jonatasgrosman / wav2vec2-sprint

MIT License
174 stars 31 forks source link

Finetune model for Japanese language with LaboroTVSpeech corpus #2

Open marlon-br opened 3 years ago

marlon-br commented 3 years ago

Hi, thank you for all the work you did with the models training.

Recently I discovered quite good speech corpus for Japanese language: https://github.com/laboroai/LaboroTVSpeech. Could you please be so kind and add it to your finetuned model for Japanese language?

jonatasgrosman commented 3 years ago

Hi @marlon-br! Thanks for the corpus link!

I'm planning to improve the Japanese model (and other languages too), but I'm running out of GPU resources, and I'm focusing on improving a small subset of my models right now, and unfortunately Japanese isn't in this subset 😢

But I'm figuring out how to get access to more GPU resources to improve my models, and Japanese is one of the models that I really want to improve soon, and the corpus you've sent will be handy for this 😃

Regards Jonatas Grosman