Training code?

[GXcells]

You did an amazing work.
I know it is very early but do you know if it will be possible to fine-tune the model and if yes are you planning to release a code for it or should I wait for community to work on it?

Thanks

[Nehc]

Yes, fine-tune of such a model is very interesting!

[yukiarimo]

I saw the fine-tuning was merged, but what does it mean no support to fine-tune the generation part? Is it possible train it on the Mel-spectrograms to turn into a text-to-speech-ish model like:

<text>Hello, this is a demo voice</text>
<audio>{audio_tokens_for_further_convertion_with_any_vocoder}