Reverse Generation on Magenta Tensorflow

[Douglas Oliveira Paschoal]

I did some research in the Magenta Tensorflow repository and identified that the model is capable of continuing polyphonic music through MIDI file input. Is there any method to do the reverse generation? That is, instead of me submitting the beginning of a song and the AI ​​generating the middle, I can submit the middle of the song and the AI ​​generating the beginning of it? Or are the Magenta generations just appended to the end of the song?