预训练模型会开源吗?
JJun-Guo opened this issue · 3 comments
JJun-Guo commented
预训练模型会开源吗?
y-ren16 commented
Limited by computing resources, we trained on the LibriTTS dataset. The weight file has been uploaded to https://huggingface.co/y-ren16/TiCodec. If you need better results, you can train on a larger dataset yourself.
Liujingxiu23 commented
@y-ren16 请问如果不使用“The time-invariant representation extraction and quantization”这部分,在单码本情况下 ,模型效果如何? How about the performace when nq=1 if I did not use the quantizer Q2 which extract the time-invariant representation?
y-ren16 commented
不加的话就是论文中实验对比的单码本HiFiCodec