Update app.py

app.py

@@ -144,7 +144,7 @@ def model_inference(input_dict, history):
         padding=True,
     ).to("cuda")
     streamer = TextIteratorStreamer(processor, skip_prompt=True, skip_special_tokens=True)
-    generation_kwargs = dict(inputs, streamer=streamer, max_new_tokens=4096)
+    generation_kwargs = dict(inputs, streamer=streamer, max_new_tokens=1024)
     thread = Thread(target=model.generate, kwargs=generation_kwargs)
     thread.start()
     buffer = ""