autoencodix 0.2.0__tar.gz → 0.2.2__tar.gz

{autoencodix-0.2.0 → autoencodix-0.2.2}/.gitignore

@@ -1,5 +1,5 @@
 uv.lock
-data/
+data/raw/*
 notebooks/census_chunks/
 notebooks/*/*
 results/*

{autoencodix-0.2.0 → autoencodix-0.2.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: autoencodix
-Version: 0.2.0
+Version: 0.2.2
 Summary: Framework for multi-omics data integration by autoencoders.
 Project-URL: Homepage, https://github.com/jan-forest/autoencodix_package
 Project-URL: Issues, https://github.com/jan-forest/autoencodix_package/issues
@@ -30,6 +30,7 @@ Requires-Dist: pandas-stubs>=2.2.2.240807
 Requires-Dist: pandas>=2.2.3
 Requires-Dist: pyarrow>=19.0.1
 Requires-Dist: pydantic>=2.10.6
+Requires-Dist: python-dotenv>=1.2.1
 Requires-Dist: pyyaml>=6.0.2
 Requires-Dist: scanpy>=1.10.3
 Requires-Dist: scikit-learn>=1.6.1

{autoencodix-0.2.0 → autoencodix-0.2.2}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "autoencodix"
-version = "0.2.0"
+version = "0.2.2"
 authors = [
   { name="Maximilian Joas & Jan Ewald", email="jan.ewald@uni-leipzig.de" },
 ]
@@ -44,6 +44,7 @@ dependencies = [
     "openai==2.31.0",
     "tensorboard-plugin-profile==2.20.1",
     "tensorboard>=2.20.0",
+    "python-dotenv>=1.2.1",
 ]
 
 [dependency-groups]

autoencodix-0.2.2/src/autoencodix/data/__init__.py

@@ -0,0 +1,39 @@
+from ._datapackage_splitter import DataPackageSplitter
+from ._datasetcontainer import DatasetContainer
+from ._datasplitter import DataSplitter
+from ._filter import DataFilter
+from ._imgdataclass import ImgData
+from ._nanremover import NaNRemover
+from ._numeric_dataset import NumericDataset, TensorAwareDataset
+from ._stackix_dataset import StackixDataset
+from ._stackix_preprocessor import StackixPreprocessor
+from .datapackage import DataPackage
+from .general_preprocessor import GeneralPreprocessor
+from ._sc_filter import SingleCellFilter
+from ._xmodal_preprocessor import XModalPreprocessor
+from ._image_dataset import ImageDataset
+from ._multimodal_dataset import MultiModalDataset
+from ._image_processor import ImagePreprocessor
+from ._sampler import BalancedBatchSampler
+
+
+__all__ = [
+    "DataSplitter",
+    "NumericDataset",
+    "DatasetContainer",
+    "GeneralPreprocessor",
+    "ImgData",
+    "DataPackage",
+    "DataPackageSplitter",
+    "DataFilter",
+    "NaNRemover",
+    "StackixPreprocessor",
+    "StackixDataset",
+    "SingleCellFilter",
+    "XModalPreprocessor",
+    "ImageDataset",
+    "TensorAwareDataset",
+    "MultiModalDataset",
+    "ImagePreprocessor",
+    "BalancedBatchSampler",
+]

