linmult.core.utils
==================

.. py:module:: linmult.core.utils

.. autoapi-nested-parse::

   Utility functions: config loading and logit aggregation.



Functions
---------

.. autoapisummary::

   linmult.core.utils.load_config
   linmult.core.utils.apply_logit_aggregation


Module Contents
---------------

.. py:function:: load_config(config_file: str | pathlib.Path) -> dict

   Load a YAML configuration file.

   :param config_file: Path to the YAML file.
   :type config_file: str | Path

   :returns: Parsed configuration dictionary.
   :rtype: dict


.. py:function:: apply_logit_aggregation(x: torch.Tensor, mask: torch.Tensor | None = None, method: str = 'meanpooling') -> torch.Tensor

   Aggregate logits across the time dimension.

   Only timesteps where the mask is ``True`` (valid) contribute to the result.
   Fully-masked samples (all ``False``) return a zero vector.

   :param x: Logit tensor of shape ``(B, T, F)``.
   :type x: torch.Tensor
   :param mask: Boolean validity mask of shape ``(B, T)``.
                ``True`` = valid timestep. If ``None``, all timesteps are treated as valid.
   :type mask: torch.Tensor, optional
   :param method: Aggregation method. One of:

                  - ``"meanpooling"``: Masked mean over the time dimension.
                  - ``"maxpooling"``: Masked max over the time dimension.
   :type method: str

   :returns: Aggregated output of shape ``(B, F)``.
   :rtype: torch.Tensor

   :raises ValueError: If ``method`` is not one of the supported values.