remove the device_map to omit the accelerate error

app.py

@@ -100,7 +100,8 @@ def load_llm_model():
     #this one is for running with GPT
     llm = HuggingFacePipeline.from_model_id(model_id= 'lmsys/fastchat-t5-3b-v1.0', 
                                             task= 'text2text-generation',
-                                            model_kwargs={ "device_map": "auto",
+                                            model_kwargs={ 
+                                                # "device_map": "auto",
                                                         "max_length": 256, "temperature": 0,
                                                         "repetition_penalty": 1.5,
                                                          "quantization_config": bitsandbyte_config}) #add this quantization config