…AN training process of updating of generator and discriminator separately.

…mprove readability

…nator fixture

… optimizer management

… forward groups derived context objects

…ndling and streamline output dimension calculations

…to use lowercase tuple syntax

…lemented to raising NotImplementedError

…od to raise KeyError

…to support a variety of forward function signatures in concrete loss classes, the old abstract AbstractLoss class has been replaced with a protocol and a non-abstract BaseLoss. Related modules and methods using the old abstract losses are also updated. And a new module is implemented to hourse brushed up wGAN training losses whereas the obsetele modules were removed.

… for inputs and outputs, enhancing type clarity and consistency.

…l instead of BaseGeneratorModel for consistency.

… for flexibility.

…nator

…flowLogger for optional dictionary parameters

…weights and add model registry for logging parameters

…with logging

… methods

- Introduced a new script for training a Wasserstein GAN (WGAN) using a U-Net generator.
- Implemented logging with MLflow to track metrics and visualize predictions during training.
- Included dataset processing utilities to handle image files and create a cropped dataset for training.
- Added functionality to visualize training outcomes and metrics from the MLflow tracking server.
- Configured hyperparameters suitable for demo purposes, including learning rates and batch size.