torch_train_loop is a general-purpose train-loop for PyTorch models, with some convenient features built-in:
- Integration with TensorBoard (via PyTorch's SummaryWriter) for
plotting:
- Training loss
- Validation loss
- Additional optional metrics
- Progress bar(s) (via tqdm) for Jupyter Notebooks or CLI environments.
- Early stopping overfitting detection.