AlibabaResearch / AlibabaResearch/efficientteacher

1、请问为何LabelMatch的更新阈值方法没有被启用？2、cls得分高但obj得分低为啥没有被送往去算obj loss？

Open

#57 opened on Mar 30, 2023 3 comments 0 reactions 0 assignees View on GitHub

Dominant language: Python
Stars: 813
Forks: 126
PR merge metrics: No merged PRs in 30d

Description

![image](https://user-images.githubusercontent.com/110707902/228746570-0e535142-c460-4891-889d-b9cf241b71af.png)
没有用论文中的方法，而是采用高斯混合模型计算阈值

![QQ截图20230329145203](https://user-images.githubusercontent.com/110707902/228746966-dec1e7cb-604b-4bd7-8d5a-89102a69dee8.png)
![QQ截图20230329145317](https://user-images.githubusercontent.com/110707902/228747028-9c992681-de1c-4bb1-afef-1cbfff5c0377.png)
![QQ截图20230329145532](https://user-images.githubusercontent.com/110707902/228747044-a2e6ddf8-9cdb-4089-b82c-ca26f084f6c7.png)
![image](https://user-images.githubusercontent.com/110707902/228747149-e66adc6b-88f7-47f4-9f21-79b4393cba55.png)
这里cls得分超过0.99，但pseudo得分低于高阈值的样本没有被处理，请问这些样本需要进一步来挖掘吗？

Contributor guide

No contributing guide indexed for this repository

Assessment

This issue has not been assessed yet.

1、请问为何LabelMatch的更新阈值方法没有被启用？2、cls得分高但obj得分低为啥没有被送往去算obj loss？

Description

Contributor guide

Assessment

Get new issues in your inbox