verify_files() function doesn't work

When I try to verify 2 audio files with the function verify_files() with the next few lines of code:

from speechbrain.pretrained import SpeakerRecognition
verification = SpeakerRecognition.from_hparams(source="speechbrain/spkrec-ecapa-voxceleb", savedir="pretrained_models/spkrec-ecapa-voxceleb")
score, prediction = verification.verify_files("voxceleb1/wav/id10270/5r0dWxy17C8/00010.wav", "voxceleb1/wav/id10305/19OCRADWjgA/00006.wav")
print(score)

The resulting behavior is weird. It works like a decreasing counter. Every time I run these lines:

score, prediction = verification.verify_files("voxceleb1/wav/id10270/5r0dWxy17C8/00010.wav", "voxceleb1/wav/id10305/19OCRADWjgA/00006.wav")
print(score)

The score is decreasing in 0.0001.

So if I run a loop:

for i in range(10):
  score, prediction = verification.verify_files("voxceleb1/wav/id10270/5r0dWxy17C8/00010.wav", "voxceleb1/wav/id10305/19OCRADWjgA/00006.wav")
  print(score)

This is the result:

I think It may be a problem in the verify_files() method, I tried it with many different files and from different computers and with different pretrained models (ECAPA, XVectors), it was always the same, a decreasing counter.

[talgoldengoren]

Yes, the score stays the same now. Thank you 😃 @Gastron

[Gastron]

I’ve edited the embedding normalization on HuggingFace. Does this solve the issue for you as well, @talgoldengoren ?