realease v0.1.0

This is a beta pretrain-model of hifi-vaegan.

Caution! This pretrain-model only rebuild speech and vocal, not a universal vocoder pretrain weight.

1. finetune models are used to finetune
2. release model are used to be a vocoder, some project use it (such as latent-diffusion-speech)

Acknowledge

Dataset Preparation:
@CNChTu
@Stardust-minus