Update app.py

app.py

@@ -69,11 +69,12 @@ pipe = pipeline("automatic-speech-recognition", model="NbAiLab/nb-whisper-large"
 @spaces.GPU()
 def transcribe_audio(audio_file, batch_size=16, sample_rate =16000):
     # --audio file to tensor
-    waveform = waveform.numpy(audio_file)
-    samples = samples.numpy(audio_file)
+    
+    waveform, samples = torchaudio.load(waveform, samples)
     if waveform.ndim > 1:
         waveform = waveform[0, :]
-    waveform, samples = torchaudio.load(waveform, samples)
+    waveform = waveform.numpy(audio_file)
+    samples = samples.numpy(audio_file)
 
     start_time = time.time()