Update configs.py

configs.py

@@ -22,6 +22,7 @@ model_info = {
                       layers_format=llama_layers_format),
     'LLAMA2-13B': dict(model_path='meta-llama/Llama-2-13b-chat-hf', 
                       token=os.environ['hf_token'], torch_dtype=torch.float16,
+                      wait_with_hidden_states=True,
                     # device_map='auto', max_memory={0: "15GB", 1: "30GB"},  dont_cuda=True, # load_in_8bit=True, 
                       original_prompt_template='<s>{prompt}',
                       interpretation_prompt_template='<s>[INST] [X] [/INST] {prompt}',