Sepformer training for 4 speakers source separation on 48KHz data #2387
Unanswered
hahmadraza
asked this question in
Q&A
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
I'm trying to train Sepformer for 4 speakers' source separation on 48KHz data (I prepared my own 48KHz data with roughly 15000 training sequences and 3000 test sequences). I prepared the data in Libri4Mix format and added code to support 4 speakers. Test-Set SI-SNRi is 8.88dB and Test-Set SDRi is 9.44dB after 78 epochs and it was not improving till 146 epocs. I stoppoed the training after that.
What are the suggestions to improve Test-Set SI-SNRi and Test-Set SDRi. What hyperparametrs should i try optimizing?
Thanks
Beta Was this translation helpful? Give feedback.
All reactions