:stuck_out_tongue_closed_eyes: TensorFlowTTS: Real-Time State-of-the-art Speech Synthesis for Tensorflow 2 (supported including English, French, Korean, Chinese, German and Easy to adapt for other languages)
I’m here again, thanks for your help. Recently, I use baker and other male voice about 40 min to train fs2 and mb_melgan, I found that the loss of mb_melgan will change suddenly after 5000 steps, from less than 1 to thousand,I also found that when I use baker and other five speaker dataset which include 3 male and 2 female voice about 30 min ~1 hour for each other, the same thing happened again after 35k steps , shouldn't I use male and female to train mb_melgan and to train fs2? The TTS model doesn't work well which fs2 is trained with baker and male voice and mb_melgan is trained with baker, is it normal?
I’m here again, thanks for your help. Recently, I use baker and other male voice about 40 min to train fs2 and mb_melgan, I found that the loss of mb_melgan will change suddenly after 5000 steps, from less than 1 to thousand,I also found that when I use baker and other five speaker dataset which include 3 male and 2 female voice about 30 min ~1 hour for each other, the same thing happened again after 35k steps , shouldn't I use male and female to train mb_melgan and to train fs2? The TTS model doesn't work well which fs2 is trained with baker and male voice and mb_melgan is trained with baker, is it normal?