Update app.py

app.py

@@ -19,7 +19,7 @@ def response(instruction, history, inputText):
 ### Response:"""
 
     input_ids = tokenizer(inp, return_tensors='pt').to("cuda")
-    beam_output = model.generate(**input_ids)
+    beam_output = model.generate(**input_ids, max_new_tokens=300)
     outputText = tokenizer.decode(beam_output[0], skip_special_token=True)
     #output = output.replace(inp,"").replace("<s>","").replace("</s>","")