- Implemented LSTM model with enhanced training features including mixed precision, early stopping, and validation via DataLoader.
- Introduced a new 1D Temporal Convolutional Network (TCN) model for predicting hysteretic curves.
- Added functionality for creating sliding windows from the dataset for both training and testing.
- Integrated scaling for features and targets using StandardScaler.
- Enhanced data handling with explicit cleanup and memory management for GPU.
- Updated main function to include model training and prediction for both LSTM and TCN.
- Saved predictions to CSV files for further analysis.