Update app.py

app.py

@@ -43,11 +43,7 @@ def inference(audio):
     with torch.no_grad():
       logits = model(input_values).logits
 
-    #pred_ids = torch.argmax(logits, dim=-1)
-    h = logits.numpy()[0,:,:]
-    v = np.pad(h, [0, 2], mode='constant')
-
-    output = processor.decode(v).text
+    output = processor.decode(logits.numpy()[0]).text
 
     return output[:-4]