finetune on which branch

[jammyWolf]

i have 2 questions：

1. released music model is trained both audio and text branch or just trained on audio branch？
2. i followed the finetune_script finding it only finetuneed audio branch for classification, is that right? if i want to finetune both text and audio branch i should unfreeze the text branch ?

[lukewys]

Hi,

1. both branch
2. Yes. If you want to fine-tune both text and audio branch, you should use the training script but not the fine-tune script. Sorry for the potential confusion of the naming, the fine-tune script is for fine-tuning the audio encoder for downstream task.

Best