from transformers import AutoTokenizer, AutoModelForCausalLM import torch model_name_or_path = 'ckiplab/CKIP-Llama-2-7b-chat' tokenizer = AutoTokenizer.from_pretrained(model_name_or_path) # int8 model = AutoModelForCausalLM.from_pretrained(model_name_or_path, device_map='auto', torch_dtype...