Needed guidance on fine-tuning whisper on custom dataset.

#1
by Naram - opened

Hello sir,
I want to fine-tune whisper medium with my dataset in kaldi format. Whenever I use transfer learning approach provided in the documentation It is running fine till stage -11 and while decoding it is giving me errors related to mismatch of tensor sizes. Can you please suggest me on how to fine-tune my data for ASR task?
Thankyou.

Naram changed discussion status to closed
ESPnet org

It would be better to ask it at https://github.com/espnet/espnet/issues
I can assign an appropriate person to answer your question.

Sign up or log in to comment