在线ps_品牌建设论文怎么写_seo是什么岗位_seo搜索引擎优化是什么

在复现一个项目时，原项目使用的是英文词汇表，换成中文的bert模型之后报错：untimeError: Error(s) in loading state_dict for BertForMaskedLM:
        size mismatch for bert.embeddings.word_embeddings.weight: copying a param with shape torch.Size([30522, 768]) from checkpoint, the shape in current model is torch.Size([21128, 768]).
        size mismatch for cls.predictions.bias: copying a param with shape torch.Size([30522]) from checkpoint, the shape in current model is torch.Size([21128]).
        size mismatch for cls.predictions.decoder.weight: copying a param with shape torch.Size([30522, 768]) from checkpoint, the shape in current model is torch.Size([21128, 768]).
        size mismatch for cls.predictions.decoder.bias: copying a param with shape torch.Size([30522]) from checkpoint, the shape in current model is torch.Size([21128]).

根据错误提示可以看到就是vocab_size不一致。 

解决方案：

第一步：打开本地模型路径下的config.jason文件，将里面的"vocab_size"改为30522。改完之后报错如下：

第二步：根据错误提示，在from_pretrained方法中加入ignore_mismatched_sizes=True

重新运行程序，能跑了！！！