How to use the model trained with lifeiteng's code to inference

Plachtaa / VALL-E-X

An open source implementation of Microsoft's VALL-E X zero-shot TTS model. Demo is available in https://plachtaa.github.io/vallex/

MIT License

7.59k stars 756 forks source link

How to use the model trained with lifeiteng's code to inference #87

Open chazo1994 opened 1 year ago

chazo1994 commented 1 year ago

I have trained a model with lifeteng's code, so how can I used this model with this repository? Could you please provide a guideline to do that.

Plachtaa commented 1 year ago

lifeiteng's model class does not contain ar_language_embedding and nar_language_embedding by default. To directly load your trained state dictionary, you have to manually initialize them to be all-zero matrix (if your trained model is monolingual, else you have to find a way to make your language embeddings compatible to this repo's), and then you should be able to use it flawlesslly.