1、导入相关包
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel
2、加载基础模型
model = AutoModelForCausalLM.from_pretrained("Langboat/bloom-1b4-zh")
tokenizer = AutoTokenizer.from_pretrained("Langboat/bloom-1b4-zh")
3、加载Lora模型
p_model = PeftModel.from_pretrained(model, model_id="./chatbot/checkpoint-500/")
4、模型推理
ipt = tokenizer("Human: {}\n{}".format("考试有哪些技巧？", "").strip() + "\n\nAssistant: ", return_tensors="pt")
tokenizer.decode(p_model.generate(**ipt, do_sample=False)[0], skip_special_tokens=True)
5、模型合并
merge_model = p_model.merge_and_unload()
模型推理
ipt = tokenizer("Human: {}\n{}".format("考试有哪些技巧？", "").strip() + "\n\nAssistant: ", return_tensors="pt")
tokenizer.decode(merge_model.generate(**ipt, do_sample=False)[0], skip_special_tokens=True)
6、模型保存
merge_model.save_pretrained("./chatbot/merge_model")
