Configurations

python-Chebai utilizes PyTorch Lightning for model development, training, and inference, offering structured project organization and enhanced configurability.

The configs are saved in the configs folder. Each component such as training, data, model etc., has their own configuration YAML file.

Configs Folder Structure

• configs/training/: Basic trainer, callbacks, and logger configurations.
• configs/data/: Configurations for different dataset types used in training and evaluation.
• configs/loss/: Custom loss function configurations for training and fine-tuning.
• configs/metrics/: Custom configurations for evaluation metrics.
• configs/model/: Configurations for different models.
• configs/weightings/: Weight values for different datasets used in training.

For detailed information on available arguments for each module, refer to the PyTorch Lightning CLI documentation: PyTorch Lightning for CLI Docs