Update configs.py

configs.py

@@ -19,9 +19,9 @@ model_info = {
                       interpretation_prompt_template='<s>[INST] [X] [/INST] {prompt}',
                       layers_format=llama_layers_format),
     'LLAMA2-13B': dict(model_path='meta-llama/Llama-2-13b-chat-hf', 
-                      token=os.environ['hf_token'], device_map='auto', 
-                       max_memory={0: "15GB", 1: "30GB"}, # load_in_8bit=True, 
-                      original_prompt_template='<s>{prompt}', dont_cuda=True,
+                      token=os.environ['hf_token'], torch_dtype=torch.float16,
+                    # 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}',
                       layers_format=llama_layers_format), 
     'GPT-2 Small': dict(model_path='gpt2', original_prompt_template='{prompt}',