Update app.py

app.py

@@ -178,10 +178,10 @@ def respond(
         streamer=streamer,
         do_sample=True,
         temperature=temperature,
-        max_new_tokens=max_new_tokens,
-        top_k=top_k,
-        repetition_penalty=repetition_penalty,
-        top_p=top_p
+        max_new_tokens=max_tokens,
+        top_k=40,
+        repetition_penalty=1.1,
+        top_p=0.95
     )
     t = Thread(target=model.generate, kwargs=generate_kwargs)
     t.start()