How to add embeddings or adding support for melody embeddings

[raghavpatnecha]

I am able to train and run the modal with text. It would be great if we can add melody embeddings as mentioned on the examples page of google research

[mansipandyaa29]

Where'd you get the Dataset from? I think this is the one https://research.google.com/audioset/download.html they used but I'm not sure I should use the CSV or the other format.