• There are several models for speaker verificaiton task (wavlm-ecapa / xvector / ...).
  • Some of those model where trained with AAM Softmax loss, which gets the number of labels as input.
  • When training speaker verificaiton task, we would split the data to train and validate (i.e different speakers in train set and different speaker in validaion set).
  • If the train and validation datasets have different number of speakers, how can we train the model while using AAM Softmax loss ?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.