Pre-trained models at paper submission

Each pre-trained model consists of the following files:

• best_thresholds.csv: Best classification thresholds for each label (not available for final models)
• model_config.json: Model hyper-parameter, e.g. hidden layer size.
• model_weights: Trained model weights
• outputs_with_ids.pickle: Prediction for validation set
• report.json/txt: Classification report (output of sklearn.metrics.classification_report, not available for final models)
• report.json/txt: Classification report only for books with available author embeddings (output of sklearn.metrics.classification_report, not available for final models)
• submission.txt: Prediction for validation/test set in output format for shared task.