autogluon.multimodal 1.4.1b20251119__tar.gz → 1.5.1b20260112__tar.gz

{autogluon_multimodal-1.4.1b20251119/src/autogluon.multimodal.egg-info → autogluon_multimodal-1.5.1b20260112}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: autogluon.multimodal
-Version: 1.4.1b20251119
+Version: 1.5.1b20260112
 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
@@ -42,9 +42,9 @@ Requires-Dist: scikit-learn<1.8.0,>=1.4.0
 Requires-Dist: Pillow<12,>=10.0.1
 Requires-Dist: tqdm<5,>=4.38
 Requires-Dist: boto3<2,>=1.10
-Requires-Dist: torch<2.8,>=2.6
-Requires-Dist: lightning<2.8,>=2.5.1
-Requires-Dist: transformers[sentencepiece]<4.50,>=4.38.0
+Requires-Dist: torch<2.10,>=2.6
+Requires-Dist: lightning<2.6,>=2.5.1
+Requires-Dist: transformers[sentencepiece]<4.58,>=4.51.0
 Requires-Dist: accelerate<2.0,>=0.34.0
 Requires-Dist: fsspec[http]<=2025.3
 Requires-Dist: requests<3,>=2.30
@@ -52,14 +52,14 @@ Requires-Dist: jsonschema<4.24,>=4.18
 Requires-Dist: seqeval<1.3.0,>=1.2.2
 Requires-Dist: evaluate<0.5.0,>=0.4.0
 Requires-Dist: timm<1.0.7,>=0.9.5
-Requires-Dist: torchvision<0.23.0,>=0.21.0
+Requires-Dist: torchvision<0.25.0,>=0.21.0
 Requires-Dist: scikit-image<0.26.0,>=0.19.1
 Requires-Dist: text-unidecode<1.4,>=1.3
 Requires-Dist: torchmetrics<1.8,>=1.2.0
 Requires-Dist: omegaconf<2.4.0,>=2.1.1
-Requires-Dist: autogluon.core[raytune]==1.4.1b20251119
-Requires-Dist: autogluon.features==1.4.1b20251119
-Requires-Dist: autogluon.common==1.4.1b20251119
+Requires-Dist: autogluon.core[raytune]==1.5.1b20260112
+Requires-Dist: autogluon.features==1.5.1b20260112
+Requires-Dist: autogluon.common==1.5.1b20260112
 Requires-Dist: pytorch-metric-learning<2.9,>=1.3.0
 Requires-Dist: nlpaug<1.2.0,>=1.1.10
 Requires-Dist: nltk<3.10,>=3.4.5
@@ -73,11 +73,11 @@ Requires-Dist: pdf2image<1.19,>=1.17.0
 Provides-Extra: tests
 Requires-Dist: ruff; extra == "tests"
 Requires-Dist: datasets<3.6.0,>=2.16.0; 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: onnxruntime<1.22.0,>=1.17.0; extra == "tests"
-Requires-Dist: onnxruntime-gpu<1.22.0,>=1.17.0; (platform_system != "Darwin" and platform_machine != "aarch64") and extra == "tests"
 Requires-Dist: tensorrt<10.9.1,>=8.6.0; (platform_system == "Linux" and python_version < "3.11") and 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
@@ -100,7 +100,7 @@ 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)
@@ -117,7 +117,7 @@ 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:
 
@@ -164,7 +164,10 @@ 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_multimodal-1.4.1b20251119 → autogluon_multimodal-1.5.1b20260112}/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:
 
@@ -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_multimodal-1.4.1b20251119 → autogluon_multimodal-1.5.1b20260112}/setup.py

@@ -41,7 +41,7 @@ install_requires = [
     "seqeval>=1.2.2,<1.3.0",
     "evaluate>=0.4.0,<0.5.0",
     "timm>=0.9.5,<1.0.7",
-    "torchvision>=0.21.0,<0.23.0",
+    "torchvision>=0.21.0,<0.25.0",
     "scikit-image>=0.19.1,<0.26.0",
     "text-unidecode>=1.3,<1.4",
     "torchmetrics>=1.2.0,<1.8",
@@ -66,11 +66,14 @@ install_requires = ag.get_dependency_version_ranges(install_requires)
 tests_require = [
     "ruff",
     "datasets>=2.16.0,<3.6.0",
-    "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'",
-    "onnxruntime>=1.17.0,<1.22.0",  # install for gpu system due to https://github.com/autogluon/autogluon/issues/3804
-    "onnxruntime-gpu>=1.17.0,<1.22.0;platform_system!='Darwin' and platform_machine!='aarch64'",
     "tensorrt>=8.6.0,<10.9.1;platform_system=='Linux' and python_version<'3.11'",
+    # Sync ONNX requirements with tabular/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'",
 ]
 
