Update app.py

app.py

@@ -55,7 +55,7 @@ def compress_history(formatted_prompt):
 
     generate_kwargs = dict(
         temperature=temperature,
-        max_new_tokens=30480,
+        max_new_tokens=10480,
         top_p=0.95,
         repetition_penalty=1.0,
         do_sample=True,
@@ -222,7 +222,7 @@ def generate(prompt, history, agent_name=agents[0], sys_prompt="", temperature=0
                 prompt=prompt.strip(' \"')
             
             formatted_prompt = format_prompt(f"{system_prompt}, {prompt}", history)
-        if len(formatted_prompt) < (50000):
+        if len(formatted_prompt) < (40000):
             print(len(formatted_prompt))