wrap automodel for zerogpu

langchainapp.py

@@ -43,7 +43,7 @@ device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 ### Utils  
 hf_token, yi_token = load_env_variables()  
 
-tokenizer = AutoTokenizer.from_pretrained(model_name, token=hf_token, trust_remote_code=True)  
+# tokenizer = AutoTokenizer.from_pretrained(model_name, token=hf_token, trust_remote_code=True)  
 # Lazy load model  
 model = None