")
+            self.model.model_ = self.model.model_.to(device_cpu)
+            self.model.device_ = device_cpu
+        path = super().save(path=path, verbose=verbose)
+        # Put the model back to the device after the save
+        if self.model is not None:
+            self.model.model_.to(self.device)
+            self.model.device_ = self.device
+
+        return path
+
+    @classmethod
+    def load(cls, path: str, reset_paths=True, verbose=True):
+        """
+        Loads the model from disk to memory.
+        The loaded model will be on the same device it was trained on (cuda/mps);
+        if the device is not available (trained on GPU, deployed on CPU), then `cpu` will be used.
+
+        Parameters
+        ----------
+        path : str
+            Path to the saved model, minus the file name.
+            This should generally be a directory path ending with a '/' character (or appropriate path separator value depending on OS).
+            The model file is typically located in os.path.join(path, cls.model_file_name).
+        reset_paths : bool, default True
+            Whether to reset the self.path value of the loaded model to be equal to path.
+            It is highly recommended to keep this value as True unless accessing the original self.path value is important.
+            If False, the actual valid path and self.path may differ, leading to strange behaviour and potential exceptions if the model needs to load any other files at a later time.
+        verbose : bool, default True
+            Whether to log the location of the loaded file.
+
+        Returns
+        -------
+        model : cls
+            Loaded model object.
+        """
+        import torch
+
+        model: TabMModel = super().load(path=path, reset_paths=reset_paths, verbose=verbose)
+
+        # Put the model on the same device it was trained on (GPU/MPS) if it is available; otherwise use CPU
+        if model.model is not None:
+            original_device_type = model.device.type
+            if "cuda" in original_device_type:
+                # cuda: nvidia GPU
+                device = torch.device(original_device_type if torch.cuda.is_available() else "cpu")
+            elif "mps" in original_device_type:
+                # mps: Apple Silicon
+                device = torch.device(original_device_type if torch.backends.mps.is_available() else "cpu")
+            else:
+                device = torch.device(original_device_type)
+
+            if verbose and (original_device_type != device.type):
+                logger.log(15, f"Model is trained on {original_device_type}, but the device is not available - loading on {device.type}")
+
+            model.set_device(device=device)
+
+        return model
+
+    def set_device(self, device):
+        self.device = device
+        if self.model is not None:
+            self.model.device_ = device
+            if self.model.model_ is not None:
+                self.model.model_ = self.model.model_.to(device)
+
     @classmethod
     def supported_problem_types(cls) -> list[str] | None:
         return ["binary", "multiclass", "regression"]

autogluon_tabular-1.4.1b20251217/src/autogluon/tabular/models/tabprep/prep_lgb_model.py

@@ -0,0 +1,21 @@
+from __future__ import annotations
+
+from ..lgb.lgb_model import LGBModel
+from .prep_mixin import ModelAgnosticPrepMixin
+
+
+class PrepLGBModel(ModelAgnosticPrepMixin, LGBModel):
+    ag_key = "GBM_PREP"
+    ag_name = "LightGBMPrep"
+
+    @classmethod
+    def _estimate_memory_usage_static(cls, **kwargs) -> int:
+        memory_usage = super()._estimate_memory_usage_static(**kwargs)
+        # FIXME: 1.5 runs OOM on kddcup09_appetency fold 2 repeat 0 prep_LightGBM_r49_BAG_L1
+        return memory_usage * 2.0  # FIXME: For some reason this underestimates mem usage without this
+
+    @classmethod
+    def _estimate_memory_usage_static_lite(cls, **kwargs) -> int:
+        memory_usage = super()._estimate_memory_usage_static_lite(**kwargs)
+        # FIXME: 1.5 runs OOM on kddcup09_appetency fold 2 repeat 0 prep_LightGBM_r49_BAG_L1
+        return memory_usage * 2.0  # FIXME: For some reason this underestimates mem usage without this