Hey! I am trying to finetune bark with your code to generate children's voices. However, coarse and fine module losses do not decrease during the finetuning. I am using default parameters as in this repo. Am I doing something wrong? Below, you can see the coarse and fine loss graphs respectively.
dagshub[bot]
commented
1 year ago
Hey! I am trying to finetune bark with your code to generate children's voices. However, coarse and fine module losses do not decrease during the finetuning. I am using default parameters as in this repo. Am I doing something wrong? Below, you can see the coarse and fine loss graphs respectively.