Questions about Training with Magnet

[gyujin-kim]

I have a few questions regarding training with Magnet:

If we train the model using our own audio dataset, what would be the license for the resulting weights?
Would it be possible to use them for commercial purposes?

In cases where a long audio file has sections with different moods, is it better to split the audio into separate tracks for each mood during training, or is it acceptable to input a multi-mood track as is?

edit :
I am interested in using the Encodec 32kHz model available on Hugging Face (link to model). Could you please clarify what the license for this model is? Specifically, I would like to know if there are any restrictions on its use, including for commercial purposes.