Models from paper "U-MusT: A Unified Framework for Cross-Modal Translation of Score Images, Symbolic Music, and Performance Audio"