Embedding of audio?

[turian]

How can I use your model to get an embedding JUST of the audio file?

[shansongliu]

If you mean music embedding, you can just use the standalone MERT model, there is no need to use the whole MU-LLaMA model. If you want to use exactly the embedding that generated from our pipeline, then you may modify the model part to let it output the embedding by yourself.