Update exist when no cuda present

app.py

@@ -3,6 +3,7 @@ import gc
 import os
 import time
 
+import gradio
 import rich
 import torch
 from huggingface_hub import snapshot_download
@@ -11,6 +12,10 @@ from transformers import AutoModelForCausalLM, AutoTokenizer
 from transformers.generation.utils import GenerationConfig
 
 model_name = "baichuan-inc/Baichuan2-13B-Chat-4bits"
+if not torch.cuda.is_available():
+    gradio.Error(f"No cuda, cant run {model_name}")
+    raise SystemError(1)
+
 # snapshot_download?
 loc = snapshot_download(repo_id=model_name, local_dir="model")