 extras_require = {"tests": tests_require}

{autogluon_multimodal-1.4.1b20251119 → autogluon_multimodal-1.5.1b20260112}/src/autogluon/multimodal/data/__init__.py

@@ -3,14 +3,14 @@ from .dataset import BaseDataset
 from .dataset_mmlab import MultiImageMixDataset
 from .infer_types import (
     infer_column_types,
+    infer_ner_column_type,
     infer_output_shape,
     infer_problem_type,
     infer_rois_column_type,
     is_image_column,
 )
-from .mixup import MixupModule
-from .infer_types import infer_column_types, infer_output_shape, infer_problem_type, is_image_column, infer_ner_column_type
 from .label_encoder import CustomLabelEncoder, NerLabelEncoder
+from .mixup import MixupModule
 from .preprocess_dataframe import MultiModalFeaturePreprocessor
 from .process_categorical import CategoricalProcessor
 from .process_document import DocumentProcessor

{autogluon_multimodal-1.4.1b20251119 → autogluon_multimodal-1.5.1b20260112}/src/autogluon/multimodal/data/dataset_mmlab/multi_image_mix_dataset.py

@@ -290,7 +290,7 @@ class Mosaic(BaseTransform):
         prob: float = 1.0,
     ) -> None:
         assert isinstance(img_scale, tuple)
-        assert 0 <= prob <= 1.0, "The probability should be in range [0,1]. " f"got {prob}."
+        assert 0 <= prob <= 1.0, f"The probability should be in range [0,1]. got {prob}."
 
         log_img_scale(img_scale, skip_square=True, shape_order="wh")
         self.img_scale = img_scale

{autogluon_multimodal-1.4.1b20251119 → autogluon_multimodal-1.5.1b20260112}/src/autogluon/multimodal/data/infer_types.py

@@ -304,11 +304,24 @@ def is_document_image_column(
     col_name: str,
     image_type: Optional[str] = IMAGE_PATH,
     sample_m: Optional[int] = 10,
-    text_len_threshold: Optional[int] = 100,
+    min_text_len_threshold: Optional[int] = 200,
+    text_density_threshold: Optional[float] = 0.001,
+    min_line_count: Optional[int] = 3,
+    min_document_ratio: Optional[float] = 0.8,
 ) -> bool:
     """
     Identify if a column is a document image column.
 
+    Document images are images that primarily contain text (e.g., scanned documents,
+    screenshots of text, PDFs rendered as images). Regular photographs, maps,
+    charts with labels, or images with watermarks/captions should NOT be
+    classified as document images.
+
+    The detection uses multiple heuristics:
+    1. Minimum absolute text length (short text like watermarks is ignored)
+    2. Text density relative to image size (documents have high text-to-pixel ratio)
+    3. Line count (documents typically have multiple lines of text)
+
     Parameters
     ----------
     data
@@ -319,46 +332,90 @@ def is_document_image_column(
         The image type to check. Set to IMAGE_PATH by default.
     sample_m
         Number of sample images used to check if images are documents images.
-    text_len_threshold
-        If the average text length is longer than text_len_threshold, the images will be considered as document images.
+    min_text_len_threshold
+        Minimum text length to even consider an image as a potential document.
+        This filters out images with just watermarks or short captions.
+    text_density_threshold
+        Minimum ratio of (text_characters / image_pixels) to consider as document.
+        Documents typically have much higher text density than photos with labels.
+    min_line_count
+        Minimum number of non-empty text lines expected in a document.
+    min_document_ratio
+        Minimum ratio of images that must be classified as documents for the
+        entire column to be treated as a document column.
+
     Returns
     -------
     Whether the column is a document image column.
     """
+    if data.empty:
+        return False
 
