Update train.py

[AdnanKahveci]

Implemented a complete training pipeline for the Glow model, including:

• Data Loading: Added dynamic data loading with image transformations.
• Training Loop: Includes loss calculation, learning rate scheduling, and early stopping.
• Validation: Periodically validates the model and saves checkpoints.
• Model Checkpointing: Saves model and optimizer state at intervals.
• Early Stopping: Halts training when no improvement is observed over a specified number of iterations.