-
paddlenlp支撑的预训练模型以及相关下游任务, 迫切需要可解释性作为工业落地的另一个支撑, 比如基于预训练的多分类模型, paddle是否提供了分类的可解释性, 看到https://github.com/PaddlePaddle/InterpretDL 似乎是这样一个目标,但是与paddlenlp并没有集成, 请给与一些建议,谢谢! |
Beta Was this translation helpful? Give feedback.
Answered by
LangDaoAI
Aug 4, 2021
Replies: 2 comments
-
刚刚调查了一下,使用LIME去做分类的可解释性应该是可以的, LIME并关注内部的模型是否是基于预训练,marcotcr/lime#356 |
Beta Was this translation helpful? Give feedback.
0 replies
-
再次看了一下interpretdl, https://github.com/PaddlePaddle/InterpretDL/blob/master/tutorials/lime_tutorial_nlp_ERNIE.ipynb 应该解决了我的问题 |
Beta Was this translation helpful? Give feedback.
0 replies
Answer selected by
ZeyuChen
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
再次看了一下interpretdl, https://github.com/PaddlePaddle/InterpretDL/blob/master/tutorials/lime_tutorial_nlp_ERNIE.ipynb 应该解决了我的问题