solve custom code issue with workaround

langchainapp.py

@@ -40,7 +40,8 @@ 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()  
   
@@ -61,7 +62,7 @@ class MyEmbeddingFunction(EmbeddingFunction):
         self.intention_client = intention_client  
         self.hf_embeddings = HuggingFaceInstructEmbeddings(  
             model_name=model_name,  
-            model_kwargs={'device': 'cuda' if torch.cuda.is_available() else 'cpu', "trust_remote_code": True},  
+            model_kwargs={'device': 'cuda' if torch.cuda.is_available() else 'cpu'}, #, "trust_remote_code": True},  
             encode_kwargs={'normalize_embeddings': True}  
         )  
   

requirements.txt

@@ -1,6 +1,6 @@
 huggingface_hub
 einops
-sentence-transformers===2.3.1
+sentence-transformers
 torch==2.2.0
 transformers
 openai