Update app.py

app.py

@@ -34,7 +34,9 @@ model.eval()
 
 def transcribe(audio):
     sample_rate, waveform = audio
-    waveform = torch.from_numpy(waveform[:, 0]).float().unsqueeze_(0)
+    if len(waveform.shape) == 2: 
+        waveform = waveform[:, 0]
+    waveform = torch.from_numpy(waveform).float().unsqueeze_(0)
     waveform = torchaudio.functional.resample(waveform, sample_rate, 16_000)
 
     transcript = model.predict(waveform)[0]