Update app.py

app.py

@@ -133,8 +133,8 @@ sys.path.append(model_path_d_local)
 # --- Model Loading ---
 
 # Constants for text generation
-MAX_MAX_NEW_TOKENS = 5120
-DEFAULT_MAX_NEW_TOKENS = 3072
+MAX_MAX_NEW_TOKENS = 4096
+DEFAULT_MAX_NEW_TOKENS = 2048
 MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")