Open chensongcan opened 2 years ago
Were the published results achieved using the bert-base or bert-large model?
Were the published results achieved using the bert-base or bert-large model?
look at the appendix
ok, thanks for your reply. Please reply the first problem, when train 50 epochs, the F1 score is still only achieved 90.5
use my dataset the f1 only reach 30%,what can i do to improve the f1?
When batch size =30, other parameters are the same as yours, F1 score is only 90.5 in NYT-star test datasets. It does not reach the 92.6 mentioned in the paper.