Increase `num_beams`

app.py

@@ -12,7 +12,7 @@ def predict(text):
         attention_mask=tokenized_inputs["attention_mask"],
         max_new_tokens=256,
         length_penalty=0.8,
-        num_beams=2,
+        num_beams=4,
         do_sample=True
     )
     summary = tokenizer.batch_decode(output, skip_special_tokens=True)[0]