Allan Jie

icon indicating a website link https://allanj.github.io icon indicating an email [email protected]

icon company Salesforce Research icon location Singapore icon location NLP Scientist at AI Lab, ByteDance, Singapore. Ph.D. from SUTD. Working on Semantic parsing, Dependency parsing and Named Entity Recognition

Results 72 comments of Allan Jie

Do you have check points for MAWPS?

what's the error that you are having?

Is there any way to use your pre-trained model?

some of the models are already in the README, how I can help you? Can you describe more detail?

Is there any way to use your pre-trained model?

1. Yes. I think so 2. when it stop, did you see the log? Not sure if the GPU memory is enough 3. Would be appreciated if you can share...

open english model

I totally agree with you. Though many papers produce dependency parsing results. I think it is better to open source the trained model. I think many people now are still...

相同输入下结果不同

雾草。。确实，每一次call都不一样

相同输入下结果不同

> 可能是inference的时候，dropout参数没有重置为0 这就尴尬了

相同输入下结果不同

想问一下大家。不同的输出结果对你们有影响吗 ----- 后来发现其实官方的elmo也是会变化的。

相同输入下结果不同

其实还好。我感觉对我NER没什么太大影响

Task specific combination of the intermediate layer representations

I think they output the weights for all the layers.

Is there a way to output not only crf 1 best, but 5 best predictions?

I'm currently on holiday until 10 Sept and will try to incorporate this function into this repo. But in the meanwhile, you can check out this repo: https://github.com/jiesutd/NCRFpp/blob/master/model/crf.py#L268-L418 and see...