Samples are organised as following:

fugue_prompted - entire fugues generated by the model, each is seeded by the first three bars of a Bach fugue.

fugue_test_set - four bars of fugue generated by the model, each is seeded by the soprano voice from
a Bach fugue. the files are organised according to:

        source - the input that the model received 
        original - bach's original harmonisation
        model - the re-harmonisation produced by the model

chorale_test_set - four bars of chorale generated by the model, each is seeded by the soprano voice from
a Bach chorale. the files are organised according to:

        source - the input that the model received 
        original - bach's original harmonisation
        model - the re-harmonisation produced by the model