Is there any .ipynb tutorial is there to fine-tune Whisper for one language without making it lose its any of previous pretrained data?

[Singhamarjeet8130]

I have successfully trained whisper, only to find to out that the all the previous pretrained data has been lost.
So is there any .ipynb tutorial is available to fine-tune whisper for any supportive language without making it lose its any of the previous pretrained data?

[phineas-pta]

what previous pretrained data ?

[phineas-pta]

after fine tuning, whisper ability to detect language drop catastrophically, do u specify lang when transcribing

[Singhamarjeet8130]

Hindi

[Singhamarjeet8130]

i specified the above language when transcribing

[phineas-pta]

so it works ?

[Singhamarjeet8130]

The training completed successfully but as you said that after fine tuning, whisper ability to detect language drop catastrophically and exactly this happened. So i have been searching for way to fine-tune whisper model for improving one language without making it lose its existing data of all other languages.