zjunlp/DocuNet

The BC5CDR dataset result

Closed this issue · 2 comments

Hello!
When I run your code with cdr dataset,get the result like this
image
Do you know why it's so high compared to your paper result and previous works!

Looking forward for your reply!
Thanks~

zxlzr commented

Hi, after the paper submission, we tune the hyper-parameters (more training steps, different learning rate, larger batch sizes, and so on) and also obtain better results on BC5CDR; so it is ok to obtain better results.

Hi, after the paper submission, we tune the hyper-parameters (more training steps, different learning rate, larger batch sizes, and so on) and also obtain better results on BC5CDR; so it is ok to obtain better results.

OK, thank you very much!