Loss stuck after 600k iteration and not improving.

[gashishoo7]

Hi, I trained yourtts model using the custom dataset using . Trained for 1M iteration, my loss is stuck at 600k iteration and not improving. Particularly, avg_loss_mel value is stuck at 20. Earlier we observed good results for 22050 Hz sampling rate when avg_loss_mel was converging somewhere between 11. How can I further reduce loss? Please suggest.