JSFakes (Dr. Tristan Behrens).
This is a tokenized version of the JS-Fakes dataset by Omar Peracha. The original dataset can be found here: js-fakes.git
The representation is four tracks with four bars per track.
Purpose.
This dataset is a good starting point for Music Generation. You could train GPT-2 on the samples to compose music.
Contact.
Find me on LinkedIn and say hello.
If you find and issue or have a feature request, please contact me.
Please be so kind and like this dataset if you find it useful.