credit.applications.preprocess
==============================

.. py:module:: credit.applications.preprocess


Functions
---------

.. autoapisummary::

   credit.applications.preprocess._scaler_probe_range
   credit.applications.preprocess._scaler_device
   credit.applications.preprocess._log_single_scaler
   credit.applications.preprocess.log_fitted_scalers
   credit.applications.preprocess.main


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

.. py:function:: _scaler_probe_range(scaler)

   Return ``(lo, hi, label)`` normalized probe endpoints for a fitted scaler.

   The endpoints span the range of normalized values the scaler typically
   produces, so an inverse transform reveals what physical values each scaler
   maps that range to:

     - minmax    -> ``[0, 1]``
     - standard  -> ``[-4, 4]`` (±4 standard deviations)
     - quantile  -> ``[0, 1]`` for a uniform output distribution, otherwise
                    ``[-4, 4]`` (e.g. normal/logistic)


.. py:function:: _scaler_device(scaler)

   Best-effort lookup of the torch device holding a scaler's fitted stats.


.. py:function:: _log_single_scaler(scaler, name, logger)

   Log the fitted parameters of one leaf scaler, one line per channel/level.


.. py:function:: log_fitted_scalers(scaler_dict, logger, path=())

   Recursively log every fitted scaler in a nested ``scaler[data_type][source][var]`` dict.


.. py:function:: main()