Configs
=======

Hydra configuration files defining models, data modules, trainers, logging,
and experiment settings.

All training and evaluation behaviour is controlled via these YAML configs.

How it works
------------

The top-level ``config.yml`` defines the default configuration by including
entries from the subfolders:

.. code-block:: yaml

   defaults:
     - model: simple
     - data: dummy
     - trainer: default
     - logging: mlflow
     - experiment: base

Hydra then loads the corresponding files:

- ``configs/model/simple.yml``
- ``configs/data/dummy.yml``
- ``configs/trainer/default.yml``
- ``configs/logging/mlflow.yml``
- ``configs/experiment/base.yml``

These files are merged together into one final config object, which is passed
into the training script.

Subdirectories
--------------

- ``model/``

  Defines model architectures and their hyperparameters.
  Each file here must contain a ``_target_`` field that points to a Python
  class implementing the model.
  Example: selecting different neural network variants or LightningModules.

- ``data/``

  Configures DataModules, dataloaders, dataset paths, and batch sizes.
  Each file should describe how training/validation/test data is created or
  loaded.

- ``trainer/``

  Contains configurations for PyTorch Lightning’s Trainer: number of epochs,
  devices, precision settings, logging frequency, etc.

- ``logging/``

  Parameters for MLflow or other logging backends.
  This keeps experiment tracking separate from model or data settings.

- ``experiment/``

  Optional presets for organising runs (name, notes, parameter overrides).
  Useful for keeping track of different experiment configurations.

Using the configs
-----------------

Hydra is invoked in ``train.py`` via:

.. code-block:: python

   @hydra.main(config_path="../../configs", config_name="config")
   def main(cfg):
       ...

You can start training with the default config:

.. code-block:: bash

   python -m FRAME_FM.training.train.py

Or override specific config options from the command line:

.. code-block:: bash

   python -m FRAME_FM.training.train.py model=simple data=cifar10 trainer.max_epochs=50