I was using the pretrained etdnn model uptil now , but now I wish to fine tune it using speaker verification data that I have. Would this script suffice :
run_xvector_1a.sh.
I am using xvectors as the embeddings currently in my pipeline as well.
If there is anything I am missing out on or unaware of, kindly help as I am new to kaldi.
I have read the paper mentioned in the script and I feel that it seems the right way to go about it.
Thanks