Update app.py

app.py

@@ -8,8 +8,8 @@ import gradio as gr
 import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
 
-MAX_MAX_NEW_TOKENS = 2048
-DEFAULT_MAX_NEW_TOKENS = 1024
+MAX_MAX_NEW_TOKENS = 1024
+DEFAULT_MAX_NEW_TOKENS = 512
 MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "8192"))
 
 if not torch.cuda.is_available():
@@ -20,11 +20,11 @@ if torch.cuda.is_available():
     tokenizer = AutoTokenizer.from_pretrained(model_id)
     model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto")
 
-@spaces.GPU
+@spaces.GPU(duration=30)
 def generate(
     message: str,
     chat_history: list[tuple[str, str]],
-    max_new_tokens: int = 1024,
+    max_new_tokens: int = 512,
     temperature: float = 0.06,
     top_p: float = 0.95,
     top_k: int = 40,