Question about the purpose and calculation of `guided_loss` in `ProsodyAligner`

Hi, I have a question about the `guided_loss` in the `ProsodyAligner` class. Could you please clarify:

1. **What is the purpose of `guided_loss`?**
2. **Why is it calculated in this specific way, using the attention mask (`attn_w_emo`) and the `guided_sigma` parameter?**

Thanks for your help!