自动加载BERT预训练模型
代码如下:
from transformers import (AutoModelForMaskedLM,
AutoTokenizer, LineByLineTextDataset,
DataCollatorForLanguageModeling,
Trainer, TrainingArguments)
# 自动加载中文预训练模型(另一种加载方法)
model_name = 'bert-base-chinese'
model = AutoModelForMaskedLM.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)
tokenizer.save_pretrained(pretrain_path) # 保存词典等文件