max new token

app.py

@@ -1,6 +1,6 @@
 import gradio as gr
 from transformers import pipeline
 
-pipe = pipeline("text-generation", model="tpzeng/prm800k_llama_joint_checkpoint1000", bad_words_ids=[[128254],[128255]], device='cuda:0')
+pipe = pipeline("text-generation", model="tpzeng/prm800k_llama_joint_checkpoint1000", bad_words_ids=[[128254],[128255]],  max_new_tokens = 1024)
 
 gr.Interface.from_pipeline(pipe).launch()