Estadísticas de Exploring the effects of silent data corruption in distributed deep learning training