Questions about pre-trained models and prediction results in the paper

[swb19]

Hello, your work is very helpful to me. However, I have some questions about the prediction accuracy achieved in the paper, when I used the pre-trained model you provided for evaluation, I found that its prediction accuracy is lower than that mentioned in Table 4 in the paper.
Could you please provide relevant information about your training parameters or pre-trained model at that time?

[BorisIvanovic]

Hi @swb19, thanks for your question! Have you seen this issue? Those results were achieved by someone else in the community running this codebase, and are more accurate to compare to with regards to the code in the current master branch.