try add token

app.py

@@ -463,7 +463,7 @@ class WhisperTranscriber:
         global _embedder
         if _embedder is None:
             # window="whole" to compute one embedding per provided chunk
-            _embedder = Inference("pyannote/embedding", window="whole", device=torch.device("cuda"))
+            _embedder = Inference("pyannote/embedding", window="whole", device=torch.device("cuda"),use_auth_token=os.getenv("HF_TOKEN"),)
         return _embedder
 
     def assign_speakers_to_transcription(self, transcription_results, diarization_segments):