Could you please make all the files related to the KIT dataset publicly available? This would include configuration files and pre-trained models.