Update logic.py

logic.py

@@ -104,7 +104,7 @@ def extract_speech_embedding(audio_path: str):
     audio_norm = audio / 32768.0
     audio_norm = audio_norm.unsqueeze(0)
     audio_norm = torch.autograd.Variable(audio_norm, requires_grad=False).to(device)
-    melspec = stft.mel_spectrogram(audio_norm).transpose(1,2).half()
+    melspec = stft.mel_spectrogram(audio_norm).transpose(1,2).float()
 
     if melspec.shape[1] <= 128:
         mel_slice = mel