Evaluation metrics don't match

[MargaridaMCampos]

Hello team,

I tried to reproduce the code for the Yes/No Question & Answering with simple BioBert and the results don't match those reported in the READ_ME. Have there been any changes?
Below are the results I obtained. Just wanted to check if everything is okay.

Thank you in advance,

Margarida

[wonjininfo]

Hi,
Thank you for your interest in our paper. Did you use BioBERT v1.1 - SQuAD v2.0 as the initial checkpoint?
Reported scores for YesNo questions are acquired from the BioBERT v1.1 - SQuAD v2.0 checkpoint.
Best,
Wonjin

[MargaridaMCampos]

Hi, I’ve been trying to run tour code with the referred checkpoints - for college research papear - but I’ve been unable.. Do you think it would be possible to have a quick Zoom call or something? Thank you so much in advance. I’m a big admirer of your work. Best, Margarida

…

No dia 27/10/2020, às 15:08, Wonjin Yoon ***@***.***> escreveu:  Hi, Thank you for your interest in our paper. Did you use BioBERT v1.1 - SQuAD v2.0 as the initial checkpoint? Reported scores for YesNo questions are acquired from the BioBERT v1.1 - SQuAD v2.0 checkpoint. Best, Wonjin — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.