-    # TODO: Add support for other types (e.g., pdf) of document.
-
-    words_len = []
     if len(data) > sample_m:
-        # Sample to speed-up type inference
         data = data.sample(n=sample_m, random_state=0)
-    failure_count = 0
+
+    document_count = 0
+    total_processed = 0
+
     for images in data:
-        success = False
+        if images is None:
+            continue
+
         if not isinstance(images, list):
             images = [images]
+
         for per_image in images:
+            if not isinstance(per_image, str):
+                total_processed += 1
+                continue
+
             try:
-                # convert images to string
-                with PIL.Image.open(per_image) as doc_image:
-                    words = pytesseract.image_to_string(doc_image)
-                    words_len.append(len(words))
+                with PIL.Image.open(per_image) as img:
+                    width, height = img.size
+                    total_pixels = width * height
+
+                    ocr_text = pytesseract.image_to_string(img)
+                    text_length = len(ocr_text.strip())
+
+                    total_processed += 1
+
+                    # Heuristic 1: Minimum absolute text length
+                    # Filters out watermarks, copyright notices, short captions
+                    if text_length < min_text_len_threshold:
+                        continue
+
+                    # Heuristic 2: Text density (characters per pixel)
+                    # Documents have dense text; photos with small labels don't
+                    text_density = text_length / total_pixels
+                    if text_density < text_density_threshold:
+                        continue
+
+                    # Heuristic 3: Line count
+                    # Documents have multiple lines; watermarks are 1-2 lines
+                    lines = [line for line in ocr_text.split("\n") if line.strip()]
+                    if len(lines) < min_line_count:
+                        continue
+
+                    # Passed all heuristics - this looks like a document
+                    document_count += 1
+
             except Exception as e:
-                words_len.append(0)
-                success = False
-                break
-            success = True
-        if not success:
-            failure_count += 1
+                logger.debug(f"Failed to process image {per_image}: {e}")
+                total_processed += 1
 
-    if (1 - failure_count / sample_m) >= 0.8:
-        logger.debug(f"Average length of words of this dataset is {sum(words_len) / len(words_len)}.")
-        if sum(words_len) / len(words_len) > text_len_threshold:
-            return True
-        else:
-            return False
-    else:
-        False
+    if total_processed == 0:
+        return False
+
+    document_ratio = document_count / total_processed
+    is_document_column = document_ratio >= min_document_ratio
+
+    logger.debug(
+        f"Column '{col_name}': {document_count}/{total_processed} images "
+        f"({document_ratio:.1%}) classified as documents. "
+        f"Column type: {'document' if is_document_column else 'regular'} images."
+    )
+
+    return is_document_column
 
 
 def is_text_column(data: pd.Series) -> bool:
@@ -769,8 +826,7 @@ def infer_output_shape(
     if problem_type in [BINARY, MULTICLASS, REGRESSION, CLASSIFICATION]:
         class_num = len(data[label_column].unique())
         err_msg = (
-            f"Problem type is '{problem_type}' while the number of "
-            f"unique values in the label column is {class_num}."
+            f"Problem type is '{problem_type}' while the number of unique values in the label column is {class_num}."
         )
         if problem_type == BINARY:
             if class_num != 2:

{autogluon_multimodal-1.4.1b20251119 → autogluon_multimodal-1.5.1b20260112}/src/autogluon/multimodal/data/preprocess_dataframe.py

@@ -456,9 +456,9 @@ class MultiModalFeaturePreprocessor(TransformerMixin, BaseEstimator):
         text_types
             The column types of these text data, e.g., text or text_identifier.
         """
-        assert (
-            self._fit_called or self._fit_x_called
-        ), "You will need to first call preprocessor.fit_x() before calling preprocessor.transform_text."
+        assert self._fit_called or self._fit_x_called, (
+            "You will need to first call preprocessor.fit_x() before calling preprocessor.transform_text."
+        )
         text_features = {}
         text_types = {}
         for col_name in self._text_feature_names:
@@ -508,9 +508,9 @@ class MultiModalFeaturePreprocessor(TransformerMixin, BaseEstimator):
         image_types
             The column types of these image data, e.g., image_path or image_identifier.
         """
-        assert (
-            self._fit_called or self._fit_x_called
-        ), "You will need to first call preprocessor.fit_x() before calling preprocessor.transform_rois."
+        assert self._fit_called or self._fit_x_called, (
+            "You will need to first call preprocessor.fit_x() before calling preprocessor.transform_rois."
+        )
 
