Update app.py

app.py

@@ -91,19 +91,20 @@ def run_interpretation(global_state, raw_interpretation_prompt, max_new_tokens,
 
 ## main
 torch.set_grad_enabled(False)
-model_name = 'meta-llama/Llama-2-7b-chat-hf' # 'mistralai/Mistral-7B-Instruct-v0.2' #
+model_name = 'LLAMA2-7B'
 
 # extract model info
 model_args = deepcopy(model_info[model_name])
+model_path = model_args.pop('model_path')
 original_prompt_template = model_args.pop('original_prompt_template')
 interpretation_prompt_template = model_args.pop('interpretation_prompt_template')
-tokenizer_name = model_args.pop('tokenizer') if 'tokenizer' in model_args else model_name
+tokenizer_path = model_args.pop('tokenizer') if 'tokenizer' in model_args else model_path
 use_ctransformers = model_args.pop('ctransformers', False)
 AutoModelClass = CAutoModelForCausalLM if use_ctransformers else AutoModelForCausalLM
 
 # get model
-model = AutoModelClass.from_pretrained(model_name, **model_args)
-tokenizer = AutoTokenizer.from_pretrained(tokenizer_name, token=os.environ['hf_token'])
+model = AutoModelClass.from_pretrained(model_path, **model_args)
+tokenizer = AutoTokenizer.from_pretrained(tokenizer_path, token=os.environ['hf_token'])
 
 # demo
 json_output = gr.JSON()