credit.preblock.norm
====================

.. py:module:: credit.preblock.norm

.. autoapi-nested-parse::

   norm.py
   -------
   ERA5Normalizer: normalizes per-variable ERA5 tensors using mean/std NC files.

   Operates on the raw batch structure from MultiSourceDataset::

       batch["era5"]["input"]["era5/field_type/3d/varname"]  = (B, n_levels, T, H, W)
       batch["era5"]["input"]["era5/field_type/2d/varname"]  = (B, 1, T, H, W)

   Variables absent from the mean/std file are passed through unchanged.

   Registered in the preblock registry as ``"era5_normalizer"`` so it can be
   included via the config's ``preblocks:`` section::

       preblocks:
         norm:
           type: era5_normalizer
           args:
             mean_path: /path/to/mean.nc
             std_path:  /path/to/std.nc



Attributes
----------

.. autoapisummary::

   credit.preblock.norm.logger


Classes
-------

.. autoapisummary::

   credit.preblock.norm.ERA5Normalizer


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

.. py:data:: logger

.. py:class:: ERA5Normalizer(mean_path: str, std_path: str, levels: list[int] | None = None)

   Bases: :py:obj:`credit.preblock.base.BasePreblock`


   Normalizes per-variable ERA5 tensors using pre-computed mean/std files.

   Normalization: ``(x - mean) / std`` applied per variable. Variables not
   found in the statistics file are passed through unchanged.

   :param mean_path: Path to NetCDF file containing per-variable means.
   :param std_path: Path to NetCDF file containing per-variable standard deviations.
   :param levels: Optional list of 1-indexed model levels to select from the
                  full 137-level stats (e.g. [60, 90, 120, 137] for a 4-level
                  smoke test).  When omitted, all levels in the stats file are
                  used.


   .. py:attribute:: _mean
      :type:  dict[str, torch.Tensor]


   .. py:attribute:: _std
      :type:  dict[str, torch.Tensor]


   .. py:method:: _normalize_tensor(key: str, tensor: torch.Tensor) -> torch.Tensor

      Normalize *tensor* using the variable name extracted from *key*.



   .. py:method:: forward(batch: dict) -> dict

      Normalize all input/target tensors, returning a new batch dict.