I am working with GMMHMM from hmm learn for speaker recognition. After training the model, I found that all the values in the covariance matrix (covars_) all have the same v