Open JunweiSUN opened 5 years ago
Thanks for your advice. We will upload the English pre-trained models today. Yes, we have to convert bert-tf model to huggingface format, then convert huggingface format to uer format. We have done these works and test UER on GLUE benchmark.
Thanks for your advice. We will upload the English pre-trained models today. Yes, we have to convert bert-tf model to huggingface format, then convert huggingface format to uer format. We have done these works and test UER on GLUE benchmark.
多谢您的回复,还有一个问题是您的tagger.py针对的是中文,如果我想用uncased模型的话是不是应该自己做大小写转换等工作呢?
I have added the links of English pre-trained models. Please see Model Zoo section. We have tested UER on GLUE benchmark and it should match the performance of HuggingFace's implementation. You need to do lowercase if you use uncased model. But I think cased model could be better choice for NER.
Embedding, thanks for this great work,
When I try to download the English pre-trained models, it said "(190049)内容可能违反相关法规", is it possible to put your model in an accessible place? Thanks a lot!
您好!首先十分感谢贵团队的工作,这项工作很有意义。 我想要将这个项目应用于英文文本上的标注任务(具体来说是NEG、CHUNKING和POS),但是我没有找到使用英文语料的预训练模型,请问这个项目有提供相关的.bin文件吗? 如果我自己生成.bin模型的话,是不是需要先用huggingface提供的脚本将tensorflow版本的预训练模型转换到huggingface版本,再用你们提供的脚本将huggingface的版本转换为本项目的.bin版本,最后使用该模型来训练下游任务(使用google提供的英文词表)呢?
期待您的回复!