kkk
kkk
The demo folder does not exist, could you please provide demo midi file for model inference?
你好,使用AISHELL-3数据集按照readme进行训练,loss的图如下,尝试推理但是不能合成出正常声音,请问是为什么呢? text/symbols.py已经更改为中文的。  
When I train with the AIShell-3 dataset, the generated audio has a strange tone. Have you encountered similar issues?