Fix audio pre-processing

app.py

@@ -29,13 +29,15 @@ def greet(image, audio):
 
     audio_file = torch.from_numpy(audio)
 
+    if len(audio_file.shape) == 2:
+        audio_file = torch.concat([audio_file[:, 0:1], audio_file[:, 1:2]], dim=0).T  # Stereo -> mono (x2 duration)
+    else:
+        audio_file = audio_file.unsqueeze(0)
+
     if desired_sample_rate != sample_rate:
         audio_file = torchaudio.functional.resample(audio_file, sample_rate, desired_sample_rate)
 
-    if audio_file.shape[0] == 2:
-        audio_file = torch.concat([audio_file[0], audio_file[1]], dim=0)  # Stereo -> mono (x2 duration)
-
-    audio_file.squeeze(0)
+    audio_file = audio_file.squeeze(0)
 
     if audio_file.shape[0] > (desired_sample_rate * set_length):
         audio_file = audio_file[:desired_sample_rate * set_length]