I trained my model and run eval.py regularly. Then I open tensorboard to listen to the audio sample in both mode. The audio in train mode is very clear but the audio in eval mode is noisy and cannot hear anything. Also the train loss is lower than 0.075 but eval loss is 1.0.
Then I run synthesize.py to generate sample. But the sample is same as eval mode.
Can anyone told me how to fix it and why?
thkxxxxxx
I trained my model and run eval.py regularly. Then I open tensorboard to listen to the audio sample in both mode. The audio in train mode is very clear but the audio in eval mode is noisy and cannot hear anything. Also the train loss is lower than 0.075 but eval loss is 1.0. Then I run synthesize.py to generate sample. But the sample is same as eval mode. Can anyone told me how to fix it and why? thkxxxxxx