Music Generation Models: Which dataset was used?

[candidexmedia]

Hi OpenVINO team! I'm curious to know how the musicgen models were trained, and what dataset was used, and how the data was sourced.

[RyanMetcalfeInt8]

Hello! The Music Generation features uses the MusicGen-small model trained by Meta, and released publicly here: https://huggingface.co/facebook/musicgen-small

Check the model card there for more information about the method(s) used, training set, etc. Thanks!

[candidexmedia]

Thank you so much for your quick reply and link!

In case anyone else is wondering:

The model was trained on licensed data using the following sources: the Meta Music Initiative Sound Collection, Shutterstock music collection and the Pond5 music collection. See the paper for more details about the training set and corresponding preprocessing.