rjk-git / bert2transformer_on_nmt Goto Github PK

能提供自定义语料训练bert的程序吗？我看代码中的bert版本具有局限性。

我搭了个环境，缺什么包就去下载什么包，最后硬是倒在了这个包上面。

报错如下:
FileNotFoundError: [Errno 2] No such file or directory: 'C:\Users\Administrator\AppData\Roaming\Python\Python36\site-packages\gluonnlp-1.0.0.dev20210115-py3.6.egg\gluonnlp\models\model_zoo_checksums\albert.txt'

那这个问题，我感觉就是gluonnlp的api有变动。

一句话，这个仓库没有把依赖说明清楚，价值不是很大。

如有成功跑出来的请指教。

您好，请问能不能发一下依赖环境，我跑cpu的可以，跑gpu的一直报错

Why do you use the pre-trained word embedding of bert as the output of transformer?
Actually, I want to make the the pre-trained word embedding of bert as the input of transformer. In this way, the context embedding processed by the transformer as an encoder can be used as others applications.

有无成功跑过的大佬告知这个环境配置应该是个怎么个样子的情况呢。。是tensorflow还是pytorch呢？万分感谢。

Hello, after I used about 10 million training parallel to the database, I didn‘t tokenizer chinese(未分词), converted the English to lower case, and run the translate.py file. No matter what the translation results are, the same translation results. Can you write an operation document (with training sample format included)，thanks