aik2mlj / polyffusion

Polyffusion: A Diffusion Model for Polyphonic Score Generation with Internal and External Controls
https://polyffusion.github.io
MIT License
71 stars 8 forks source link

how to create unique dataset #1

Closed taktak1 closed 9 months ago

taktak1 commented 12 months ago

Thank you for developing such a framework.

What I want to do is learn a unique model using a unique dataset, but how is this possible?

https://github.com/music-x-lab/POP909-Dataset/blob/master/data_process/data_process.ipynb For example, the file format created in pop909's notebook is npy, not npz. Also, you are loading the list file when loading the dataset, right? I don't even know how to make it.

@aik2mlj

aik2mlj commented 11 months ago

Hi! I will look into this and try to provide a more convenient way to load custom dataset in the following days! As for now, you may take a look at the midi_to_data.py and datasample.py file here: https://github.com/aik2mlj/polyffusion/blob/sdf_prmat2c%2Bpop909/polyffusion/data/midi_to_data.py https://github.com/aik2mlj/polyffusion/blob/sdf_prmat2c%2Bpop909/polyffusion/data/datasample.py

They can get a midi file input and transform it into the data format we use in the model!

aik2mlj commented 9 months ago

See https://github.com/aik2mlj/polyffusion/pull/3