-
Notifications
You must be signed in to change notification settings - Fork 150
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
I find a bug in your train_log #5
Comments
Hello, The source of the domain-adapted BERT model in this repository has been declared in the README.md, |
sorry,this is my fault, very sorry and pleasure your reply. |
你好,可以分享一下你的训练日志吗? |
当然可以,谢谢你帮我看问题所在。
在这之后,就没有任何提升了,锁死在了max_acc:76.3 f1:75.22,我想要的是更高的点。 |
从日志来看, 如果不是这个问题,可以再联系我。 |
另外,使用 BERT-ADA存储库中提供的 joint domain-adapted BERT可以在Twitter数据集上取得78.3+的acc(使用LCF-net模型,该模型已经被我废弃),你可以使用批量训练脚本多测试几次,一般3-5次就可以达到最有效果。 |
非常感谢你的回答。确实是shell的指定有问题,附上原因一份:https://cloud.tencent.com/developer/ask/188470 但在更正了代码后,依然达不到paper的效果。下面附上训练日志:
|
你好,抱歉回复晚了。这几天才有时间测试代码,发现现在模型确实很难在twitter上达到77+的acc了(最高76.88),推测是代码重构和迁移到pytorh-transfomers库导致的,你可以使用自己实际得出的结果而不是paper上的结果。后面如果有时间我会再检查问题并更新repo,LCF-net模型确实可以在几个数据集上达到比较好的成绩,但是存在暂时无法解决的问题,所以暂时不会开源该模型。 |
非常感谢你的回答。我觉得你做出了了不起的工作。谢谢你。 |
In your log ,we can see ,you use a
I dont know where this pretrained_bert come from
and
you use restaurant bert_pretrained_models in twitter dataset , I think this is not Ok
can you explain and fix this ?
The text was updated successfully, but these errors were encountered: