support fish speech TTS model

[Tint0ri]

Feature request / 功能建议

support the fish speech model. Which sounds better than ChatTTS and CosyVoice

Motivation / 动机

Better TTS solution

Your contribution / 您的贡献

https://huggingface.co/fishaudio/fish-speech-1.2-sft

[qinxuye]

Thanks, we need to some investigation first.

[Tint0ri]

According to my simple test, far more better than ChatTTS and CosyVoice

[qinxuye]

@codingl2k1 I think we can support this next week.

[Tint0ri]

I try to run fish speech tts on cpu. Get the error log: hydra.errors.MissingConfigException: [address=0.0.0.0:34645, pid=158] Cannot find primary config 'firefly_gan_vq'. Check that it's in your config search path.

Config search path: provider=hydra, path=pkg://hydra.conf provider=main, path=file:///opt/conda/lib/python3.11/site-packages/xinference/thirdparty/fish_speech/fish_speech/configs provider=hydra-colorlog, path=pkg://hydra_plugins.hydra_colorlog.conf provider=schema, path=structured://

inference version: 0.14.3

[qinxuye]

Could you open a new issue about it?

[Tint0ri]

open new issue: https://github.com/xorbitsai/inference/issues/2195