         x = self.transform_image(df)
         ret_data = x[0]
@@ -552,9 +552,9 @@ class MultiModalFeaturePreprocessor(TransformerMixin, BaseEstimator):
         image_types
             The column types of these image data, e.g., image_path or image_identifier.
         """
-        assert (
-            self._fit_called or self._fit_x_called
-        ), "You will need to first call preprocessor.fit_x() before calling preprocessor.transform_semantic_segmentation_img."
+        assert self._fit_called or self._fit_x_called, (
+            "You will need to first call preprocessor.fit_x() before calling preprocessor.transform_semantic_segmentation_img."
+        )
 
         ret_data = {}
         ret_type = {}
@@ -597,9 +597,9 @@ class MultiModalFeaturePreprocessor(TransformerMixin, BaseEstimator):
             All the image data stored in a dictionary.
         image_types
             The column types of these image data, e.g., image_path, image_bytearray or image_identifier."""
-        assert (
-            self._fit_called or self._fit_x_called
-        ), "You will need to first call preprocessor.fit_x() before calling preprocessor.transform_image."
+        assert self._fit_called or self._fit_x_called, (
+            "You will need to first call preprocessor.fit_x() before calling preprocessor.transform_image."
+        )
 
         image_features = {}
         image_types = {}
@@ -650,9 +650,9 @@ class MultiModalFeaturePreprocessor(TransformerMixin, BaseEstimator):
         document_types
             The column types of these document data.
         """
-        assert (
-            self._fit_called or self._fit_x_called
-        ), "You will need to first call preprocessor.fit_x() before calling preprocessor.transform_document."
+        assert self._fit_called or self._fit_x_called, (
+            "You will need to first call preprocessor.fit_x() before calling preprocessor.transform_document."
+        )
         document_features = {}
         document_types = {}
         for col_name in self._document_feature_names:
@@ -687,9 +687,9 @@ class MultiModalFeaturePreprocessor(TransformerMixin, BaseEstimator):
         None
             The column types of numerical data, which is None currently since only one numerical type exists.
         """
-        assert (
-            self._fit_called or self._fit_x_called
-        ), "You will need to first call preprocessor.fit before calling preprocessor.transform_numerical."
+        assert self._fit_called or self._fit_x_called, (
+            "You will need to first call preprocessor.fit before calling preprocessor.transform_numerical."
+        )
         numerical_features = {}
         for col_name in self._numerical_feature_names:
             generator = self._feature_generators[col_name]
@@ -720,9 +720,9 @@ class MultiModalFeaturePreprocessor(TransformerMixin, BaseEstimator):
         None
             The column types of categorical data, which is None currently since only one categorical type exists.
         """
-        assert (
-            self._fit_called or self._fit_x_called
-        ), "You will need to first call preprocessor.fit before calling preprocessor.transform_categorical."
+        assert self._fit_called or self._fit_x_called, (
+            "You will need to first call preprocessor.fit before calling preprocessor.transform_categorical."
+        )
         categorical_features = {}
         for col_name, num_category in self._categorical_num_categories.items():
             col_value = df[col_name]
@@ -758,9 +758,9 @@ class MultiModalFeaturePreprocessor(TransformerMixin, BaseEstimator):
         label_types
             The label column types.
         """
-        assert (
-            self._fit_called or self._fit_y_called
-        ), "You will need to first call preprocessor.fit_y() before calling preprocessor.transform_label."
+        assert self._fit_called or self._fit_y_called, (
+            "You will need to first call preprocessor.fit_y() before calling preprocessor.transform_label."
+        )
         # Creating deep copy of the DataFrame, which allows writable buffer to be created for the new df
         # This is needed for 1.4.1 < scikit-learn < 1.5.0, versions <=1.4.0 and >=1.5.1 do not need a writable buffer
         df = df.copy(deep=True)
@@ -784,9 +784,9 @@ class MultiModalFeaturePreprocessor(TransformerMixin, BaseEstimator):
         self,
         df: pd.DataFrame,
     ) -> Tuple[Dict[str, NDArray], Dict[str, str]]:
-        assert (
-            self._fit_called or self._fit_x_called
-        ), "You will need to first call preprocessor.fit_x() before calling preprocessor.transform_ner."
+        assert self._fit_called or self._fit_x_called, (
+            "You will need to first call preprocessor.fit_x() before calling preprocessor.transform_ner."
+        )
         ret_data, ret_type = {}, {}
         ner_text_features = {}
         ner_text_types = {}
@@ -831,12 +831,12 @@ class MultiModalFeaturePreprocessor(TransformerMixin, BaseEstimator):
         -------
         Ground-truth labels ready to compute metric scores.
         """
-        assert (
-            self._fit_called or self._fit_y_called
-        ), "You will need to first call preprocessor.fit_y() before calling preprocessor.transform_label_for_metric."
-        assert (
-            self._label_column in df.columns
-        ), f"Label {self._label_column} is not in the data. Cannot perform evaluation without ground truth labels."
+        assert self._fit_called or self._fit_y_called, (
+            "You will need to first call preprocessor.fit_y() before calling preprocessor.transform_label_for_metric."
+        )
+        assert self._label_column in df.columns, (
+            f"Label {self._label_column} is not in the data. Cannot perform evaluation without ground truth labels."
+        )
         y_df = df[self._label_column]
         if self.label_type == CATEGORICAL:
             # need to encode to integer labels
@@ -875,9 +875,9 @@ class MultiModalFeaturePreprocessor(TransformerMixin, BaseEstimator):
         -------
         Predicted labels ready to compute metric scores.
         """
-        assert (
-            self._fit_called or self._fit_y_called
-        ), "You will need to first call preprocessor.fit_y() before calling preprocessor.transform_prediction."
+        assert self._fit_called or self._fit_y_called, (
+            "You will need to first call preprocessor.fit_y() before calling preprocessor.transform_prediction."
+        )
 
