Why was this parameter set to a non-zero value for training the speaker diarization model?The concept of masking is just for learning representation. For the substream task, you should not do that.
· Sign up or log in to comment
