Update configs.py

configs.py

@@ -20,7 +20,7 @@ model_info = {
                       layers_format=llama_layers_format),
     'LLAMA2-13B': dict(model_path='meta-llama/Llama-2-13b-chat-hf', 
                       token=os.environ['hf_token'], device_map='auto', 
-                       max_memory={0: "10GB", 1: "10GB"}, # load_in_8bit=True, 
+                       max_memory={0: "15GB", 1: "30GB"}, # load_in_8bit=True, 
                       original_prompt_template='<s>{prompt}', dont_cuda=True,
                       interpretation_prompt_template='<s>[INST] [X] [/INST] {prompt}',
                       layers_format=llama_layers_format),