         if self.label_type == CATEGORICAL:
             assert len(y_pred.shape) <= 2

{autogluon_multimodal-1.4.1b20251119 → autogluon_multimodal-1.5.1b20260112}/src/autogluon/multimodal/data/template_engine.py

@@ -32,9 +32,9 @@ class TemplateEngine:
         self.template_length = self.template_config.template_length
 
         if self.preset_templates:
-            assert (
-                len(self.preset_templates) == 2
-            ), f"Preset templates has the wrong format. Needs to be [DATASET, SUBSET]."
+            assert len(self.preset_templates) == 2, (
+                f"Preset templates has the wrong format. Needs to be [DATASET, SUBSET]."
+            )
             dataset_templates = DatasetTemplates(self.preset_templates[0], self.preset_templates[1])
             current_templates = list(dataset_templates.templates.values())
             self.templates += current_templates[: self.num_templates]

{autogluon_multimodal-1.4.1b20251119 → autogluon_multimodal-1.5.1b20260112}/src/autogluon/multimodal/data/trivial_augmenter.py

@@ -210,8 +210,8 @@ def set_image_augmentation_space():
 def download_nltk():
     """
     Download required NLTK resources with singleton pattern to prevent multiple downloads.
-    
-    This function handles NLTK 3.9+ changes where resource names changed and 
+
+    This function handles NLTK 3.9+ changes where resource names changed and
     the quiet=True parameter behavior was affected. Uses a global flag to ensure
     downloads happen only once even when TrivialAugment is instantiated multiple times.
     """
@@ -232,7 +232,6 @@ def download_nltk():
         try:
             nltk.data.find(resource_path)
         except LookupError:
-
             nltk.download(download_name, quiet=True)
     _nltk_downloaded = True
 

{autogluon_multimodal-1.4.1b20251119 → autogluon_multimodal-1.5.1b20260112}/src/autogluon/multimodal/learners/base.py

