AdaSeq icon indicating copy to clipboard operation
AdaSeq copied to clipboard

[Feature] pipleline NER返回结果问题

Open tianchiguaixia opened this issue 2 years ago • 7 comments

Is your feature request related to a problem?

1689576213474_B0235B15-0449-4ecc-8584-2540D9826C2F 按照transformers里面的piplelines如果输入超过512时候,也会返回大于512的位置,而不是报错

Describe the solution you'd like.

No response

Describe alternatives you've considered.

No response

Additional context.

No response

Code of Conduct

  • [X] I agree to follow this project's Code of Conduct

tianchiguaixia avatar Jul 17 '23 09:07 tianchiguaixia

还有paddlenlp里面的taskflow也是,超出512也会返回大于512以后的位置。希望可以修改一下后处理代码

tianchiguaixia avatar Jul 17 '23 11:07 tianchiguaixia

请问一下,训练好的模型在用这个方法调用的时候,一直出现如下报错: KeyError: "SequenceLabelingPipeline: 'text-classification is not in the models registry group token-classification. Please make sure the correct version of ModelScope library is used.'" 以及 KeyError: 'text-classification is not in the models registry group token-classification. Please make sure the correct version of ModelScope library is used.' 我目前使用的modelscope是1.9.2,一直没搞懂错在哪里,请问你有遇到过类似的报错吗

cuberJ avatar Oct 12 '23 15:10 cuberJ

模型地址传错了

tianchiguaixia avatar Oct 13 '23 01:10 tianchiguaixia

不好意思,我没太理解您的意思。能否再向您请教一下:

  1. 您说的这个模型地址出错,是指pipeline中的调用出现错误了吗?
  2. 这种错误是发生在训练的过程中,还是在执行预测的过程中产生的呢?
  3. 如果是训练的过程中出现的问题,需要修改的是yaml配置文件,还是源码本身呢

cuberJ avatar Oct 13 '23 02:10 cuberJ

补充一下: 我目前调用的任务是做named_entity_recognition, 但是在调试追踪的时候发现,pipeline里总是会调用到text_classification的步骤,用作者官方提供的训练配置训练后,也会出现这种错误的调用链

cuberJ avatar Oct 13 '23 03:10 cuberJ

你看官方的,你的可能老版本,传参问题

tianchiguaixia avatar Oct 16 '23 02:10 tianchiguaixia

也是NER任务,自己微调后,遇到了与三楼同样的问题 KeyError: "SequenceLabelingPipeline: 'text-classification is not in the models registry group token-classification. Please make sure the correct version of ModelScope library is used.'"

我的modelscope版本是1.9.3

TianyiLu avatar Nov 24 '23 08:11 TianyiLu