To train, run:
python script/train.py --path absolute_path_to_data --logdir log/ --other_params
It will (1) load the training data by reading absolute_path_to_data/labels.csv; (2) save config.json, tensorboard log file, and model.pt to current_path/log/.
To evaluate a trained model, run:
python script/eval.py --data_path absolute_path_to_data --model_path absolute_path_to_model --logdir log/
It will (1) load the evaluation data from absolute_path_to_data; (2) load trained model from absolute_path_to_model/model.pt; (3) save pred_labels.csv to current_path/log/.