@@ -530,9 +530,9 @@ class BaseLearner(ExportMixin, DistillationMixin, RealtimeMixin):
     def fit_sanity_check(self):
         assert not self._resume or not self._is_hpo, "You can not resume training with HPO."
         if self._is_hpo and hasattr(self, "_teacher_learner") and self._teacher_learner is not None:
-            assert isinstance(
-                self._teacher_learner, str
-            ), "HPO with distillation only supports passing a path to the learner."
+            assert isinstance(self._teacher_learner, str), (
+                "HPO with distillation only supports passing a path to the learner."
+            )
 
     def prepare_fit_args(
         self,
@@ -683,9 +683,9 @@ class BaseLearner(ExportMixin, DistillationMixin, RealtimeMixin):
                 overrides=hyperparameters,
             )
         if self._model is None:
-            assert (
-                len(self._config.model.names) == 1
-            ), f"Zero shot mode only supports using one model, but detects multiple models {self._config.model.names}"
+            assert len(self._config.model.names) == 1, (
+                f"Zero shot mode only supports using one model, but detects multiple models {self._config.model.names}"
+            )
             self._model = create_fusion_model(
                 config=self._config,
                 pretrained=self._pretrained,
@@ -836,8 +836,7 @@ class BaseLearner(ExportMixin, DistillationMixin, RealtimeMixin):
         )
         if mixup_active and (config.env.per_gpu_batch_size == 1 or config.env.per_gpu_batch_size % 2 == 1):
             warnings.warn(
-                "The mixup is done on the batch."
-                "The per_gpu_batch_size should be >1 and even for reasonable operation",
+                "The mixup is done on the batch.The per_gpu_batch_size should be >1 and even for reasonable operation",
                 UserWarning,
             )
         return mixup_active, mixup_func
@@ -1053,9 +1052,9 @@ class BaseLearner(ExportMixin, DistillationMixin, RealtimeMixin):
         if (
             config.env.strategy == DEEPSPEED_OFFLOADING and num_gpus == 1 and DEEPSPEED_MODULE not in sys.modules
         ):  # Offloading currently only tested for single GPU
-            assert (
-                version.parse(pl.__version__) >= version.parse(DEEPSPEED_MIN_PL_VERSION)
-            ), f"For DeepSpeed Offloading to work reliably you need at least lightning version {DEEPSPEED_MIN_PL_VERSION}, however, found {pl.__version__}. Please update your lightning version."
+            assert version.parse(pl.__version__) >= version.parse(DEEPSPEED_MIN_PL_VERSION), (
+                f"For DeepSpeed Offloading to work reliably you need at least lightning version {DEEPSPEED_MIN_PL_VERSION}, however, found {pl.__version__}. Please update your lightning version."
+            )
             from ..optim.deepspeed import CustomDeepSpeedStrategy
 
             strategy = CustomDeepSpeedStrategy(
@@ -1909,15 +1908,15 @@ class BaseLearner(ExportMixin, DistillationMixin, RealtimeMixin):
         return_prob: Optional[bool] = False,
     ):
         query_embeddings = self.extract_embedding(query_data, as_tensor=True)
-        assert (
-            len(query_embeddings) == 1
-        ), f"Multiple embedding types `{query_embeddings.keys()}` exist in query data. Please reduce them to one type."
+        assert len(query_embeddings) == 1, (
+            f"Multiple embedding types `{query_embeddings.keys()}` exist in query data. Please reduce them to one type."
+        )
         query_embeddings = list(query_embeddings.values())[0]
 
         candidate_embeddings = self.extract_embedding(candidate_data, as_tensor=True)
-        assert (
-            len(candidate_embeddings) == 1
-        ), f"Multiple embedding types `{candidate_embeddings.keys()}` exist in candidate data. Please reduce them to one type."
+        assert len(candidate_embeddings) == 1, (
+            f"Multiple embedding types `{candidate_embeddings.keys()}` exist in candidate data. Please reduce them to one type."
+        )
         candidate_embeddings = list(candidate_embeddings.values())[0]
 
         if return_prob:
@@ -2157,9 +2156,9 @@ class BaseLearner(ExportMixin, DistillationMixin, RealtimeMixin):
         state_dict = {k: v for k, v in state_dict.items() if k not in buffer_names_to_filter}
 
         load_result = self._model.load_state_dict(state_dict, strict=strict)
-        assert (
-            len(load_result.unexpected_keys) == 0
-        ), f"Load model failed, unexpected keys {load_result.unexpected_keys.__str__()}"
+        assert len(load_result.unexpected_keys) == 0, (
+            f"Load model failed, unexpected keys {load_result.unexpected_keys.__str__()}"
+        )
 
     @staticmethod
     def _replace_model_name_prefix(