Update app.py

app.py

@@ -9,7 +9,7 @@ from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStream
 
 MAX_MAX_NEW_TOKENS = 100
 DEFAULT_MAX_NEW_TOKENS = 20
-MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
+MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "200"))
 
 DESCRIPTION = """\
 # Löwolf Chat
@@ -37,7 +37,7 @@ def generate(
     message: str,
     chat_history: list[tuple[str, str]],
     system_prompt: str,
-    max_new_tokens: int = 1024,
+    max_new_tokens: int = 50,
     temperature: float = 0.6,
     top_p: float = 0.9,
     top_k: int = 50,