Ruoyi
there's a python file called quantize.py, you can view the file and choose to quantize. 
因为auto_rerank功能需要下载funasr的权重，时间长的话请自行检查网速是否太慢。
Mac is not supported for training/Inferring.
planned.
because we haven't use other language in training for 1.2 version
please use torch==2.3.0
Closed as issue had been solved.
请点Reference Audio，然后勾选启用参考音频，之后上传你的参考音频。
4060推理大约15tokens/s，具体效果可以自行测试。 编译后大概是100tokens/s.
> 请问下，“编译后”是指加streaming=True这个编译参数吗，我这边加上之后，速度没怎么变，还是30tokens/s呢。 自己看文档