increase temp, add top_k=20

handler.py

@@ -29,7 +29,8 @@ class EndpointHandler():
             max_new_tokens=50, 
             return_full_text=False, 
             bad_words_ids=[[3070]], 
-            temperature=0.7
+            temperature=0.8,
+            top_k=20,
         )
         return prediction