omniloader.transforms
=====================

.. py:module:: omniloader.transforms

.. autoapi-nested-parse::

   Configurable, per-sample transforms: normalization and augmentation.

   This subpackage groups the transform pipeline used after unification:

   * :mod:`~omniloader.transforms.base` — :class:`Transform`, :class:`Compose` and
     shared helpers;
   * :mod:`~omniloader.transforms.normalize` — :class:`Normalize`,
     :class:`MinMaxNormalize`, :class:`RobustNormalize`, :class:`InstanceNormalize`;
   * :mod:`~omniloader.transforms.augment` — :class:`GaussianNoise`,
     :class:`FeatureDropout`, :class:`SpanMasking`, :class:`TimeWarp`,
     :class:`FeatureMasking`;
   * :mod:`~omniloader.transforms.crop` — :class:`RandomCrop`, :class:`CenterCrop`;
   * :mod:`~omniloader.transforms.mix` — :class:`MixupCollator` (collate-time);
   * :mod:`~omniloader.transforms.stats` — statistics helpers for normalization.

   Compose transforms with :class:`Compose`, or build one from a list of config dicts
   with :func:`build_transform`.



Submodules
----------

.. toctree::
   :maxdepth: 1

   /autoapi/omniloader/transforms/augment/index
   /autoapi/omniloader/transforms/base/index
   /autoapi/omniloader/transforms/crop/index
   /autoapi/omniloader/transforms/mix/index
   /autoapi/omniloader/transforms/normalize/index
   /autoapi/omniloader/transforms/stats/index


Attributes
----------

.. autoapisummary::

   omniloader.transforms.TRANSFORMS


Functions
---------

.. autoapisummary::

   omniloader.transforms.build_transform


Package Contents
----------------

.. py:data:: TRANSFORMS
   :type:  dict[str, type[base.Transform]]

.. py:function:: build_transform(configs: collections.abc.Sequence[collections.abc.Mapping[str, Any]], schema: omniloader.schema.spec.UnifiedSchema | None = None) -> base.Compose | None

   Build a :class:`Compose` from a list of transform config dicts.

   Each config dict has a ``name`` (a key of :data:`TRANSFORMS`) and the
   remaining keys are forwarded as keyword arguments. Schema-aware transforms
   receive ``schema`` automatically when they declare no explicit ``keys``;
   stats-based normalizers read their ``stats`` from the config — either inline,
   or from a saved JSON file via ``stats_path`` (and ``union_stats_path`` for
   :class:`~omniloader.transforms.normalize.PerDatasetNormalize`), loaded with
   :func:`~omniloader.transforms.stats.load_stats`. This lets a config reference
   stats produced once by ``omniloader compute-stats`` instead of inlining them.

   :param configs: A list of transform specifications.
   :param schema: The unified schema, injected into schema-aware transforms.

   :returns: A :class:`Compose`, or ``None`` if ``configs`` is empty.

   :raises ValueError: If a config names an unknown transform.