solve custom code issue with workaround

langchainapp.py

@@ -40,11 +40,11 @@ os.environ['CUDA_LAUNCH_BLOCKING'] = '1'
 os.environ['CUDA_CACHE_DISABLE'] = '1'  
   
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")  
-tokenizer = AutoTokenizer.from_pretrained(model_name, token=token, trust_remote_code=True)  
-nvidiamodel = AutoModel.from_pretrained(model_name, token=token, trust_remote_code=True).to(device)  
 ### Utils  
 hf_token, yi_token = load_env_variables()  
-  
+
+tokenizer = AutoTokenizer.from_pretrained(model_name, token=hf_token, trust_remote_code=True)  
+nvidiamodel = AutoModel.from_pretrained(model_name, token=hf_token, trust_remote_code=True).to(device)    
 def clear_cuda_cache():  
     torch.cuda.empty_cache()