huggingface model模型本地化

以t5-base为例：

 

储存模型：

from transformers import AutoModelForSeq2SeqLM
if __name__ == '__main__':
    model_name = 't5-base'
    prefix = 'your_path'
    save_path = prefix+model_name

    model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
    model.save_pretrained(save_path)

 

加载模型：

model = AutoModelForSeq2SeqLM.from_pretrained(save_path)