ruidan / IMN-E2E-ABSA

Code and dataset for ACL2019 ‘‘An Interactive Multi-Task Learning Network for End-to-End Aspect-Based Sentiment Analysis’’.
Apache License 2.0
99 stars 18 forks source link

请问您的结果是怎么取的 #11

Open dongyang12138 opened 4 years ago

dongyang12138 commented 4 years ago

请问您的结果是取的是测试集loss最小值时候的结果还是取的F1-I最高的时候的结果? 还是其他的取值方式,或者多次试验取平均?

dongyang12138 commented 4 years ago

看您的源代码,save_model是按照验证集f_absa取的最高值时进行的,请问为什么不可以按照测试集f值取最高时取值?

bigredbug47 commented 3 years ago

Hi @dongyang12138 , have you successfully reproduce the source code with python3 version and get same result as author paper?

Could you kindly provide me which elements that you have to updated when convert source code from python2 to python3?

Thank you a lot.