Text classification model for threat detection in Spanish.

We define threat messages as messages containing threats, support of violence, or harm towards a person or group of people.

It is trained on the XML-RobertA (xlm-roberta-base) model with a dataset of 364.793 messages from online sources.

Parameters "learning_rate": 1e-5, "weight_decay": 0.01, "opimizer": "Adam", "eps": 1e-08, "num_labels": 2, "output_attentions": "True", "output_hidden_states": "False", "number_epoch": 3, "batch-size": 16