add new speech language？

[zhusy09]

No description provided.

[avidale]

Hi @zhusy09,
Could you please elaborate on your question?
Do you want to contribute a speech encoder for a new language?
If so, we could probably help you by sharing a training recipe based on an existing text encoder and an ASR dataset.

[JAVI897]

Hi, I would like to get the training recipe to train a new speech encoder based on an existing text encoder and an ASR dataset

[laleye]

Hi @avidale
I'm also planning to add a new language. Can I have the training recipe for integrating a new language?

[MayYard]

Hi @avidale , I'd also like to ask for the training recipe for the speech encoder to add a new language.

[avidale]

Hi @JAVI897, @laleye and @MayYard!
I found some time to prepare a speech encoder training recipe that would depend on the Fairseq2 code in a minimal way, and would work in a limited computational environment (Google Colab pro).
Here is a notebook with it: https://colab.research.google.com/drive/1e8Ta7b5z1loIEfDI3pstW_AXZOMAX-_f

So far, I am not happy neither with the quality of the resulting model, nor with the compactness of the code, so I don't commit the notebook to the repo yet.

Still, if you have some time to play around it, I would be happy to get your feedback.

[JAVI897]

Thanks @avidale . I will take a look!