fairseq
fairseq copied to clipboard
letter vocabulary for other language?
Hello guys i have just trained my wav2vec2 model on Nepali speech dataset from this page. I want to now fine tune this model but looks like i need a letter vocabulary for this step. Can anyone suggest me how should i complete this fine tuning part also with nepali vocabulary? or can i use the english vocabulary itself?
@roboticsai did you make any progress on this? I was excited to see the MMS launch but then there's no TTS support for Nepali. https://github.com/facebookresearch/fairseq/tree/main/examples/mms
I'm hoping to fine tune that model to add Nepali support. Let me know if you have any thoughts.