autoencodix-0.2.2/src/autoencodix/data/_datapackage_splitter.py

@@ -0,0 +1,151 @@
+import copy
+from typing import Any, Dict, Optional
+
+import numpy as np
+import pandas as pd
+from anndata import AnnData  # type: ignore
+from mudata import MuData  # type: ignore
+
+from autoencodix.data.datapackage import DataPackage
+from autoencodix.configs.default_config import DefaultConfig
+
+
+class DataPackageSplitter:
+    """Splits DataPackage objects into training, validation, and testing sets.
+
+    Supports paired and unpaired (translation) splitting.
+
+    Attributes:
+        data_package: The original DataPackage to split.
+        config: The configuration settings for the splitting process.
+        indices: The indices for each split (train/val/test).
+    """
+
+    def __init__(
+        self,
+        data_package: DataPackage,
+        config: DefaultConfig,
+        indices: Dict[str, Dict[str, Dict[str, np.ndarray]]],
+    ) -> None:
+        self._data_package = data_package
+        self.indices = indices
+        self.config = config
+
+        if not isinstance(self._data_package, DataPackage):
+            raise TypeError(
+                f"Expected data_package to be of type DataPackage, got {type(self._data_package)}"
+            )
+
+    def _shallow_copy(self, value: Any) -> Any:
+        try:
+            return copy.copy(value)
+        except AttributeError:
+            return value
+
+    def _indexing(self, obj: Any, indices: np.ndarray) -> Any:
+        """Indexes pd.DataFrame, list, AnnData, or MuData objects using the provided indices.
+
+        Args:
+            obj: The object to index (can be pd.DataFrame, list, AnnData, MuData, or None).
+            indices: The indices to use for indexing.
+        Returns:
+            The indexed object, or None if the input object is None.
+        Raises:
+            TypeError: If an unsupported type is encountered.
+        """
+
+        if obj is None:
+            return None
+        if isinstance(obj, pd.DataFrame):
+            return obj.iloc[indices]
+        elif isinstance(obj, list):
+            return [obj[i] for i in indices]
+        elif isinstance(obj, (AnnData, MuData)):
+            # print(f"shape of obj: {obj.shape}")
+            # print(f"obj: {obj}")
+            # print(f"len(ind): {len(indices)}")
+            # print(f"max of index{np.max(indices)}")
+            # print(f"ind: {indices}")
+            return obj[indices]
+        else:
+            raise TypeError(
+                f"Unsupported type for indexing: {type(obj)}. "
+                "Supported types are pd.DataFrame, list, AnnData, and MuData."
+            )
+
+    def _split_data_package(self, split: str) -> Optional[DataPackage]:
+        """Creates a new DataPackage where each attribute is indexed (if applicable)
+        by the given indices. Returns None if indices are empty.
+
+        Args:
+            indices: The indices to use for splitting the DataPackage.
+        Returns:
+            A new DataPackage with attributes indexed by the provided indices,
+            or None if indices are empty.
+        """
+        if len(self.indices) == 0:
+            return None
+
+        split_data = {}
+        for key, value in self._data_package.__dict__.items():
+            if value is None:
+                continue
+            split_data[key] = {
+                modality: self._indexing(data, self.indices[key][modality][split])
+                for modality, data in value.items()
+            }
+        return DataPackage(**split_data)
+
+    def _split_mudata(
+        self,
+        mudata: MuData,  # ty: ignore[invalid-type-form]
+        indices_map: Dict[str, Dict[str, np.ndarray]],
+        split: str,
+    ) -> MuData:  # ty: ignore[invalid-type-form]
+        """Splits a MuData object based on the provided indices map.
+
+        Args:
+            mudata: The MuData object to split.
+            indices_map: A dictionary mapping modalities to their respective indices.
+            split: The split type ("train", "valid", or "test").
+        Returns:
+            A new MuData object with the specified splits applied.
+        """
+        for modality, data in mudata.mod.items():
+            indices = indices_map[modality][split]
+            mudata.mod[modality] = self._indexing(data, indices)
+        return mudata
+
+    def _requires_paired(self) -> bool:
+        return self.config.requires_paired is None or self.config.requires_paired
+
+    def split(self) -> Dict[str, Optional[Dict[str, Any]]]:
+        """Splits the underlying DataPackage into train, valid, and test subsets.
+        Returns:
+            A dictionary containing the split data packages for "train", "valid", and "test".
+            Each entry contains a "data" key with the DataPackage and an "indices" key with
+            the corresponding indices.
+        Raises:
+            ValueError: If no data package is available for splitting.
+            TypeError: If indices are not provided for unpaired translation case.
+        """
+        if self._data_package is None:
+            raise ValueError("No data package available for splitting")
+
+        splits = ["train", "valid", "test"]
+        result: Dict[str, Optional[Dict[str, Any]]] = {
+            "train": {},
+            "valid": {},
+            "test": {},
+        }
+
+        for split in splits:
+            if self.indices is None:  # or split not in self.indices:
+                result[split] = None
+                continue
+            result[split] = {
+                "data": self._split_data_package(split=split),
+                "indices": self.indices,
+            }
+
+        return result

autoencodix-0.2.2/src/autoencodix/data/_datasetcontainer.py

@@ -0,0 +1,33 @@
+from typing import Optional, Union
+from dataclasses import dataclass
+from autoencodix.base._base_dataset import BaseDataset
+
+
+@dataclass
+class DatasetContainer:
+    """A container for datasets used in training, validation, and testing.
+
+    Attributes:
+    train : Dataset
+        The training dataset.
+    valid : Dataset
+        The validation dataset.
+    test : Dataset
+        The testing dataset.
+    """
+
+    train: Optional[Union[BaseDataset, None]] = None
+    valid: Optional[Union[BaseDataset, None]] = None
+    test: Optional[Union[BaseDataset, None]] = None
+
+    def __getitem__(self, key: str) -> BaseDataset:
+        """Allows dictionary-like access to datasets."""
+        if key not in {"train", "valid", "test"}:
+            raise KeyError(f"Invalid key: {key}. Must be 'train', 'valid', or 'test'.")
+        return getattr(self, key)
+
+    def __setitem__(self, key: str, value: BaseDataset):
+        """Allows dictionary-like assignment of datasets."""
+        if key not in {"train", "valid", "test"}:
+            raise KeyError(f"Invalid key: {key}. Must be 'train', 'valid', or 'test'.")
+        setattr(self, key, value)