Update app.py

app.py

@@ -26,11 +26,11 @@ model = AutoModelForCausalLM.from_pretrained(model_id,
                                              # load_in_8bit=True,
                                              quantization_config=nf4_config,
                                              torch_dtype = torch.bfloat16,
-                                             # device_map="auto"
+                                             device_map="auto"
                                             )
 
 # replace_linears_in_hf(model)
-model.to('cuda').eval()
+model.eval()
 @spaces.GPU
 def generate_response(user_input, max_new_tokens, temperature):
     os.system("nvidia-smi")