vdobrovolskii / wl-coref

This repository contains the code for EMNLP-2021 paper "Word-Level Coreference Resolution"

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

about the selection of the best model

leileilin opened this issue · comments

Hello, I think the code of saving the model is only saved every epoch, and the optimal model is not saved. Will it be reduced when evaluating the performance of the test set?
thanks.

What do you mean, the optimal model is not saved? All the models are saved, and the logs are there, so you are free to choose the one whose metrics you like

What do you mean, the optimal model is not saved? All the models are saved, and the logs are there, so you are free to choose the one whose metrics you like

Yeah, of course, you can select the model by checking the log, but I think it is not as convenient as saving the optimal model.

Pull requests with that functionality are most welcome :)

Pull requests with that functionality are most welcome :)

i test with different models, the experimental results are basically similar.