transformers quantization bitsandbytes实时量化方法使用load_in_4bit，load_in_8bit；量化模型保存及加载使用

参考：
https://huggingface.co/docs/transformers/main/en/quantization/bitsandbytes?bnb=8-bit
https://www.atyun.com/57101.html
https://huggingface.co/blog/4bit-transformers-bitsandbytes

bitsandbytes错误参考：
https://cnloong.blog.csdn.net/article/details/141607933

代码

使用qwen2-7b

1）load_in_4bit

from transformers import AutoTokenizer,AutoModelForCausalLM,BitsAndBytesConfig

quantization_config = BitsAndB