【peft】用peft调大模型，加载checkpoint失败，报ValueError: Can‘t find a valid checkpoint at

接上文，用peft的lora训练bloomz，训练1轮之后，发现可能不太充分，打算加载检查点继续训练，代码如下：

trainer.train(resume_from_checkpoint = 'checkpoint目录')

然后报错：raise ValueError(f"Can't find a valid checkpoint at {resume_from_checkpoint}")
ValueError: Can't find a valid checkpoint at checkpoint目录

参考Peft Model not resuming from Checkpoint · Issue #24252 · huggingface/transformers · GitHub

就是_load_from_checkpoint有点问题

解决，新建一个Trainer子类，子类里重写了加载检查点的函数，调用时用这个子类来创建trainer对象

from transformers import Trainer
import os
from peft import PeftModel
from transformers.utils import (
    ADAPTER_SAFE_WEIGHTS_NAME,
    ADAPTER_WEIGHTS_NAME,
    is_sagemaker_mp_enabl